Decision-Making

Information & Management 53 (2016) 767–786

Contents lists available at ScienceDirect

Information & Management

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / i m

Process innovation as creative problem solving: An experimental

study of textual descriptions and diagrams

Kathrin Figl a,*, Jan Recker b,1

a WU – Vienna University of Economics and Business, Institute for Information Systems & New Media, Welthandelsplatz 1, Building D2, 1020 Vienna, Austria b Queensland University of Technology, Information Systems Discipline, Office 510/126 Margaret Street, Brisbane, QLD 4000, Australia

A R T I C L E I N F O

Article history:

Received 11 March 2014

Received in revised form 23 February 2016

Accepted 26 February 2016

Available online 4 March 2016

Keywords:

Process innovation

Business process models

Business process reengineering

Creative problem solving

Diagrams

A B S T R A C T

The use of process models to support business analysts’ idea-generation tasks has been a long-standing

topic of interest in process improvement. We examine how two types of representations of

organizational processes – textual and diagrammatic – assist analysts in developing innovative

solutions to process-redesign tasks. The results of our study clarify the types of process-redesign ideas

generated by analysts who work with text versus those who work with models. We find that the volume

and originality of process-redesign ideas do not differ significantly but that appropriateness of ideas

varies. We discuss the implications of these findings for research and practice in process improvement.

� 2016 Elsevier B.V. All rights reserved.

1. Introduction

When analyzing and/or designing information systems, analysts frequently use process models to document and analyze current organizational operations. These models help business personnel understand the work domain and identify opportu- nities in the improvement of business processes and related information systems [35]. This exercise typically involves developing process models that capture the current organiza- tional reality and then giving them to analysts in the hope that the models will stimulate creative ideas about how the processes can be improved. However, whether process models actually assist analysts in their idea generation tasks, that is, in finding innovative solutions for future processes, or limit them to narrow ways of thinking remains in doubt. This question is far from trivial. For example, some claim that process modeling focuses on the shortcomings of an existing solution, so model-based process innovation aims on overcoming existing problems rather than achieving inspirational new goals [69]. Others suggest that good process models can be an important determinant in process- improvement success [41].

* Corresponding author. Tel.: +43 650 979 48 89; fax: +43 1 31336 90 4467.

E-mail addresses: [email protected] (K. Figl), [email protected] (J. Recker). 1 Tel.: +61 7 3138 9479; fax: +61 7 3138 9390.

http://dx.doi.org/10.1016/j.im.2016.02.008

0378-7206/� 2016 Elsevier B.V. All rights reserved.

We study whether and how various ways of modeling organizational processes aid process innovation. We conceptualize process innovation as creative problem solving, where analysts generate appropriate and original ideas on how processes could be redesigned. We draw on problem-solving and visual representa- tion theory (e.g., Refs. [26,43,88,91,92]) to hypothesize how textual and diagrammatic process models affect the creativity and type of redesign solutions. Subsequently, we report an experiment in which we tested our hypotheses.

Our study contributes to the extant literature in three primary ways. First, it adds to the body of knowledge on the use of process modeling in practice. The literature to date tends to explain how analysts understand visual models of organizational systems and processes (e.g., Refs. [53,67]) but not how the use of such models may influence the type and creative quality of ideas in process- redesign initiatives. However, input to process redesign remains the main outcome expected from process modeling [35]. Second, our study contributes to the literature on process redesign [66,81] by evaluating the types of creative solutions analysts generate by working with various types of process models. Third, we offer a new methodology for evaluating process-redesign ideas in terms of their originality, appropriateness, impact, and locus of change.

For industry, our study provides an answer to two deceptively simple questions: Do the outcomes of process redesigns vary with the process models analysts use? What type of representation

K. Figl, J. Recker / Information & Management 53 (2016) 767–786768

format should analysts use based on the objective of process improvement?

2. Background

Our study relates to three streams in the literature: (i) how business process redesign is conducted and how creative problem solving is part of these efforts, (ii) how information about organizational processes can be represented, and (iii) how process representations can act as stimuli for creative redesign. We discuss each stream in turn.

2.1. Business process redesign and creative problem solving

Business processes are sets of logically related organizational tasks that are performed to achieve defined business outcomes [17]. Organizations often document their business processes in order to understand where weaknesses and performance deficien- cies in processes manifest and to generate ideas about how new processes, supported by existing or future information systems, could be enacted.

Process innovation projects tend to unfold in a set pattern [40]: After a process innovation project is initiated, the diagnosis phase begins with evaluation of a current process and its attributes. Information representations, such as semi-structured texts, process flowcharts, and other types of diagrams, are used to capture information about the process [20]. In the subsequent redesign phase, analysts use these process models and creativity support techniques such as brainstorming to envision and choose among possible alternatives. In the reconstruction phase, changes to the process are introduced in the organization, and the new process is evaluated in the evaluation stage.

Our study addresses the redesign stage of process innovation projects [40], particularly the generation of ideas about a current process in the form of a ‘‘future’’ process model. This task can address several components of a business process:

1. C

hanging the control-flow components of a process by, for example, cutting unnecessary, non-value-adding tasks or inserting additional tasks for quality assurance.

2. C

Table 1 Real-world examples of text and diagrammatic representations.

Sector Examples

hanging the technology component on which processes operate by changing the systems, applications, tools, or infrastructure required to execute a process [7]. Examples include changes to manufacturing machines in a production process, the use of new tools and techniques in a decision process, and the use of different digital platforms for communication processes.

3. C

Education Processes in textbooks about business process management

usually provide both textual description and corresponding

diagrammatic models (e.g., Ref. [21]; Example 1.1 and Fig. 1.6)

hanging the organizational component of a process by allocating process tasks to organizational actors (e.g., Ref. [94]) or outside organizations (e.g., Ref. [47]).

Other textbooks provide both diagrams and structured text

4. C

models of business processes (e.g., Ref. [33]; Figs. 6.1 and 6.2)

Research Many experimental studies involving models of business

processes provide textual and diagrammatic models. For instance,

textual and graphic models are used for the processes of:

hanging an information system component of a process by changing how a process is enacted within it or supported by it (e.g., Ref. [86]). An example is implementing a workflow solution for supply chain processes [45].

- Creating a software solution [59]

5. C

- Providing financial services [65, p. 97 and p. 100]

- Providing room service in hotels [44, pp. 66, p.75]

Industry Business process management information and material provided

by industry associations typically include process models (e.g.,

reference models or best practice models) both as textual

descriptions and as diagrams (e.g., system or flowchart diagrams).

Examples include

- The American Productivity and Quality Center: http://www.

apqc.org/pcf

- The American Production and Inventory Control Society: http://

www.apics.org/sites/apics-supply-chain-council/frameworks/

scor

- The Massachusetts Institute of Technology process handbook:

http://process.mit.edu/Default.asp

hanging the data component of business processes by modifying how information is produced or consumed in the course of the process tasks [83] (e.g., through electronic patient records).

The literature on the process of redesign in process innovation, rather than the outcome of redesign, is sparse [66,85]. Sharp and McDermott stated [75, p. 323]: ‘‘How to get from the as-is to the to- be [in a process-redesign project] isn’t explained, so we conclude that during the break, the famous ATAMO procedure is invoked – And Then, A Miracle Occurs.’’

As there is no widely accepted theoretical frame for the redesign phase, we conceptualize process redesign as the conjuring of creative changes to a business process, and process

innovation as the actual implementation of these changes. Our distinction follows West and Farr [90, p. 10], who distinguish between creativity as ‘‘the ideation component of innovation’’ and innovation as ‘‘the proposal and applications of the new ideas.’’

Following this distinction, we can view process redesign as a creative problem-solving activity – that is, an activity that creates solutions that are both original/novel and worthwhile/valuable [80]. Process redesign as a creative problem-solving task involves three steps: idea generation, composition, and evaluation [1]. Typ- ically, a process problem is presented to analysts in the form of information about the current way of working and an objective to introduce changes or overcome issues such as bottlenecks or quality concerns. Then, analysts develop one or more redesign solutions to the problem, identify one preferred solution, and develop and implement the corresponding future process. Finally, the implemented solution is evaluated for its ability to meet the original objective.

2.2. Representing information about organizational processes

To redesign processes to resolve issues, analysts require information about how the processes are currently executed. Current processes are documented using approaches that range from textual documentation, such as policy documents or even e- mails, to structured texts (e.g., in Excel spreadsheets) and visual approaches such as flowcharts and formal diagrams. A global study of process-modeling initiatives in 130 companies [60] showed that 55.9% of the organizations documented their processes as text and 31.5% as tables. The most popular diagrammatic formats were Business Process Model and Notation (BPMN, 21.3%) and Unified Modeling Language (UML, 15.0%). Table 1 provides real-world examples of textual and diagrammatic representations of business processes used in education, research, and industry. In these examples, the text format typically uses sentences and subsen- tences to describe the flow of work, whereas diagrammatic forms use markers such as boxes, circles, and diamonds to illustrate the flow of work.

Our study addresses whether and how the representation format – purely text and purely diagram – influences process redesign as a creative problem-solving activity. Because we are primarily interested in the existence and magnitude of the contrasts between representation formats, we consider these two opposing types of

K. Figl, J. Recker / Information & Management 53 (2016) 767–786 769

representation formats for a methodological reason and a theoreti- cal reason. Methodologically, this choice allows us to focus on strong contrasts between the two end points of the spectrum of representation formats, from textual to visual, in order to ensure instrumentation validity. To facilitate post hoc analyses to examine the results’ ecological validity, we define a control group that receives an intermediary format, structured text.

As for the theoretical reason, several studies have compared text versus diagram formats (e.g., Refs. [25,59]), including unstructured and structured text (such as those in use cases), and traditional diagramming notations, such as flowcharts or BPMN. These studies suggest that diagrammatic representations can help overcome working-memory limitations and improve information acquisition because knowledge put down in models as ‘‘external storage’’ need not be maintained in the working memory [6,74]. For example, Gemino and Parker [25] found that participants’ understanding improved when they had supporting diagrams in addition to textual descriptions. However, Ottensooser et al. [59] reported that readers’ understanding of a process based on a diagrammatic model improved only when they were appropriately trained, while all readers, independent of formal training, benefited from textual description.

Our choice is thus motivated by conclusion validity. Although the literature reports on a variety of benefits of diagrammatic visualizations, results regarding the combination of textual and diagrammatic process descriptions remain inconclusive. Results also appear to depend on the application and task addressed in the research setting. Our interest lies in the differences between textual and diagrammatic representation formats specifically for the task of process redesign, so our study is situated in an information-processing task-setting (e.g., using the information in a problem-solving task). Thus, it extends the prevalent focus of the literature on information acquisition (e.g., developing an under- standing of a business process) to task settings [25,59].

Next, we discuss theoretical viewpoints concerning how representation formats influence process redesign as a creative problem-solving activity.

2.2.1. Process representations as stimuli for creative redesign

The environment in which a creative design activity takes place can affect the creative performance. McCoy and Evans [49] demonstrated that environmental characteristics such as highly complex visual detail and naturalness may positively influence creativity. In addition, creativity-enhancing techniques such as creativity support systems use external stimuli/triggers to inspire designers in searching design and solution spaces [46]. Such stimuli may be inspirational, support analogical creative problem solving, and/or help designers to structure mental representations of the problem domain.

The effect of the format of a process representation as a stimulus to trigger creativity in process-redesign tasks is particularly important because the representation format may significantly affect the quantity and quality of creative ideas [11]. One likely reason for this effect is related to the effort required to do creative problem solving, which, similar to other information processing tasks, is constrained by the limits of the working memory [72]. Bilda and Gero [6] demonstrated that designing blindfolded instead of sketching had a negative effect on idea generation because of the attendant demand on working memory.

It is likely that the demand on working memory during creative problem solving differs with the process representation format. One stream of the literature suggests that visual representations may be best suited to idea generation tasks related to processes because diagrammed process models may have lower working-memory demands than textual representations do. Working memory provides separate subsystems for storing and manipulating visual

and verbal information for cognitive tasks [2]. Purely verbal text involves only one system, while two systems are involved with diagrammatic process models – including visual (symbols and syntax of the modeling language) and verbal information (activity labels). Visual process models may use directed edges (‘‘arrows’’) to depict the process flow and visualize the relationships among the model’s elements, leaving more working memory for idea genera- tion, leading to more and better redesign ideas.

Other studies suggest that visual models may support retention of information and their use in problem solving better than textual models do [48] because ‘‘cues to the next logical step in the problem may be present at an adjacent location’’ [43, p. 65]. One piece of relevant information is located near other relevant information.

Process models may also be closer than text is to the structure of internal human semantic memory (networks with nodes and pathways) [13]. Glenberg and Langston [26] demonstrated that simple process diagrams assist in building mental models because their visual structure is similar to that of the mental model. Therefore, ideas could be generated that are semantically close to the problem at hand.

However, a second stream of research suggests that process models may hinder the development of truly innovative solutions [69]. Diagrammatic and other visual representation formats may evoke fixedness and result in problem solutions that are too similar to the original representation. According to Sarkkinen and Karsten [73, p. 184], ‘‘Visual representations are likely to constrain discussions more than verbal representations. Talk and written language construct rather abstract versions of the subject matter, leaving it open to various interpretations. [. . .] A software process diagram with a strict notation constrains the audience’s imagina- tion more than, for example, a quite freely drawn rich picture diagram.’’

Finally, stimuli may influence not only the quality and quantity of ideas but also the types of ideas generated in a creative task. The desired outcomes of process redesign can affect any of the components of a business process–control flow, information systems, data, technology, organizational resources, etc. Process- redesign ideas may address any of these components, depending on the activation cues in the information material. These cues act as anchors upon which analysts can fixate when they generate ideas. For example, visual examples are known to constrain idea generation because designers tend to conform to such examples. Jansson and Smith [37] demonstrated that both students and experienced engineers in a creative problem-solving task tended to become fixated on a particular type of solution when shown a picture of a suboptimal design along with a textual description.

The root cause of the effect of information representation on types of ideas can be found in the associative theory of creativity [50]. This theory states that stimuli facilitate specific cognitive associations when people are creating new ideas because different stimuli activate different concepts in the knowledge structure. This idea is supported by the theory of spreading activation [14], which proposes that activation of one concept in the internal semantic knowledge network in long-term memory spreads to concepts in the neighborhood. Activated concepts are transferred to working memory and may influence idea generation; hence, diagrams may generate different ideas than text does. As process diagrams are used in the context of information system development, designers’ exposure to them might generate redesign ideas in that realm. For instance, process diagrams highlight the flow of work through arrows and rectangles, so analysts exposed to these diagrams may focus primarily on control flow instead of, say, organizational resources in their ideas about process redesign. It is unlikely that such a thematic fixation would occur with textual representations because of text’s prevalence in everyday life.

K. Figl, J. Recker / Information & Management 53 (2016) 767–786770

In conclusion, although there is some evidence that visual process representations are more effective than text as a cognitive aid in creative process redesign, there is also evidence to the contrary. Table 2 summarizes the studies discussed here and the relevant implications for our study.

We position our research as follows. Our dependent variables contrast those of previous experiments. Most studies have measured creativity in terms of the number of ideas, raters’ creativity scores, and categorized types of ideas. To these measures, we add a process- redesign-specific categorization of ideas that capture the wider context of a business process. Most experiments have included open-ended divergent idea/solution-generation tasks for ill-defined design problems of various subject areas in which participants wrote or drew either one solution or as many solutions as possible.

Table 2 Summary of literature on representation formats and creative problem solving.

Reference Independent variable Dependent variable

Comparison studies

[46] Use of words versus pictures

(e.g., a photograph of a cake) as

stimuli

Open-ended idea generation for

new ice cream flavors (creativity

score was based on judges’

ratings of ideas’ novelty and

feasibility)

[31] Pictorial and textual (cyclone)

distant stimuli

Open-ended generation of ideas

about transportation in 2050;

drawings with short descriptions

(fluency, originality ratings, and

type of ideas)

[8] Pictorial and textual stimuli and

a control group without stimulus

Design of a device to pick up a

book from a shelf that is out of

reach (as many designs as

possible)

Studies of visual stimuli only

[79] Designs with and without

pictorial examples, with textual

labels

Open-ended generation of ideas

for new toy creatures (sketching

and labeling)

[37] Designs with and without

pictorial examples

Ill-defined mechanical

engineering tasks (as many

designs as possible)

[9] Pictures (photographs and

architectural drawings) with and

without the instruction to use

visual analogy

Solving ill-designed

architectural design problems

(one solution)

Studies of textual stimuli only

[28] Designs with and without

textual examples

Design of a chair for children or a

desk clock (one solution)

[30] Textual stimuli with differing

levels of abstraction (related,

distant–a book excerpt,

unrelated text) and control

group without stimulus

Open-ended generation of ideas

for transportation in 2050

[82] Designs with and without

textual stimuli

Sketching and describing one

idea for a chair

Although the results of previous studies on design activities can be generalized (to some degree) to generating process-redesign ideas, we use a specific categorization so as to ensure that the domain specificity of creativity is included [93]. Our study also differs from extant research in terms of the independent variable, as we use diagrammatic process models to compare to textual models, rather than the pictures, photographs, or sketches previous studies have used. Another difference is that we present the problem setting, rather than example solutions, in textual/visual form. Finally, our literature review suggests that clarification is required to resolve inconsistencies. As Table 2 shows, the literature reports both positive and constraining effects of textual and pictorial/visual stimuli on idea generation and that textual representations alone may evoke creative ideas, but not always. Determining whether and

Summary of study Implications for this study

The independent variable had no effect

on the number of ideas, but picture

stimuli led to more creative ideas than

words or the combination of words and

pictures. Word stimuli might lead to

inappropriate ‘‘design fixation.’’

Visual representations of processes

might lead to different and

increasingly creative–but not

more–process-redesign ideas than

textual representations do.

The textual condition outperformed the

pictorial condition in terms of the

number and originality of ideas. The

authors argue that text is

underestimated as an inspirational

source and that pictures can both

stimulate and hamper creativity. Both

conditions triggered various categories

of ideas. This was also explained with a

recency effect (focus on last words of a

text).

Counterargument to [46]: Textual

representations might lead to more

original redesign ideas. Empirical

testing will be required to

determine which of the two

hypotheses hold for process

redesign. In addition, textual

representations might lead to

different types of ideas than visual

representations do.

Neither pictorial nor textual stimuli

affected the number of ideas. The

pictorial stimuli led to ideas similar to

the example (fixation effect), while text

led to no more fixation as compared

with the control group.

There might be no discernible

difference between representation

formats in terms of the number of

ideas, but visual process

representations might lead to

process-redesign ideas that are less

appropriate than textual

representations do.

The ideas generated conformed closely

to the examples presented, and their

originality was constrained [79].

Visual process representations

might lead to unoriginal

process-redesign ideas that are also

less appropriate.

Design fixation occurred in the

experimental groups that were

provided with the pictorial example

(conformance to stimulus, reusing

parts of the example even when

inappropriate).

Visual process representations

might lead to inappropriate

process-redesign ideas.

Pictorial representations stimulated

design solutions even when

participants were not explicitly

instructed to use visual analogies.

Visual process representations

might lead to process-redesign

ideas that focus on the visually

highlighted elements (notably

control flow).

Text stimuli led to a higher level of

originality than no stimuli did, but

practicality was not affected.

Textual process representations

might lead to process redesigns that

are more original but not more

appropriate.

The ‘‘appropriate’’ abstraction level–the

distant text–led to a higher number of

flexible (ideas in slightly different

categories) and more original ideas.

Textual process representations

might lead to different types of

process redesigns and to ideas that

are more original if they have an

appropriate abstraction level.

Text led to a higher number of creative

ideas, and more possibilities were

explored, but the overall creative

quality did not increase.

Textual process representations

might lead to more process-

redesign ideas, but the ideas might

not be more original.

K. Figl, J. Recker / Information & Management 53 (2016) 767–786 771

how process redesign is affected by textual representations, which are a dominant form of process information in practice, can clarify which representations to suggest for redesign projects. A final verdict on this matter requires an empirical analysis.

3. Research model

The debate about the relative merits of information represen- tation formats, together with the dearth of empirical research on idea generation in process-redesign tasks, indicates the absence of theory with which to structure our empirical study. Therefore, we follow a scientific exploratory approach, rather than a purely confirmatory approach. To guide this investigation, we first develop a framework that describes the elements to be included in an empirical research design. Next, we develop two sets of hypotheses with which to investigate the effects of representations of business processes on readers’ ability to perform creative problem-solving tasks by generating process-redesign ideas. Fig. 1 shows the research model that frames our empirical study.

The model shown in Fig. 1 frames our primary research interest: the influence of the type of process representation on the creativity and type of the process-redesign solutions. Based on findings in the literature on how individual characteristics relate to creative problem-solving processes, the model acknowledges the relevance of the individual as a creative person by using creative competence [16] and creative attitude [4] as control variables. Table 3 provides construct definitions for all factors in the model and lists the literature on which the definitions were based.

Based on our model, we present two sets of hypotheses that describe our expectations about the effects of a type of representation in a process-redesign task on the solutions conceived in this task. First, we explore whether the quantity and quality of process-redesign solutions varies. Although the literature on the effects of textual and visual stimuli on idea generation has shown both enabling and inhibiting effects of both types of representation, several arguments suggest visual descrip- tions of process models and may be superior to textual descriptions. Research has shown that diagrammatic representa- tions can improve understanding of a business process if readers are sufficiently familiar with the diagrammatic notation [59], as the spatial arrangement of information in diagrammatic models[(Fig._1)TD$FIG]

Fig. 1. Research model.

improves information search and reduces the need to store information in the working memory [92], leaving capacity for finding creative solutions. Therefore, we expect that diagrammatic process representations lead users to create increasingly more creative solutions to the process-redesign task. The creativity literature has typically differentiated the number of solutions (fluency of ideas) found in a creative problem-solving task from their quality (originality and appropriateness) [19]. Accordingly, we suggest the following:

Hypothesis 1a. Users of diagrammatic process representations develop more solutions to the process-redesign problem than do users of text process representations.

Hypothesis 1b. Users of diagrammatic process representations develop higher-quality solutions to the process-redesign problem than do users of text process representations.

Second, we suggest that the type of process-redesign solutions varies. We speculate that textual and diagrammatic process representations activate different kinds of knowledge in the long-term memory, thereby spreading to concepts in different ‘‘neighborhoods’’ of the memory [72]. The difference in the ‘‘neighborhoods’’ is important because the resources that process redesigns require can differ. Organizations typically require to document various resources in their process representations: apart from the sequence of tasks in a process (i.e., control flow components), Patig et al. [60] found that information (i.e., data components), personnel (i.e., organizational components), soft- ware (i.e., information systems components) and machines, material, and applications (i.e., technology components) are the process resources listed most frequently.

Which of these resources a process-redesign solution requires depends on the information provided to the analysts because different representations highlight these resources in different ways. For example, if a representation conveys nothing about, say, the software a process uses, then it is unlikely that process-redesign solutions will address this component. By contrast, if the representa- tion draws attention to the actors involved in a process, then analysts are more likely to consider solutions that focus on organizational components. Therefore, we suggest that the type of process representation influences the types of process-redesign ideas.

There is no strong theory with which to speculate ex ante which type of process design will feature prominently in process-redesign solutions, but our understanding of the literature leads us to believe two effects will occur. First, we expect that analysts who work with diagrammatic process representations generate more ideas that focus on improving a process’s control flow components than other components because the most prominent feature of visual process diagrams is the logical and temporal sequence of activities using boxes and arrows. The spatial arrangement of the predecessor– successor relationships between tasks is likely to facilitate more efficient information processing than does neutral text, where the mind must first identify the activities (e.g., by finding verbs in the text) and then infer their temporal and logical relationships from the text [48]. This added complexity is likely to reduce the amount of working memory available to focus on generating creative solutions. Therefore, we posit as follows:

Hypothesis 2a. Users of diagrammatic process representations develop more process-redesign solutions that feature control flow components than do users of text process representations.

Second, we expect that analysts who work with diagrammatic process representations will generate more ideas that focus on improving the organizational resource components of a business process than other components. In visual process diagrams, organizational resources are often modeled using swim lanes,

Table 3 Construct definitions.

Construct Definition Relevant literature

Creativity of process-redesign ideas The number and quality of ideas that are novel and purposeful and that provide an

effective solution

[19]

Type of process-redesign ideas The component of a business process that is addressed by the process-redesign idea:

- control flow (the sequence and order of tasks)

- information system (the application software in use in the process)

- data (information consumed or produced in the process)

- technological resources (technology required to execute the process)

- organizational resources (organizational actors involved in the process)

New construct based on theories of

control flow components [87] and the

wider context of a business process [70]

Type of process representation The textual or visual notation used to convey information about a business process,

including the available graphic or textual symbols and the relevant compositional

rules

[59]

Creative competence The individuals’ creative thinking competence in terms of fluency (number of ideas),

originality (novelty of ideas), and elaboration (embellishment of ideas with details)

[16,27]

Creative attitude The individuals’ attitude toward creative problem-solving tasks in terms of intrinsic

motivation, preference for ideation, and tendency for premature critical evaluation

of ideas

[4]

K. Figl, J. Recker / Information & Management 53 (2016) 767–786772

which allow the easy identification of the number of actors involved and their various responsibilities in executing the process [5,38]. These spatial arrangements are not present in text; hence, the identification of actors and their roles requires careful reading of the text, which consumes working memory and leaves less for generating ideas. Therefore, we expect

Hypothesis 2b. Users of diagrammatic process representations develop more process-redesign solutions that feature organiza- tional resource components than do users of text process repre- sentations.

These two sets of hypotheses help us to evaluate empirically a long-standing assumption about the beneficial use of diagram- matic process representation formats in process-redesign projects – that is, whether and how types of process models assist analysts in process improvement.

4. Method

4.1. Design

We conducted an experiment to provide evidence about the impact of process models while controlling for other factors. Our research design was motivated more by internal validity than by external validity. We used a controlled repeated measures design with one primary between-group factor, two covariates, and one within-group factor.

The between-group factor, type of process representation, had three levels: (a) a textual and (b) a diagrammatic representation of information about a pizza-delivery process (Fig. 2), and (c) an intermediary representation format ‘‘structured text’’ [88]. The ‘‘structured text’’ group helps us to differentiate the main results from those from a third experimental group for manipulation- check purposes, so we can distinguish the factors that cause the differences between text and diagrams (particularly structure vs. use of symbolic vocabulary). Structured texts introduce only structure with no symbolic vocabulary and resemble the process descriptions that are available in tables, which is a third common representation format used in industry [60]. Fig. 5 in Appendix A depicts the ‘‘structured text’’ treatment. We report on these data in a post hoc analysis of the main results below.

Congruent with our research model, we included two types of control variables in the experiment – creative competence and creative attitude – described in Section 4.3.3.

The within-group factor trial had three levels, operationalized as three creative problem-solving tasks with differing process- improvement objectives. The purpose of multiple trials was to

strengthen the external validity of the findings by examining task solutions across three process-redesign objectives.

We used two categories of dependent variables. First, we measured the solutions’ creativity in terms of fluency (number of ideas), appropriateness, and originality, as is common in the creativity literature (e.g., Ref. [27]), and in terms of their impact [62]. We added the impact dimension in order to relate creative problem-solving solutions back to the original business objective of changing a process (thus to differentiate process-redesign solutions that are truly relevant to the business from other creative solutions). Measuring fluency addresses Hypothesis 1a, while the other measures relate to Hypothesis 1b.

Second, we used a measure we developed for the type of solutions in terms of the locus of change, that is, as affecting the control flow, information systems, or the organizational, techno- logical, or the data component of a business process. This measure relates to Hypotheses 2a and 2b.

4.2. Participants

The goal of our study was to ascertain whether different process representation formats would lead to differences in process redesign solutions produced by novice analysts. The population of interest to our study thus consists of business users of process representation formats that would be involved in process-redesign activities. This business cohort is thus wider than process modelers alone, whose tasks typically consist of describing a process in a particular representation format. Instead the cohort also includes process managers, analysts, and domain experts, many of whom do not have method experience [20].

Following recommendations for sample selection [15], we recruited university students from a business school as proxies for future end-users of process representations who have at least some knowledge about business domains and business-process management. Students are also less likely than experienced (and difficult to recruit) practitioners to vary in their knowledge about and experience with modeling methods, creativity, and innovation management. A particular advantage of using a novice sample is that they have not been ‘‘brainwashed’’ into a particular format through years of experience with process-representation formats as industry experts would be; thus, we minimized the chance of bias [23].

Participation incentives included access to a copy of the summarized study results and s15 cash. We used a balanced block-randomization strategy with blocks of variable length to assign the 120 students who volunteered to participate in the experiment to the three experimental groups.

[(Fig._2)TD$FIG]

Fig. 2. Representation of the pizza-delivery service process.

K. Figl, J. Recker / Information & Management 53 (2016) 767–786 773

4.3. Materials and procedures

We used a paper-based experimentation system. The experi- ment took place in a daylight computer laboratory. Appendix A includes the experimental material used except for the test used to measure creative competence and the scales used to measure creative attitude due to copyright restrictions.

4.3.1. Experimental tasks

We asked participants to analyze and redesign a business process for a pizza-delivery service business process. We chose this process because it is widely understood and requires little specialized knowledge. We could safely assume that all student participants had some knowledge of pizza delivery from a consumer perspective. We created the process scenario based on a simplified version of the pizza-delivery example given in the BPMN 2.0 standard [57].

All participants were provided with a process description of the pizza-delivery service, either as a textual description or as a diagrammatic process model. The process descriptions for the three experimental groups were created to contain the same information, although the representation format varied (text, diagram, or structured text).

In creating these representations, we started with the text and diagram representations of the process in Ref. [58] and then amended them to ensure the information provided in each was equivalent.

The unstructured text uses full sentences (e.g., ‘‘the pizza chef checks the oven temperature,’’ ‘‘the delivery person gives change’’), which we adapted for the diagram and structured text representa- tions to match established verb–object style conventions [52] (e.g., ‘‘prepare dough,’’ ‘‘give change’’). In addition, we used vertically arranged swim lanes to group process activities performed by an actor (pizza costumer, clerk, pizza chef, and delivery person), so each actor is mentioned only once, as the label of a swim lane. Next, we replaced textual terms in the unstructured text that indicate the activities’ tempo-logical flow (e.g., ‘‘first,’’ ‘‘then,’’ ‘‘next’’) with spatial–visual placements in the structured text and with symbolic vocabulary in the diagram [58]. By superimposing printing

techniques, we used the same visual layout in the structured text condition that was used in the process diagram but used symbols (routing symbols and start and end symbols, as proposed by BPMN [58]) in the process diagram but pseudocode identifiers (e.g., ‘‘Begin,’’ ‘‘If,’’ ‘‘Terminate,’’ Endif,’’ ‘‘End’’) in the structured text [91]. The visual process diagram also uses edges to visualize the flow of the process and boxes to visualize the swim lanes. Apart from these changes, we held other variables constant over all three representations to facilitate fair comparison and minimize confounding. All textual elements were set to the 12-point Arial font, with the representations spanning the width of an A4 page. (Textual description was justified.) Finally, we kept the reading direction uniform (left to right and top to bottom). Fig. 2 shows the final text and diagram representations of the pizza-delivery process. The structured text version is provided in Fig. 5 in Appendix A.

Because we wanted to eliminate potentially confounding effects of model complexity [64], we ensured that the chosen business process was represented in a moderately complex, rather than a very simple or very complex, manner. To gauge the level of complexity, we compared the example of the pizza with the complexity of an average model in the collection of 1400 practi- tioner models reported in Ref. [42], using the complexity metrics defined in Ref. [51]. Our case contains more arcs than an average model (28 or 158%) and nodes (27 or 221%) and maintains an approximately average connector degree (3 or 92%). Therefore, it lies between a low-complexity model and a high-complexity model [64]. Its number of tasks (19 vs. 19.1) is similar to the processes examined in Ref. [60].

With each participant using one of these three process representations, participants worked on three open-ended prob- lem-solving questions. The objective of these tasks was to identify areas of improvement in the process in terms of efficiency and effectiveness gains, in alignment with a specified process objective. Because of the high-level nature of the process descriptions, participants had to make some appropriate and probable assumptions about the business case.

To avoid monomethod bias in the creativity assessment and determine whether noted effects would differ or be consistent

Table 4 Idea-generation tasks.

Task 1 Improvement-invoked ‘‘The pizza-delivery service wants to

improve its processes, so that customers

know at all times when their pizza will

arrive. How can the process be changed to

implement that improvement? Provide as

many options as you can think of.’’

Task 2 Pattern-invoked ‘‘The pizza-delivery service gets a new

employee. How could the employee be used

most effectively to improve the process?

Provide as many options as you can think of.’’

Task 3 Measure-invoked ‘‘The pizza-delivery service wants to cut

down costs. How could the process be

changed to most effectively reduce costs?

Provide as many options as you can think of.’’

K. Figl, J. Recker / Information & Management 53 (2016) 767–786774

across differing task settings, we used three tasks to measure creativity as the trial levels in our experiment (Table 4). All three tasks focused on idea generation for process innovation. We designed these tasks to allow for variation in the salience of process-improvement ideas and to cover differing approaches to process innovation. Following Shtub and Karni [76], we chose three tasks, corresponding to three types of process innovation procedures (Table 4).

We designed Task 1 as an improvement-invoked procedure, which Shtub and Karni [76, p. 222] indicated requires that designers be cognizant of the improvement objective. We implemented this requirement in the instruction that customers know at all times when their pizzas will arrive.

We designed Task 2 as a pattern-invoked procedure. Specifi- cally, we included pattern no. 26 (‘‘increase the number of performers carrying out a process’’ (see Ref. [76], p. 223)) in the description by asserting that a new employee would be available to assist in the process.

Finally, we designed Task 3 as a measure-invoked procedure: we included in the task description a specific process-quality metric (cost) and the associated objective (cost reduction), thereby adhering to the objective of upgrading process performance in terms of cost, quality, time, or flexibility [76].

4.3.2. Result coding

Three research assistants coded the creativity of process- redesign ideas. All had experience in business process manage- ment – through either university education or job experience – as well as experience as data coders for research projects and familiarity with the domain of the model from a consumer perspective (ordering pizzas). The research assistants, who were

Table 5 Measurement of creativity of ideas.

Attribute Measurement Definition

Fluency Continuous scale The number of relevant id

provided.

Originality 5-point scale (where 1 = not at all

original, 3 = medium originality,

and 5 = very original)

Something that is original,

unexpected, and novel [19

Appropriateness 5-point scale (where 1 = not at all

appropriate, 3 = medium

appropriateness, and 5 = very

appropriate)

Something that is useful, m

task constraints, and is

purposeful.

Impact 5-point scale (where 1 = no

positive impact at all,

3 = medium positive impact, and

5 = very positive impact)

The positive tangible and

intangible effects

(consequences) of one ent

action or influence upon an

unfamiliar with the purpose of the study, scored all ideas according to a pre-developed coding schema. Instead of relying on a general creativity score, we defined four attributes of creative performance that are predominantly used in creativity research – fluency, originality, and appropriateness [55,62] – plus impact [62]. To this, we added our self-developed categorization of the type of process- improvement idea (‘‘the locus of change’’). Table 5 summarizes the rationale and the construct definitions for the four dimensions of creativity. We gave the coders an explanation of each dimension and examples (e.g., ‘‘using a tissue as a napkin’’ is less original than ‘‘using tissues to make a costume for the next Halloween party (e.g., ghost or fairy)’’). Coders rated the dimensions on a five-point scale (where, 1 = not at all appropriate, 3 = medium appropriate- ness, and 5 = very appropriate), with the exception of fluency, which was coded as the number of solutions provided. The final coding schema is shown in Appendix B.

All coders were instructed on the coding schema, and several iterations with sample data from the participants were used to increase familiarity with the process and the definitions. Once the coders were sufficiently familiar with the criteria and the process, they eliminated responses that were clearly unrelated to the situation (e.g., ‘‘costumer called the wrong pizza service’’). Three such answers were eliminated. Then they coded all remaining responses against the coding scheme, scoring each response independently. Next, they met to discuss their ratings, which led to revisions to the individual interpretations of the coding scheme and revisions in scoring. This iterative process was repeated to eliminate inconsistencies in the ratings and ensure the reliability of the coding. Therefore, by design, the inter-rater reliability was 100%. Table 6 provides illustrative solutions and their final coding.

� T

ype of process-improvement idea (locus of change):

We also measured the key focus of the proposed improvement solutions – that is, the process component that a solution primarily addressed. Table 7 shows how we coded example answers given for task 3 (cutting costs).

4.3.3. Posttest evaluation: demographics, modeling experience,

creative competence, and creative attitude

We collected demographic data and data on task-related (participation in process-improvement initiatives) and domain- related (i.e., ordering pizza as a customer) experience and on experience with process models (how many process models participants had read or created), and we asked the participant to rate their work intensity with process models on a five-point scale

Rationale

eas This attribute measures ideational fluency as the number of

semantically different ideas [32].

].

Originality is the attribute most often used in divergent creativity tests

[55] and the most commonly mentioned attribute of creativity [71]. It is

defined as resulting in ‘‘ideas that are not only rare but that also have

the characteristic of being ingenious or imaginative’’ [19,p. 659].

eets Appropriateness ‘‘refers to the extent to which a proposed solution can

satisfy the demands posed in a problem context’’ [93,p. 31] and

determines whether a solution makes sense in its context [36].

ity’s

other.

Impact refers to the benefits that can be derived from implementing a

proposed solution–that is, the profit or gain in terms of monetary and/

or non-monetary advantages (e.g., in terms of cost or time savings,

increased customer or staff satisfaction, or other criteria).

The impact or influence of a creative idea differs from its

appropriateness and ‘‘indicates the extent to which an idea changes a

particular domain’’ [62], which is particularly applicable to the task of

innovating a business process.

Table 6 Sample solutions for Task 1: The pizza-delivery service wants to improve its processes, so that customers know at all times when their pizza will arrive. How can the process

be changed to implement this improvement?

Originality

(Low, 1)

� ‘‘Tell them to set an alarm clock.’’

Originality

(Low, 2)

� ‘‘Promise that the pizza will be delivered in a certain number of minutes.’’ � ‘‘Oven temperature display near phone.’’

Originality

(High, 5)

� ‘‘Webcam in the kitchen with live stream. Pizzas get name cards and can be observed while baking.’’ � ‘‘On the Internet, there might be an ‘‘avatar’’ chef, and so on, where customers can watch a cartoon about the process while waiting: Chef preparing pizza – oven – baking – street delivery. The moment the cartoon delivery person rings the bell, the real one will be there.’’

Appropriateness

(Low, 1)

� ‘‘Have multiple locations around the city, making the delivery quicker.’’ � ‘‘Prepare the grated cheese in advance.’’ � ‘‘Pizza chef and delivery person regularly visit doctors as a precautionary measure.’’ � ‘‘Keep pizza oven turned on continuously.’’

Appropriateness

(High, 5)

� ‘‘Each order gets a barcode number. The clerk can then scan it and pass it to the kitchen. The chef scans it when he or she starts to cook, then when it is in the oven, and again when the delivery person gets it. On the website, the customer can look up in which production stage the pizza is now.’’

� ‘‘GPS tracking of the delivery car so the customer can watch online where it is right now, see when the pizza is done, etc.’’ Impact

(Low, 2)

� ‘‘Calculating more time as is needed, so the customer knows before.’’ � ‘‘Very simple estimation by average times.’’ � ‘‘An ‘‘all for one’’ option: every pizza is delivered within half an hour after an order is placed.’’ � ‘‘Customers can order pizza in advance for delivery at a particular time.’’

Impact

(High, 5)

� ‘‘The pizza-delivery service could buy or order software for the delivery cars that the customers can watch online. At any time, the customer could follow the car coming to him or her. Every car could have a number and an option for showing the estimated time until it reaches the target (the

customer).’’

� ‘‘Use web facilities (separate login, where customers can track the status of their pizzas – leads to more playful engagement, where customers can view their pizzas’ progress (computer animation of process) from baking to delivery). Delivery will be visualized via GPS sensors on a map so the

customer is always informed about the status and location of his or her pizza.’’

Table 7 Sample answers for types of process-improvement ideas.

Type Characterization Example answers

Control flow The nature, sequence, and order of the tasks to be executed

in a pizza-delivery service (e.g., prepare dough, bake pizza,

select toppings)

� ‘‘Take out as many pizzas at the same time as possible (no half empty deliveries!).’’

� ‘‘Receipt-making in the restaurant – saves time.’’ � ‘‘The delivery person should be informed before the pizza is ready.’’

Organizational resources The staff involved in the pizza-delivery process (e.g.,

delivery person, pizza chef)

� ‘‘Pizza chef is replaced by another cook less well educated – cheaper.’’ � ‘‘Cut salary for employees.’’ � ‘‘Outsource clerk call center to low-wage country.’’ � ‘‘Clerk is let go and pizza chef takes orders as well.’’

Technological resources The tools and infrastructure involved in the pizza-delivery

process (e.g., oven, fridge, car)

� ‘‘Invest in an automatic pizza oven that always has the right temperature, can take more pizzas, and knows when a pizza is ready.’’

� ‘‘Delivery by bike on smaller routes.’’ Information system Any computerized system that might be involved in

managing information about the pizza-delivery process

(e.g., online ordering system, short messaging services,

pizza status dashboard, electronic payment system)

� ‘‘Automatic clerk system (standard answers, voice recording of orders).’’ � ‘‘Electronic order service – voice recognition – clerk is not needed, information automatically forwarded to chef.’’

� ‘‘Replace the leaflets with a website where you can also place the orders.’’ � ‘‘Switch to internet-based order system instead of the phone. More orders can be taken at one time and easily monitored by the clerk. More orders –

more pizzas sold.’’

Data Any input or output information required or created in the

pizza-delivery process (e.g., recipe, pizza orders, etc.)

� ‘‘Discounts for orders with more than five items.’’ � ‘‘Add minimum order-value.’’ � ‘‘A higher minimum of the order value.’’

K. Figl, J. Recker / Information & Management 53 (2016) 767–786 775

(from never to always). We also used the three-item process- modeling familiarity scale from Recker [63] to measure their perceived familiarity with process model diagrams.

The participants’ last task measured creative competence and creative attitude. This task came last to avoid task-order bias because instruction to be creative can influence creative output [56]. To measure creative competence, we relied on the shortened version of the Torrance Tests of Creative Thinking (TTCT), a widely used instrument (e.g., Ref. [10]) that measures divergent thinking abilities and assesses the quantity and quality of creative ideas [12]. The test includes verbal and figural subtests, and the scores in the test are based on fluency (number of ideas), originality, and elaboration (the amount of additional details). Two psychologists who had the knowledge and skill required scored the TTCT. The use of certified professionals for psychological test administration is also a legal requirement in Austria, where the study took place. In addition, we wanted to comply with the American Psychological Association’s standards for test-user qualifications [84].

We used three scales to measure creative attitude: the ‘‘preference for ideation’’ scale [4] with the example item ‘‘One new idea is worth ten old ones,’’ and the ‘‘tendency for premature critical evaluation of ideas’’ scale [4] with the example item ‘‘Quality is a lot more important than quantity in generating ideas.’’ We measured intrinsic motivation to perform the process-redesign activities using items from Davis et al. [18]. Together, these three scales provided a meaningful evaluation of the respondents’ creative attitudes.

5. Results

5.1. Data screening

In examining the data for outliers, we excluded participants who were not currently enrolled in business administration or had already completed a degree in that field because business students score differently on creativity tests than other students do [22]. A variation in areas of study could introduce an experimental bias.

K. Figl, J. Recker / Information & Management 53 (2016) 767–786776

Four participants indicated that they had participated in more than ten process-improvement initiatives, while the rest of the participants had participated in five or fewer, so the original sample size of 120 was reduced to 108 to achieve homogeneity. Table 8 summarizes the demographic statistics, including those for the structured text group that we used for a post hoc analysis of the results.

To screen for differences among the three experimental groups, we computed appropriate statistical tests, shown in Table 8. The results did not suggest significant differences, with the exception of domain-related experience, where the participants who worked with a diagrammatic representation of the process had, on average, less experience ordering pizzas than the other participants did. We did not anticipate any result bias because of this demographic difference.

Next, we performed manipulation checks, described in Appendix C. Given the results from our manipulation checks, we performed several supplementary analyses to determine the influence of gender, domain experience, and task order. These analyses are summarized in Appendix D.

Finally, we examined our multi-item scales for reliability and internal consistency. The scale used to measure familiarity with process diagrams, adopted from Ref. [63], the scale used to measure the tendency toward premature critical evaluation of ideas [4], and the scale used to measure intrinsic task motivation [18] had Cronbach’s alpha coefficients of 0.92 (familiarity with process diagrams), 0.76 (tendency toward premature evaluation of ideas), and 0.92 (intrinsic motivation), indicating sufficient reliability and internal consistency. The preference for ideation scale [4] had a Cronbach’s alpha of 0.45, indicating a lack of reliability; hence, we eliminated this factor from all subsequent analyses.

5.2. Hypothesis testing

Data analysis was performed using SPSS Version 20. To identify differences between the main experimental groups (‘‘diagram’’ and ‘‘text’’), we performed analysis of covariance (ANCOVA) for repeated measures tests, with the treatment (text or diagram) as the independent variable for each dependent variable (fluency, appropriateness, originality, and impact of a future process; number of control flow-/information system-/data-/technological resources-related ideas) in all three creativity tasks. To determine which measures of creative competence and creative attitude to include in the analysis as covariates, we first checked to see whether any of these control variables had a significant linear correlation (see Appendix C2) with any of the dependent measures

Table 8 Participants’ demographic data.

Text (n = 36) Diagr

M/Count SD/Percentage M/Co

Age 25.03 3.54 23.43

Gender

Male 19 53% 12

Female 17 47% 23

Highest degree completed

High school 5 14% 6

One or more years of university 24 67% 22

Bachelor’ degree 4 11% 6

Master’s degree 3 8% 1

Work intensity with process models

(5-point scale)

2.72 0.62 2.83

Number of models created or read 4.39 4.51 7.29

Familiarity with BPMN process

model diagrams

0.33 0.80 0.27

Task-related experience (participation

in process-improvement initiatives)

0.67 1.29 0.74

Domain-related experience (pizzas ordered) 39.31 44.82 17.54

(indicating that they needed to be controlled). We retained one measure for individual creative competence (fluency) as a covariate for three of the eight ANCOVAs with repeated measures.

While intrinsic motivation was not significantly correlated to the dependent variables, it was generally high, averaging 5.46 (standard deviation (SD) = 1.27) on a seven-point scale. Therefore, the participants considered the tasks to be enjoyable, which was likely to spur creativity.

5.2.1. Hypotheses 1a and 1b

Hypotheses 1a and 1b suggested that solutions proposed by diagram and text users would differ in terms of creativity (i.e., fluency for Hypothesis 1a and originality, appropriateness, and impact for Hypothesis 1b). Results from our ANCOVA for repeated measures tests are summarized in Table 9. Significant main results (at the p = 0.05 level) are printed in bold. Results on the influence of the covariates and interaction effects are only reported if significant. Wherever the assumption of sphericity was violated, we report Greenhouse–Geisser corrected values. Fig. 3 presents the results graphically.

The results shown in Table 9 and Fig. 3 indicate that the ‘‘diagram’’ group generated ideas that were more appropriate than those of the ‘‘text’’ group (MDiagram = 3.56, SDDiagram = 0.28; MText = 3.36, SDText = 0.38; p = 0.01). They also produced ideas of greater originality (MDiagram = 3.10, SDDiagram = 0.33; MText = 2.97, SDText = 0.34; p = 0.09) and impact (MDiagram = 3.59, SDDia- gram = 0.16; MText = 3.50, SDText = 0.25; p = 0.09), although these results were not significant at the p = 0.05 level. The significance levels stayed similar when the control variables of gender, domain experience, and task order were included in the analyses.

The results also indicate that either the within-subject effect on the creativity task or the interaction effect between the individual creative competence and the creativity task was significant for all dependent variables. Therefore, the innovation objectives influ- ence the creative outcome of the problem-solving task.

Although most of the results were in line with our expectations, the number of ideas produced was similar between the two groups (MDiagram = 4.01, SDDiagram = 1.84; MText = 3.65, SDText = 1.37). Only the individual creative competence factor affected the number of ideas produced, confirming the widely held assumption that participants with higher creativity produce more ideas. Therefore, we found no support for Hypothesis 1a from the data on fluency of ideas. However, results strongly supported Hypothesis 1b regard- ing appropriateness and partially supported Hypothesis 1b regarding originality and impact (in terms of directionality but

am (n = 35) Structured text (n = 37) Statistical test

unt SD/Percentage M/Count SD/Percentage

2.78 24.73 3.57 Fdf=105 =�2.32; p = 0.10

34% 20 54% X2 df¼2 ¼ 2:51; p ¼ 0:29

66% 17 46%

17% 5 14%

63% 20 54% X2 df¼6 ¼ 3:51; p ¼ 0:74

17% 8 22%

3% 4 11%

0.89 2.59 0.60 Fdf=105 = 0.98; p = 0.38

10.80 4.24 6.15 Fdf=105 = 1.82; p = 0.17

0.66 0.17 0.48 Fdf=105 = 0.57; p = 0.57

1.67 0.73 1.43 Fdf=105 = 0.03; p = 0.97

16.32 41.46 51.61 Fdf=103 = 3.73; p = 0.03

[(Fig._3)TD$FIG]

4.01

3.65 3.56

3.36

3.10 2.97

3.59 3.50

2.00

2.50

3.00

3.50

4.00

4.50

Diagram Text Diagram Text Diagram Text Diagram Text

Fluency Appropriateness Originality Impact

M ea

n sc

or e

pe r g

ro up

Fig. 3. The influence of the type of representation on the creativity of solutions.

Table 9 Experimental results: influence of representation format on the creativity of process-innovation solutions.

Dependent variable Factor F (dfHypothesis, dfError) p h 2

Fluency Between-subject effect Representation type

(text vs. diagram)

>0.10

Within-subject effect Creativity task >0.10

Covariate Individual creative

competence (fluency)

9.70 (1, 68) 0.003 0.13

Interaction Individual creative

competence

(originality)�creativity task

6.19 (2, 136) 0.003 0.08

Appropriateness Between-subject effect Representation type

(text vs. diagram)

6.47 (1, 69) 0.01 0.09

Within-subject effect Creativity task 5.20 (2, 118) 0.01 0.07 Originality Between-subject effect Representation type

(text vs. diagram)

2.96 (1, 69) 0.09 0.04

Within-subject effect Creativity task 22.71 (2, 119) <0.001 0.25 Impact Between-subject effect Representation type

(text vs. diagram)

2.92 (1, 69) 0.09 0.04

Within-subject effect Creativity task 75.30 (2, 138) <0.001 0.52

p-values�0.05 are printed in bold.

Table 10 Experimental results: influence of representation format on creative problem solving.

Dependent variable Factor

Control flow-related ideas Between-subject effect Representat

(text vs. dia

Within-subject effect Creativity ta

Covariate Individual c

competence

Interaction Individual c

competence

Information system-related ideas Between-subject effect Representat

(text vs. dia

Within-subject effect Creativity ta

Data-related ideas Between-subject effect Representat

(text vs. dia

Within-subject effect Creativity T

Technological resources-related ideas Between-subject effect Representat

(text vs. dia

Within-subject effect Creativity ta

Covariate Individual c

competence

Interaction Individual c

competence

Organizational resources Between-subject effect Representat

(text vs. dia

Within-subject effect Creativity ta

p-values�0.05 are printed in bold.

K. Figl, J. Recker / Information & Management 53 (2016) 767–786 777

not significance of the effect). Overall, then, we see enough evidence to accept Hypothesis 1b, that the use of diagrammatic process representations leads to higher quality of creative ideas in process-redesign solutions than the use of textual representations, without necessarily influencing the number of outcomes in the idea-generation process (Hypothesis 1a).

5.2.2. Hypotheses 2a and 2b

Hypotheses 2a and 2b suggested that uses of process diagrams versus process texts would lead to differing types of process-redesign solutions, and in particular that the use of visual diagrams would lead to more process-redesign solutions featuring control flow (Hypoth- esis 2a) or organizational resources (Hypothesis 2b). We again ran ANCOVA for repeated measures tests with the same independent factors and covariates and using the count of ideas per categorization as dependent variable. Table 10 summarizes the significant results from the test, including F-statistics, test result (p-value), and effect size (h2), and Fig. 4 visualizes the differences graphically.

We noticed that there are significant differences in the categories of information systems (p = 0.02) and data (p = 0.04) between the

F (dfHypothesis, dfError) p h 2

ion type

gram)

>0.10

sk >0.10

reative

(fluency)

8.56 (1, 68) 0.005 0.11

reative

(fluency)�creativity task 4.03 (2, 98) 0.02 0.06

ion type

gram)

6.10 (1, 69) 0.02 0.08

sk 47.27 (2, 138) <0.001 0.41 ion type

gram)

4.55 (1, 69) 0.04 0.06

ask 7.34 (2, 138) 0.001 0.10 ion Type

gram)

>0.10

sk >0.10

reative

(fluency)

6.09 (1, 68) 0.02 0.08

reative

(originality)�creativity task 7.25 (2, 86) 0.001 0.10

ion type

gram)

>0.10

sk 10.81 (2, 122) <0.001 0.14

[(Fig._4)TD$FIG]

5.91 5.44

3.34

2.22

0.31 0.78

1.54 1.42 0.86 1.03

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

D ia

gr am Te

xt

D ia

gr am Te

xt

D ia

gr am Te

xt

D ia

gr am Te

xt

D ia

gr am Te

xt

Control-Flow Informa�on System

Data Technological Resources

Organiza�onal Resources

N um

be r o

f p ro

po se

d so

lu �

on s

Fig. 4. The influence of representation on types of process innovation solutions.

K. Figl, J. Recker / Information & Management 53 (2016) 767–786778

diagram and the text group. Participants in the diagram group produced more ideas related to information systems (MDia- gram = 3.34, SDDiagram = 2.09; MText = 2.22, SDText = 1.73) and fewer ideas related to data than the text group did (MDiagram = 0.31, SDDiagram = 0.80; MText = 0.78, SDText = 1.02). Individual creative competence is positively associated with the number of ideas related to control flow, data, and technological resources. The data did not provide support for Hypotheses 2a and 2b. Diagram users produced more control flow ideas but fewer organizational resource ideas, but neither difference was significant. In sum, the type of process representation influenced some but not all types of process- redesign ideas. Further empirical research and theorizing are required regarding the specific effects hypothesized.

We performed several supplementary analyses, summarized in Appendix D, for additional evaluation of our hypotheses. These analyses provide further insights into how the tasks, the representation format, and user demographics influence the process-redesign solutions produced in the experiment.

6. Discussion

We set out to determine the impact of diagrammatic process representations on creativity in redesign tasks for process innovation. Our findings indicate that diagrammatic process representations led to more creative process changes than textual representations do. The findings confirm a commonly held notion that diagrammatic process models are a useful aid to process analysts in designing future processes. Although these results demonstrate that diagrammatic models do not make analysts more creative or lead to a higher number of ideas, the redesign solutions offered appear to be beneficial in terms of dimensions such as appropriateness and type of idea.

Differences in the originality ratings of ideas generated appear to be sensible to gender and task selection. However, the level of originality did not differ significantly between text and diagram models, so our result differs from that of Ref. [31], who found higher originality of ideas with textual stimuli, and [46] who reported more creative ideas for users of pictures. A likely explanation for the differences is that originality depends on the type of visual/pictorial stimulus used, and process models cannot be compared directly with photographs or illustrations.

Our findings do not support the argument that process models evoke fixation and hinder the generation of creative, appropriate ideas, as other researchers have reported for pictorial stimuli in design tasks [8,37,79]. On the contrary, our results suggest that users develop a higher number of appropriate ideas when they work with

a diagram than with text. This outcome can be interpreted in light of Smith’s [78] suggestion to use paraphrasing for a problem setting to overcome fixation. Our study suggests that not only paraphrasing of the problem setting but also further transformation to a diagram- matic representation is helpful. Abstraction, two-dimensional (2D) structuring of information, and the additional use of symbolic vocabulary in the diagrammatic representation seem to support creative thinking about process redesign.

These results also align with Ward’s [89] observation that abstraction from the solution initially presented can help designers to escape fixation. Making the elements of a business process explicit – for example, the temporal and logical order of process activities in a diagram – can allow analysts to stick only to the essential elements of the business process.

This interpretation is also supported by the results of our third experimental group, the structured text group. As the results of the structured text representation format fell between the textual and the diagrammatic representation format for all dependent vari- ables, we speculate that both structure and symbolic vocabulary support abstraction, reduce working memory demands, and improve creative performance.

One unanticipated finding was that the number of ideas produced was similar among the experimental groups, which reflects other studies’ [8,46] findings that not the number, but the range of ideas was affected in idea-generation tasks that used textual or pictorial stimulus.

Our results on Hypotheses 2a and 2b, concerning the types of ideas generated, indicate that process models may both limit and expand the range of ideas produced. We found a significant effect of representation type for two types of ideas: information systems and data components of a business process. The increase in information system-related ideas in the diagram group could be attributed to the frequent use of process models in requirements engineering for information systems, while the lower number of data-related ideas could stem from the focus on process instead of data in the diagrammatic representations.

Measuring the semantic similarity of concepts [61] may help to explain the result through the associative theory of creativity [50] (Appendix E). Similarity measures based on WordNet [54] demonstrate that English concepts taken from the description of the information-system type of process-improvement ideas are somewhat closer to the concept of ‘‘diagram,’’ while concepts from the data type of description are closer to the concept of ‘‘text.’’ This analysis lends some support to the idea that diagrams are more closely associated with information systems, while data are associated more closely with text. However, the two explanations are similar: they show that an association between the repre- sentation’s format and the type of process-redesign ideas may stem from practice in information systems development or from a linguistic context.

7. Implications

7.1. Implications for research

To our knowledge, this study is the first to examine the effects of how information is represented (text versus diagrams) on process redesign as a creative problem-solving task. We identify three central contributions.

First, our study is the first to examine process models in the context of redesigning an organizational process. Thus, our work extends the stream of research on how well individuals understand differing forms of process representations (e.g., Refs. [53,59,67]). Our research sheds light on how the representation of the problem situation as a process model influences ideas for solutions. This contribution is important both in clarifying how textual or visual

K. Figl, J. Recker / Information & Management 53 (2016) 767–786 779

process representations influence process redesign and in clarify- ing the differences in solutions when analysts use either. Our findings also support the development of task-specific theories of process modeling and help to clarify how process models can be designed to be effective in differing types of task settings (such as in analyzing the performance of a current process versus designing future organizational realities, which is this paper’s focus).

Second, the paper contributes to the literature on process redesign as a creative problem-solving activity [66,81]. In particu- lar, we extend the literature on fixation effects in creative problem solving. In this literature, experiments have predominately provided participants with example solutions and have focused less on how the problem or the context is presented [79]. In addition, our paper adds to these research streams by accounting for the domain-specificity of creativity; we investigated a specific work-related task on process redesign, rather than using the architectural or mechanical design tasks on which several studies have focused [9,11,28,29].

Third, we offer a new, nuanced representation of process redesign in terms of the creativity dimensions of originality, appropriateness and impact, and the type of process-redesign ideas (locus of change: control flow, information system, data, and technological and organizational resources). This conceptualiza- tion and our newly developed measurement instrument can be used to guide researchers in evaluating business process redesigns.

7.2. Implications for practice

Our findings have implications primarily for process-improve- ment projects, a key focus of information professionals [24]. We addressed a long-standing debate about the relative merits of process modeling for process-redesign tasks and determined the type of process-redesign suggestions that can be expected when users work with diagrammatic or textual process descriptions.

One useful interpretation of our findings is that managers can, at least to some extent, guide the development of future processes by selecting a process representation format, that is, more or less conducive to producing changes to the control flow, data, resource, or technology components of a business process. Given that some or several components might not be a core focus (or, alternatively, might be taboo) in many projects, our results can aid managers in making decisions about the type of redesign solutions they wish to foster in their teams.

A second implication concerns the application of textual versus diagrammatic process representations for other kinds of tasks. We examined one task, process redesign, and our results suggest that the two kinds of representations influence the outcomes in process redesign. Some might interpret our results to mean that visual models serve this purpose better, but our findings do not suggest that textual diagrams are useless in such tasks or in other tasks. For instance, textual representations offer the advantage of packing in more contextual details than a diagrammatic representation can, which advantage might be helpful in tasks others than process redesign, such as in designing information systems support for the process. Additional contextual information may also be helpful in process redesign itself. Therefore, we suggest that organizations find an optimal trade-off between representation formats by considering the task setting in which process descriptions are used. Based on our results, where the results for the structured text lies between those for diagrams and texts (Fig. 3), the common practice of using diagrammatic process models supplemented by struc- tured process descriptions is useful. One implication of this observation is that organizations should maintain representations of processes in a variety of formats (e.g., text and diagram) and offer some or all of these representations to analysts, depending on the objective of a redesign project.

A final implication concerns the dependence of process-redesign solutions based on the type of redesign task to be undertaken. We noted that the type of process redesign varied not only across representation formats but also across the kinds of improvements required in the three tasks. This result suggests that managers should be conscientious about setting appropriate process-redesign objectives and suggesting process innovation procedures (e.g., using existing improvement patterns) to govern a project. Our results suggest that the choice of task objective acts as a focusing lens for analysts that will vary the solutions they generate.

8. Limitations

Our study has several limitations. The process representations used in the experiment are simplified versions of models used in practice, although the models themselves are not necessarily simple, and we restricted our investigation to one process scenario, so that we could examine the problem in a controlled setting. Therefore, external validity in the sense of being able to generalize the findings to other process scenarios (e.g., in terms of process complexity, type of domain, or extent of pre-existing knowledge) is limited. However, because our study is the first experimental study in this domain, internal validity was more important than other forms of validity. Therefore, we used a controlled experimental setting in a daylight computer laboratory because environmental characteristics such as the environment [34], the physical location (e.g., paintings and drawings in a room [9]), or even the simple presence of a light bulb [77] could influence performance in creative problem-solving tasks.

Second, our choice of a student sample limits the external generalizability of our results. Given our research objective, we selected a sample that was representative of future employees who would be concerned with process redesign. This selection strategy misses students’ groups who perform better in creativity tasks and practitioners with higher levels of domain knowledge, method knowledge, or more experience in creative problem solving. Concerning domain experience with the pizza-delivery industry, we refrained from including experts to avoid having participants with such elaborate understanding of the pizza-delivery industry that they might answer questions without looking at the stimulus of the process model. We included a variable that measured experience with pizza-delivery services as a customer, but we did not measure experience working in a pizza-delivery business because we assumed that a large population of such workers was unlikely in our sample. Still, we acknowledge that work experience may influence activities and outcomes in our setting. Our main reason for using a student sample was that bias related to knowledge about the method would be at most minimal. Senior business users and analysts would be more likely to be ‘‘brainwashed’’ into a particular representation format through years of training and/or practice. The study in Ref. [23] confirms such a preferential bias.

Third, the rating method for the ideas may have influenced our results. We pursued a multi-rater approach to mitigate subjective bias during the coding and to ensure that any influence would be consistent across all three groups.

Fourth, process innovation (and other forms of creative problem solving) is often conducted in groups to encourage idea composition and evaluation in the group. These interaction effects were not considered in this experiment, which focused primarily on idea generation by individuals. Follow-up work could consider group dynamics in this or subsequent phases.

Fifth, our coding of process-redesign solutions through the team of research assistants, while subjectively reliable, remains subject to interpretation. Our description of materials and processes ensured transparency in our chosen approach, but the

[(Fig._5)TD$FIG]

Structured Text Description of Process

pizza costumer

BEGIN

select a pizza from leaflet

order pizza by phone clerk pizza chef delivery person

redrohtiwetonetirw

nehctikotetonkcits

erutarepmetnevokcehc

hguoderaperp

esabotamotdda

sgnippotdetcelesdda

eseehcetarg

azzipekab

IF pizza = done

nosrepyreviledyfiton

IF setunim06>emitgnitiaw

azzipehtrofksa

calm costumer IF costumer not calmed

cancel order

ETANIMRET

FIDNE

ENDIF

ehtreviled pizza by car

azzipehteviecer

azzipehtyap

egnahcevig

tpieceretirw

END

Fig. 5. Control group treatment: structured text.

K. Figl, J. Recker / Information & Management 53 (2016) 767–786780

possibility of alternative interpretations remains. Replication research will be required to determine whether interpretation bias by our coders influenced the findings.

The generalizability of our findings is constrained by these limitations, especially in terms of external validity. Still, our study provides some insights into the process and outcomes of process redesign that could be useful in real-world settings. Our choice of research design was motivated by a desire to maximize internal validity, while maintaining some ecological validity. Internal validity was important because the practice of redesign is relevant and popular in today’s businesses, and research has so far relied largely on descriptive or observational studies. Our reasoning was that if we can detect differences on outcomes that are due to different representation formats in a controlled and simplified version using student subjects, then these effects will be even more significant in other, perhaps more realistic scenarios involving highly trained and experienced practitioners who use complex representations of key business processes.

Finally, our results are susceptible to the quality of the chosen representation. Our results suggest that visual process representa- tions may be superior to textual formats, but a badly constructed graphical representation may well be worse than a good textual representation. We described in detail how we constructed our materials and believe the quality was high for all three formats used.

9. Conclusion

The purpose of our study was to determine whether diagram- matic process models differ from textual representations in terms of how well they support analysts in developing creative process- redesign ideas. Our results suggest that diagrammatic models provide better assistance than text in terms of the generation of appropriate ideas. Our findings also suggest that users generate more ideas that are related to information systems and fewer ideas that are related to data when they work with a model than they do when they work with text. In general, participants with more individual creativity produced more ideas, independent of representation type.

Acknowledgment

Dr Recker’s contributions were supported by a grant from the Australian Research Council (DE120100776).

Appendix A. Experimental material used (selection)

Instructions

Imagine you are a business consultant, and a pizza-delivery service contacts you to help them improve their business processes. They provide you with a short description of their main business process, which is shown below. Please study this description carefully before proceeding (Fig. 5).

Your task as a consultant is to generate ideas on how to improve the process of the pizza-delivery service from several points of view. (These viewpoints could be – but do not have to be – cost, quality, turnaround time, customer satisfaction, increased market share, etc.). Please note that you do not have the complete information about the pizza-ordering service’s processes and that it is important to use your imagination. For each of the following questions, briefly describe as many improvement ideas as you can in the space provided. You do not need to make complete sentences when writing the ideas – just use simple phrases, and do not worry about grammar. You can use English and/or German. You have 5 min to complete each of the following three tasks, for a total of 15 min.

Process innovation tasks (creative redesign task)

Task 1: The pizza-delivery service wants to improve its processes so customers know at all times when their pizzas will arrive. How can the process be changed to implement this improvement? Write down as many options as you can think of.

Task 2: The pizza-delivery service is willing to hire a new employee. How could the employee be used to improve the process? Provide as many options as you can think of.

Task 3: The pizza-delivery service wants to cut costs. How could the process be changed to reduce costs? Provide as many options as you can think of.

Demographics

[TD$INLINE]

[TD$INLINE]

[TD$INLINE]

K. Figl, J. Recker / Information & Management 53 (2016) 767–786 781

Appendix B. Coding schema

Table B1

Criteria Subdimension

(where applicable)

Definition Explanation and examples

Creativity of

process-improvement idea

Fluency The quantity of relevant ideas provided. This attribute measures the ability to prod

of ideas that are relevant to the task instru

Originality Something that is original, unexpected, and novel. If the task is to list as many creative uses

possible, the answer ‘‘using a tissue as a n

original than the response ‘‘using tissues t

costume for the next Halloween party (e.g

fairy).’’

Appropriateness Something that is useful, meets task constraints, and is

purposeful.

Something can be original but not appropr

instance, a participant describes how to im

quality of pocket tissues instead of listing

creative uses for tissues as possible. These

be original, but they are not in line with th

requirements. Another example, knotting ti

to make a rope to escape from a fire on th

might be original but not appropriate, as th

tear and could catch fire easily.

Impact Measure of the positive tangible and intangible effects

(consequences) of one thing’s or entity’s action or influence

upon another. We refer to ‘‘impact’’ as the benefits that can

be derived from implementing a proposed solution, that is,

the profit or gain in monetary and/or nonmonetary terms

(e.g., cost or time savings, increased customer or staff

satisfaction, or other criteria).

Raising the price of cigarettes has a greate

immediate impact on the reduction of toba

than large antismoking campaigns do. Bein

active in old age has a significant positive im

the individual’s well-being (intangible bene

status (both intangible and tangible benefi

expenses for doctor visits and increased qu

Short-term solutions have a greater impac

term solutions because the benefits (tangi

intangible) can be collected immediately.

Type of

process-improvement idea

Locus of change* Measure of the key focus of the improvement idea. We refer

to the area that the improvement idea primarily addresses.

The relevant areas in relation to the pizza-delivery process

are as follows:

- Control flow: the nature, sequence, and order of the tasks

that need to be executed in pizza delivery (e.g., prepare

dough, bake pizza, and select toppings).

- Organizational resources: staff involved in the pizza-

delivery process (e.g., delivery person, pizza chef).

- Technological resources: tools and infrastructure involved

in the pizza-delivery process (e.g., oven, refrigeration, car).

- Information system: any computerized system that might

be involved in the pizza-delivery process (e.g., online

ordering system, short messaging services, pizza status

dashboard, electronic payment system).

- Data: any input or output information required or created

in the pizza-delivery process (e.g., recipe and pizza orders).

Hiring more experienced staff to execute ta

associated with the organizational resource

Using a web-based system for online pizza

information-system-related change. Elimina

assurance task is a change in the control flow

Buying a new oven is a technological resou

* Self-developed criteria, specific for the domain of process innovation.

7 8

2

Coding instructions

uce quantities

ctions.

Count the number of answers provided.

for a tissue as

apkin’’ is less

o make a

., ghost or

Evaluate each answer on a 5-point scale

(1 = not at all original, 3 = medium

originality, and 5 = very original).

iate. For

prove the

as many

answers may

e task

ssues together

e 30th floor,

e rope would

Evaluate each answer on a 5-point scale

(1 = not at all appropriate, 3 = medium

appropriateness, and 5 = very appropriate).

r and more

cco consumed

g physically

pact on both

fit) and health

ts, e.g., fewer

ality of life).

t than long-

ble and/or

Evaluate each answer on a 5-point scale

(1 = no positive impact at all, 3 = medium

positive impact, and 5 = very positive

impact).

sks is a change

s in a process.

ordering is an

ting a quality

of a process.

rce idea.

For each answer, identify whether the

focus of the change idea falls into any of the

five change areas: control flow, data,

information system, technological resource,

or organizational resource. Denote the locus

of change, only if one area is clearly the

most prevalent one (e.g., not if a change

simultaneously addresses data and

organizational resources).

K .

F ig

l, J.

R e ck

e r

/ In

fo rm

a tio

n &

M a

n a

g e m

e n

t 5

3 (2

0 1

6 )

7 6

7 –

7 8

6

Table C3 Descriptive statistics for dependent measures.

Mean SD

Fluency Diagram 4.01 1.84

Structured text 3.92 1.52

Text 3.65 1.37

Appropriateness Diagram 3.56 0.28

Structured text 3.50 0.29

Text 3.36 0.38

Originality Diagram 3.10 0.33

Structured text 3.07 0.28

Text 2.97 0.34

Impact Diagram 3.59 0.16

Structured text 3.54 0.18

Text 3.50 0.25

Control Flow Diagram 5.91 3.52

Structured text 5.65 3.00

Text 5.44 2.29

Information System Diagram 3.34 2.09

Structured text 3.05 1.53

Text 2.22 1.73

Data Diagram 0.31 0.80

Structured text 0.41 0.72

Text 0.78 1.02

Technological Resources Diagram 1.54 1.07

Structured text 1.51 1.28

Text 1.42 1.34

Organizational Resources Diagram 0.86 0.91

Structured text 0.97 1.07

Text 1.03 1.32

K. Figl, J. Recker / Information & Management 53 (2016) 767–786 783

Appendix C. Manipulation checks

First, we examined correlation statistics (Appendices C1 and C2) and the descriptive statistics for our key measures (Appendix C3). Correlations with significance levels p � 0.05 (two-sided) are printed in bold. Correlations are based on the subsamples of the diagram and the text group (n = 71). In particular, we examined the influence of domain-related experi- ence to determine whether the group difference introduced bias. To do so, we calculated Pearson’s correlations of the measure domain-related experience with the means of all dependent variables over the three tasks. Domain-related experience correlated with originality (r = �0.20, p = 0.05), as the more pizzas partici- pants had ordered, the less creative were their ideas.

Second, some studies reported gender-based influences on creative achievement [3]; hence, we examined gender differences in our results. Independent sample t-tests showed that gender influenced the number of control flow-related ideas (t = 2.20, p = 0.03) and the originality of ideas (t = 2.65, p = 0.01). Female participants developed more ideas concerning control flow (MFemales = 6.22, SDFemales = 3.00; MMales = 4.98, SDMales = 2.77), but their answers were rated less original (MFemales = 2.97, SDFemales = 0.30; MMales = 3.13, SDMales = 0.32).

Third, we examined potential effects of experiment fatigue. To avoid task-order effects, we used two orders of tasks. Subsequent t- tests showed that task order did not have a significant effect on the dependent variables, with one exception: the number of ideas related to organizational resources was related to task order (t = 2.79, p = 0.01). Overall, however, we argue that our experi- mental task setting was largely robust.

Table C1 Correlations of control variables with dependent measures (Pearson’s correlation coefficient).

Types of ideas Creativity of redesign

Control Variables Control

Flow

Information

System

Data Technological

Resources

Organizational

Resources

Fluency Appropriateness Originality Impact

Creative competence: fluency 0.34 0.12 0.13 0.29 0.07 0.36 0.19 0.05 0.07 Creative competence: originality �0.03 0.02 0.20 0.14 0.11 0.10 0.00 0.05 �0.11 Creative competence: elaboration 0.09 �0.03 0.07 0.18 0.04 0.11 0.10 �0.07 0.17 Creative attitude: intrinsic motivation 0.23 �0.05 0.05 0.09 0.05 0.16 0.12 0.14 0.08 Creative attitude: tendency toward

premature evaluation of ideas

0.13 �0.06 �0.03 �0.06 �0.10 0.01 0.00 �0.14 0.06

Task-related experience: experience

with process-improvement initiatives

�0.12 �0.07 �0.02 0.02 0.23 �0.05 0.00 0.14 �0.05

Domain-related experience:

number of times ordered a pizza

�0.09 S0.20 0.03 �0.04 0.23 �0.09 �0.15 S0.20 �0.03

Age �0.13 �0.11 �0.15 S0.27 �0.13 S0.25 0.12 0.13 �0.04

Correlations with significance levels p�0.05 (two-sided) are printed in bold.

Table C2 Intercorrelations of dependent measures (Pearson’s correlation coefficient).

Types of ideas Creativity of redesign

Control

Flow

Information

System

Data Technological

Resources

Organizational

Resources

Fluency Appropriateness Originality Impact

Types of ideas

Control Flow – 0.14 0.12 0.29 �0.01 0.75 �0.09 S0.25 �0.15 Information System – 0.07 0.32 �0.11 0.56 0.22 0.27 0.13 Data – 0.27 0.08 0.39 S0.35 �0.22 S0.29 Technological Resources – 0.27 0.68 �0.16 �0.13 �0.15 Organizational Resources – 0.28 S0.29 �0.10 S0.30

Creativity of redesign

Fluency – �0.14 �0.14 �0.20 Appropriateness – 0.60 0.75 Originality – 0.45 Impact –

Correlations with significance levels p�0.05 (two-sided) are printed in bold.

Table D1 Subsample analysis for gender effects.

Subsample Comparison Significant difference on dependent variable?

DV: originality DV: count of control flow ideas

Males Representation type (text vs. diagram) Yes/No (Fdf=1;38 = 3.97, p = 0.06) Yes/No (Fdf=1;37 = 0.01, p = 0.94)

Females Representation type (text vs. diagram) Yes/No (Fdf=1;29 = 1.59, p = 0.22) Yes/No (Fdf=1;28 = 0.00, p = 1.00)

Table E1 Similarity according to WordNet. The path-length measure gives the inverse of the

shortest path length between two concepts [61]. The maximum value is 1.

Types of

process-improvement

ideas

Word input Path length

‘‘Diagram’’

Path length

‘‘Text’’

K. Figl, J. Recker / Information & Management 53 (2016) 767–786784

Appendix D. Supplementary analyses

Post hoc analysis: task

Further analyses showed that the number of ideas related to data, information systems, and the organization differed among the three innovation tasks. For both sets of hypotheses and for all dependent variables, we found that either the within-subject effect of the creativity task or the interaction effect between individual creative competence and the creativity task was significant. Thus, our results confirm that the specified objectives for the task setting determine the number and types of ideas generated.

Post hoc analysis: gender

Our manipulation checks revealed that gender differences accounted for differences in the number of control flow ideas and originality. To determine differences between diagrams and text on the development of process innovation solutions, we performed subsample tests that kept the gender factor constant while varying the other factors (i.e., diagram or text). We performed this analysis for all of the affected dependent measures. Appendix D1 sum- marizes the results. The number of control flow ideas is not influenced by the representation type in neither the male nor the female subgroup. In both subgroups, the difference between the originality scores in the text and the diagram groups is not significant (MDiagram = 3.00; SDDiagram = 0.27; MText = 2.88; SDText = 0.33 for females and MDiagram = 3.30; SDDiagram = 0.36; MText = 3.04; SDText = 0.34 for males). This result might also be affected by the task objectives. A detailed look at the results at the task level revealed that, in tasks 1 and 3, gender and the type of

[(Fig._6)TD$FIG]

4.01 3.92

3.65 3.56 3.50

3.36

3.10 3.07 2.97

3.59 3.54 3.50

2.00

2.50

3.00

3.50

4.00

4.50

D ia

gr am

St ru

ct ur

ed T

ex t

Te xt

D ia

gr am

St ru

ct ur

ed T

ex t

Te xt

D ia

gr am

St ru

ct ur

ed T

ex t

Te xt

D ia

gr am

St ru

ct ur

ed T

ex t

Te xt

ImpactOriginalityAppropriatenessFluency

M ea

n s

co re

p er

g ro

u p

Fig. 6. The influence of representation (Diagram, Structured text, and Text) on the creativity of process innovation solutions.

representation are significant influencing factors for originality, but not in task 2. One reason for this result might be that tasks 1 and 3 have more room for being original than task 2 does, as task 2 specifically asked how to use an employee instead of how to change the process in general.

Post hoc analysis: structured text

Next, we examined our main results in light of the data collected on the intermediary representation format, ‘‘structured text,’’ which was included to ease interpretation. The main results of this post hoc analysis are summarized in Fig. 6, which demonstrates that, for all dependent variables for which we identified significant differences between the ‘‘text’’ and the ‘‘diagram’’ group, the results of the ‘‘structured text’’ group fall between the textual and the diagrammatic representation format. Values for structured text are based on a sample of 37 participants drawn from the same basic population as the two main experimental groups; see Appendix C3 for descriptive statistics for all dependent measures.

Appendix E. Associations of ‘‘Diagram’’ and ‘‘Text’’ with types of process-improvement ideas.

Control Flow ‘‘Sequence’’ 0.13 0.09

‘‘Order’’ 0.13 0.17

‘‘Task’’ 0.11 0.08

‘‘Control’’ 0.14 0.08

‘‘Process’’ 0.11 0.13

Organizational Resources ‘‘Organization’’ 0.08 0.11

‘‘Staff’’ 0.14 0.13

‘‘Employee’’ 0.09 0.08

Technological Resources ‘‘Technology’’ 0.06 0.08

‘‘Tools’’ 0.13 0.10

‘‘Infrastructure’’ 0.07 0.09

Information System ‘‘Information system’’ 0.13 0.10

‘‘Computer’’ 0.11 0.09

‘‘System’’ 0.14 0.13

Data ‘‘Data’’ 0.08 0.11

‘‘Input’’ 0.10 0.14

‘‘Output’’ 0.17 0.17

‘‘Information’’ 0.08 0.14

References

[1] T.M. Amabile, The social psychology of creativity: a componential conceptualiza- tion, J. Personal. Soc. Psychol. 45 (2), 1983, pp. 357–376.

[2] A.D. Baddeley, Working memory, Science 255 (5044), 1992, pp. 556–559. [3] J. Baer, J.C. Kaufman, Gender differences in creativity, J. Creat. Behav. 42 (2), 2008,

pp. 75–105. [4] M. Basadur, C.T. Finkbeiner, Measuring preference for ideation in creative prob-

lem-solving training, J. Appl. Behav. Sci. 21 (1), 1985, pp. 37–49.

K. Figl, J. Recker / Information & Management 53 (2016) 767–786 785

[5] P. Bera, Does cognitive overload matter in understanding BPMN models? J. Comput. Inf. Syst. 52 (4), 2012, pp. 59–69.

[6] Z. Bilda, J.S. Gero, The impact of working memory limitations on the design process during conceptualization, Des. Stud. 28 (4), 2007, pp. 343–367.

[7] M. Broadbent, P. Weill, The implications of information technology infrastructure for business process redesign, MIS Q. 23 (2), 1999, pp. 159–182.

[8] C. Cardoso, P. Badke-Schaub, Idea fixation in design: the influence of pictures and words, in: Proceedings of the ICORD 09: Proceedings of the 2nd international conference on research into design, Bangalore, India, 2009.

[9] H.P. Casakin, G. Goldschmidt, Reasoning by visual analogy in design problem- solving: the role of guidance, environment and planning, Plan. Des. 27 (1), 2000, pp. 105–119.

[10] P.-K. Cheung, P.Y.K. Chau, A.K.K. Au, Does knowledge reuse make a creative person more creative? Decis. Support Syst. 45 (2), 2008, pp. 219–227.

[11] B.T. Christensen, C.D. Schunn, The relationship of analogical distance to analogical function and preinventive structure: the case of engineering design, Memory Cognit. (pre-2011) 35 (1), 2007, pp. 29–38.

[12] M.M. Clapham, The convergent validity of the Torrance tests of creative thinking and creativity interest inventories, Educ. Psychol. Meas. 64 (5), 2004, pp. 828–841.

[13] A. Collins, M. Quillian, Retrieval time from semantic memory, J. Verbal Learn. Verbal Behav. 8 (2), 1969, pp. 240–247.

[14] A.M. Collins, E.F. Loftus, A spreading activation theory of semantic processing, Psychol. Rev. 82 (6), 1975, pp. 407–428.

[15] D.R. Compeau, B.L. Marcolin, H. Kelley, C.A. Higgins, Generalizability of informa- tion systems research using student subjects – a reflection on our practices and recommendations for future research, Inf. Syst. Res. 23 (4), 2012, pp. 1093–1109.

[16] B. Cramond, J. Matthews-Morgan, D. Bandalos, L. Zuo, A report on the 40-year follow-up of the Torrance tests of creative thinking: alive and well in the new millennium, Gift. Child Q. 49 (4), 2005, pp. 283–291.

[17] T.H. Davenport, J. Short, The new industrial engineering: information technology and business process redesign, Sloan Manag. Rev. 1990, pp. 11–27.

[18] F.D. Davis, R.P. Bagozzi, P.R. Warshaw, Extrinsic and intrinsic notivation to use computers in the workplace, J. Appl. Soc. Psychol. 22 (14), 1992, pp. 1111–1132.

[19] D.L. Dean, J.M. Hender, T.L. Rodgers, E.L. Santanen, Identifying quality, novelty, and creative ideas: constructs and scales for idea evaluation, J. Assoc. Inf. Syst. 7 (10), 2006, pp. 646–699.

[20] A.R. Dennis, G. Hayes, R.M. Daniels, Business process modeling with group support systems, J. Manag. Inf. Syst. 15 (4), 1999, pp. 115–142.

[21] M. Dumas, M. La Rosa, J. Mendling, H.A. Reijers, Fundamentals of Business Process Management, Springer, Berlin, Germany, 2013.

[22] R. Eisenman, Creativity and academic major: business versus English majors, J. Appl. Psychol. 53 (5), 1969, pp. 392–395.

[23] K. Figl, J. Recker, Exploring cognitive style and task-specific preferences for process representations, Requir. Eng. 21 (1), 2016, pp. 63–85.

[24] Gartner Group, Leading in Times of Transition: The 2010 CIO Agenda, EXP Premier Report, 2010.

[25] A. Gemino, D. Parker, Use case diagrams in support of use case modeling: deriving understanding from the picture, Database 20 (1), 2009, pp. 1–24.

[26] A.M. Glenberg, W.E. Langston, Comprehension of illustrated text: pictures help to build mental models, J. Mem. Lang. 31 (2), 1992, pp. 129–151.

[27] K. Goff, E.P. Torrance, Abbreviated Torrance Test for Adults (ATTA), Scholastic Testing Service, Bensenville, IL, 2002.

[28] G. Goldschmidt, A.L. Sever, Inspiring design ideas with texts, Des. Stud. 32 (2), 2011, pp. 139–155.

[29] G. Goldschmidt, M. Smolkov, Variances in the impact of visual stimuli on design problem solving performance, Des. Stud. 27 (5), 2006, pp. 549–569.

[30] M. Gonçalves, C. Cardoso, P. Badke-Schaub, Find your inspiration: exploring different levels of abstraction in textual stimuli, in: Proceedings of the 2nd international conference on design creativity (ICDC 2012), Glasgow, UK, 2012.

[31] M. Gonçalves, C. Cardoso, P. Badke-Schaub, How far is too far? Using different abstraction levels in textual and visual stimuli in: Proceedings of the design 2012: 12th international design conference, Dubrovnik, Croatia, 2012.

[32] J.P. Guilford, Three faces of intellect, Am. Psychol. 14 (8), 1959, pp. 469–479. [33] P. Harmon, Business Process Change: A Guide for Business Managers and BPM and

Six Sigma Professionals, Morgan Kaufmann, San Francisco, CA, 2007. [34] S. Hemlin, C.M. Allwood, B.R. Martin, Creative knowledge environments, Creat.

Res. J. 20 (2), 2008, pp. 196–210. [35] M. Indulska, P. Green, J. Recker, M. Rosemann, Business process modeling:

perceived benefits, in: S. Castano, U. Dayal, A.H.F. Laender (Eds.), Conceptual Modeling – ER 2009, Springer, Gramado, Brazil, 2009, pp. 458–471.

[36] P.W. Jackson, S. Messick, The person, the product, and the response: conceptual problems in the assessment of creativity, J. Personal. 33 (3), 1965, pp. 309–329.

[37] D.G. Jansson, S.M. Smith, Design fixation, Des. Stud. 12 (1), 1991, pp. 3–11. [38] A. Jeyaraj, V.L. Sauter, Validation of business process models using swimlane

diagrams, J. Inf. Technol. Manag. 25 (4), 2014, p. 27. [40] W. Kettinger, J. Teng, S. Guha, Business process change: a study of methodologies,

techniques, and tools, MIS Q. 21 (1), 1997, pp. 55–80. [41] N. Kock, J. Verville, A. Danesh-pajou, D. Deluca, Communication flow orientation

in business process modeling and its effect on redesign success: results from a field study, Decis. Support Syst. 46 (2), 2009, pp. 562–575.

[42] M. Kunze, A. Luebbe, M. Weidlich, M. Weske, Towards understanding process modeling – the case of the BPM academic initiative, in: Proceedings of business process model and notation, Lecture notes in business information processing 95, Lucerne, Switzerland, 2011.

[43] J.H. Larkin, H.A. Simon, Why a diagram is (sometimes) worth ten thousand words, Cognit. Sci. 11 (1), 1987, pp. 65–100.

[44] H. Leopold, J. Mendling, A. Polyvyanyy, Generating natural language texts from business process models, in: Proceedings of advanced information systems engineering, Lecture notes in computer science, Springer, Berlin, Heidelberg, 2012.

[45] J. Liu, S. Zhang, J. Hu, A case study of an inter-enterprise workflow-supported supply chain management system, Inf. Manag. 42 (3), 2005, pp. 441–454.

[46] R.A. Malaga, The effect of stimulus modes and associative distance in individual creativity support systems, Decis. Support Syst. 29 (2), 2000, pp. 125–141.

[47] D. Mani, A. Barua, A.B. Whinston, An empirical analysis of the impact of informa- tion capabilities design on business process outsourcing performance, MIS Q. 34 (1), 2010, pp. 39–62.

[48] R.E. Mayer, Multimedia Learning, Cambridge University Press, Cambridge, MA, 2009.

[49] J.M. McCoy, G.W. Evans, The potential role of the physical environment in fostering creativity, Creat. Res. J. 14 (3), 2002, pp. 409–426.

[50] S.A. Mednick, The associative basis of the creative process, Psychol. Rev. 69 (3), 1952, pp. 220–232.

[51] J. Mendling, Metrics for Process Models: Empirical Foundations of Verification, Error Prediction and Guidelines for Correctness, Springer, Berlin, Germany, 2008.

[52] J. Mendling, H.A. Reijers, J. Recker, Activity labeling in process modeling: empiri- cal insights and recommendations, Inf. Syst. 35 (4), 2010, pp. 467–482.

[53] J. Mendling, M. Strembeck, J. Recker, Factors of process model comprehension – findings from a series of experiments, Decis. Support Syst. 53 (1), 2012, pp. 195– 206.

[54] G.A. Miller, WordNet – a lexical database for English, Commun. ACM 38 (11), 1995, pp. 39–41.

[55] M.D. Mumford, K. Hester, I. Robledo, Chapter 3 – Methods in creativity research: multiple approaches, multiple levels, in: D.M. Michael (Ed.), Handbook of Orga- nizational Creativity, Academic Press, San Diego, 2012, pp. 39–65.

[56] L.A. O’Hara, R.J. Sternberg, It doesn’t hurt to ask: effects of instructions to be creative, practical, or analytical on essay-writing performance and their interac- tion with students’ thinking styles, Creat. Res. J. 13 (2), 2001, pp. 197–210.

[57] Object Management Group, BPMN 2.0 by Example, 2010 http://www.omg.org/ spec/BPMN/2.0/examples/PDF/10-06-02.pdf , (accessed 22.02.16).

[58] Object Management Group, Business Process Model and Notation (BPMN) (Ver- sion 2.0.2), 2013 http://www.omg.org/spec/BPMN/2.0.2 , (accessed 22.02.16).

[59] A. Ottensooser, A. Fekete, H.A. Reijers, J. Mendling, C. Menictas, Making sense of business process descriptions: an experimental comparison of graphical and textual notations, J. Syst. Softw. 85 (3), 2012, pp. 596–606.

[60] S. Patig, V. Casanova-Brito, Requirements of process modeling languages – results from an empirical investigation, in: Proceedings of the Wirtschaftsinformatik, 2011.

[61] T. Pedersen, S. Patwardhan, J. Michelizzi, WordNet::Similarity: Measuring the Relatedness of Concepts, Demonstration papers at HLT-NAACL 2004, Association for Computational Linguistics, Boston, MA, 2004 pp. 38–41.

[62] D. Piffer, Can creativity be measured? An attempt to clarify the notion of creativity and general directions for future research Think. Skills Creat. 7 (3), 2012, pp. 258– 264.

[63] J. Recker, Continued use of process modeling grammars: the impact of individual difference factors, Eur. J. Inf. Syst. 19 (1), 2010, pp. 76–92.

[64] J. Recker, Empirical investigation of the usefulness of gateway constructs in process models, Eur. J. Inf. Syst. 22 (6), 2013, pp. 673–689.

[65] J. Recker, M. Rosemann, E. Roohi Goohar, A. Hjalmarsson, M. Lind, Modeling and analyzing the carbon footprint of business processes, in: J. vom Brocke, S. Seidel, J. Recker (Eds.), Green Business Process Management – Towards the Sustainable Enterprise, Springer, Heidelberg, Germany, 2012, pp. 93–110.

[66] H.A. Reijers, S.L. Mansar, Best practices in business process redesign: an overview and qualitative evaluation of successful redesign heuristics, Omega 33 (4), 2005, pp. 283–306.

[67] H.A. Reijers, J. Mendling, A study into the factors that influence the understand- ability of business process models, IEEE Trans. Syst. Man Cybern. – Part A 41, 2011, pp. 449–462.

[69] M. Rosemann, Potential pitfalls of process modeling: part B, Bus. Process Manag. J. 12 (3), 2006, pp. 377–384.

[70] M. Rosemann, J. Recker, C. Flender, Contextualization of business processes, Int. J. Bus. Process Integr. Manag. 3 (1), 2008, pp. 47–60.

[71] M.A. Runco, Operant theories of insight, originality, and creativity, Am. Behav. Sci. 37 (1), 1993, p. 54.

[72] E.L. Santanen, R.O. Briggs, G.-J.D. Vreede, Causal relationships in creative problem solving: comparing facilitation interventions for ideation, J. Manag. Inf. Syst. 20 (4), 2004, pp. 167–198.

[73] J. Sarkkinen, H. Karsten, Verbal and visual representations in task redesign: how different viewpoints enter into information systems design discussions, Inf. Syst. J. 15 (3), 2005, pp. 181–211.

[74] M. Scaife, Y. Rogers, External cognition: how do graphical representations work? Int. J. Hum. Comput. Stud. 45 (2), 1996, pp. 185–213.

[75] A. Sharp, P. McDermott, Workflow Modeling: Tools for Process Improvement and Application Development, Artech House, 2001.

[76] A. Shtub, R. Karni, Business process improvement, ERP, Springer, US, 2010, pp. 217–254.

[77] M.L. Slepian, M. Weisbuch, A.M. Rutchick, L.S. Newman, N. Ambady, Shedding light on insight: priming bright ideas, J. Exp. Soc. Psychol. 46 (4), 2010, pp. 696– 700.

[78] S.M. Smith, Fixation, incubation, and insight in memory and creative thinking, in: S.M. Smith, T.B. Ward, R.A. Finke (Eds.), The Creative Cognition Approach, The MIT Press, Cambridge, MA, 1995, pp. 135–156.

K. Figl, J. Recker / Information & Management 53 (2016) 767–786786

[79] S.M. Smith, T. Ward, J. Schumacher, Constraining effects of examples in a creative generation task, Mem. Cognit. 21 (6), 1993, pp. 837–845.

[80] R.J. Sternberg, Cognitive Psychology, Wadsworth, Cengage Learning, Belmont, 2009.

[81] D.B. Stoddard, S.L. Jarvenpaa, Business process redesign: tactics for managing radical change, J. Manag. Inf. Syst. 12 (1), 1995, pp. 81–107.

[82] L. Sun, W. Xiang, C. Chai, C. Wang, Z. Liu, Impact of text on idea generation: an electroencephalography study, Int. J. Technol. Des. Educ. 23 (4), 2013, pp. 1047– 1062.

[83] S.X. Sun, J.L. Zhao, J.F. Nunamaker Jr., O.R. Liu Sheng, Formulating the data-flow perspective for business process management, Inf. Syst. Res. 17 (4), 2006, pp. 374–391.

[84] S.M. Turner, S.T. DeMers, H.R. Fox, G. Reed, APA’s guidelines for test user qualifications: an executive summary, Am. Psychol. 56 (12), 2001, pp. 1099–1113.

[85] G. Valiris, M. Glykas, Critical review of existing BPR methodologies, Bus. Process Manag. J. 5 (1), 1999, pp. 65–86.

[86] W.M.P. van der Aalst, M. Rosemann, M. Dumas, Deadline-based escalation in process-aware information systems, Decis. Support Syst. 43 (2), 2007, pp. 492– 511.

[87] W.M.P. van der Aalst, A.H.M. ter Hofstede, B. Kiepuszewski, A.P. Barros, Workflow patterns, Distrib. Parallel Databases 14 (1), 2003, pp. 5–51.

[88] I. Vessey, R. Weber, Structured tools and conditional logic: an empirical investi- gation, Commun. ACM 29 (1), 1986, pp. 48–57.

[89] T.B. Ward, What’s old about new ideas? in: S.M. Smith, T.B. Ward, R.A. Finke (Eds.), The Creative Cognition Approach, MIT Press, Cambridge, MA, 1995, pp. 157–178.

[90] M.A. West, J.L. Farr, Innovation and Creativity at Work: Psychological and Orga- nizational Strategies, John Wiley, Chichester, 1990.

[91] K. Whitley, Visual programming languages and the empirical evidence for and against, J. Vis. Lang. Comput. 8 (1), 1997, pp. 109–142.

[92] W.D. Winn, An account of how readers search for information in diagrams, Contemp. Educ. Psychol. 18 (2), 1993, pp. 162–185.

[93] L. Zeng, R.W. Proctor, G. Salvendy, Can traditional divergent thinking tests be trusted in measuring and predicting real-world creativity? Creat. Res. J. 23 (1), 2011, pp. 24–37.

[94] M. zur Muehlen, Organizational management in workflow applications – issues and perspectives, Inf. Technol. Manag. 5 (3), 2004, pp. 271–291.

Kathrin Figl is an Assistant Professor at the Institute for Information Systems and New Media at the Vienna University of Economics (WU). Most of her research and

teaching focuses on human-centered development and design of information

systems. She is especially interested in evaluating conceptual modeling languages

and methods and investigating human interaction with models. She has published

more than 60 research papers and articles, among others in the Journal of the

Association for Information Systems, Decision Support Systems, Requirements

Engineering, the Journal of Visual Languages and Computing and the International

Journal of Human-Computer Studies.

Jan Recker is Full Professor of Information Systems and Retail Innovation at Queensland University of Technology. Jan’s research focuses on processes-oriented

systems analysis and design, Green IS and IT-enabled innovation. He has published

in journals such as MIS Quarterly, Journal of the Association for Information Systems,

Information & Management, Journal of Information Technology and European Journal

of Information Systems. He is Editor-in-Chief for the Communications of the

Association for Information Systems and an Associate Editor for the MIS Quarterly.