cognitive psychology short essay

12/4/2019

1

Constitution, Complexity, and

Dynamic Systems

Constitution Hypothesis

• What “constitutes” cognition?

• Standard cog – cognition is brain only

• Conceptualization – cognition is brain and body

• Constitution – cognition is embodied, extended, and embedded (in addition to utilizing brain) • Mental addition is cognitive • Counting on fingers is cognitive • Doing long division on a piece of paper is cognitive • Using a calculator is cognitive

Conceptualization vs. Constitution • Conceptualization hypothesis

• Cognition is constrained by sensorimotor processes; we cannot conceptualize that which we cannot directly perceive/experience/act on

• Mental representations must be analog

• Constitution hypothesis • Cognition utilizes sensorimotor processes to develop

concepts and representations • Embodiment is a constituent of cognition, and is more heavily

emphasized early in development (learning of sensorimotor contingencies)

• Mental representation can be propositional, once link between symbol and grounded process are learned

• Perceptual experience depends on knowledge of sensorimotor contingencies (mirror neuron system), which can be represented propositionally (once learned)

12/4/2019

2

Constituent vs. causal contributor • Where do we draw the line between something that

just impacts cognition and something that is a part of cognition? • Head/eye movements

• Clearly impacts cognition through altering input of visual data • But tight perception/action loop between eye movements and

cognition implies that it goes the other way also – cognition impacts how we move eyes

• Process of visual cognition and eye movements coupled together – can’t have one without the other – therefore it is a part of cognition

• Is this a semantic argument or something deeper?

• Constitutionalists claim something deeper. “Mind” as we know it becomes fundamentally different as body and environment change

Constituent vs. Cause

• Another way of looking at the debate: • Causal processes are

independent, once cause is applied • Process A causes process B, but

process B does not cause (or effect) process A

• Constituent processes are coupled together • One process can not be

described without explicit reference to other process

• In mathematical terms, describing one requires term representing other

• Spontaneous synchronization

https://www.youtube.com/watch?reload=9&v=T58lGKREubo

12/4/2019

3

Gesture and Constitution

• Recap on gesture… • Improves lexical access

• Improves spatial cognition

• People do it even when others can’t see them gesturing

• Gesture is a constituent of cognition? • Does cognition produce gesture? Yes

• Does gesture produce cognition? Yes

• So under this view, processes are coupled together and thus gesture is a constituent of cognition

• What exactly is the role of gesture?

Gesture and Constitution

• Gesture being used to offload cognition • Simplifies work of brain by

offloading WM onto body • Provide people with pencil and

paper while solving problems – now extend cognition using paper, not gestures

• Two ways of talking about this • Gesture aids cognition, just

like paper and pencil • Gesture is a part of cognition,

just like paper and pencil • Cognition is now extended

beyond brain (and even beyond body!)

Tversky, 2009

Wide Computationalism

• Cognitive off-loading

• Argument for wide computationalism • Cognitive processes are computational

• If the computational processes that comprise some cognitive systems have constituents outside the head, then these cognitive systems extend outside the head

• The computational processes that comprise some cognitive systems do have constituents outside the head

• Therefore, some cognitive systems extend beyond the head

12/4/2019

4

Supporting Extended Cognition • Some stuff we have mentioned that can be taken as

support for constitution/extended cognition • Using tools changes body schema • Relying on maps associated with hippocampal changes

• Some stuff we didn’t mention coming out of UCSC, but is quite relevant • Offloading onto internet resources (i.e., searching Google)

encourages future internet use and discourages storing information in memory (Storm et al., 2017)

• Taking photos of events reduces memory of those events (even when photos are “temporary”; Soares & Storm, 2018)

• If these processes are occurring in order to reduce processing necessity of brain, they are technically adaptive (but might come with unwanted consequences!)

Replacement Hypothesis • Most extreme form of

embodied cognition with respect to departure from standard cognitive science

• Brain is not a computer, because brain is not a symbolic processor, is not discrete, and does not use representation

• Brain is a dynamic system and should be conceptualized as such Centrifugal Governor

What do replacement theories look like? • Evolution of behaviorism – pushback against

“computational” cognition in general

• No representation (at least, emphasis away from representation to greatest degree possible)

• Emphasis on interacting units and subsystems

• Emphasis on how things change and manifest over time (dynamic processes)

• Emphasis on learning and development

• Example: A-not-B error

12/4/2019

5

A-not-B Error

https://www.youtube.com/watch?v=4jW668F7HdA

What do replacement models look like? • A-not-B error

• You’ve read about this, but as a recap • Standard cog assumes that child has incomplete object

permanence concept (can’t yet guide behavior) • Dynamic approach emphasizes reinforcing behavior

based on prior successes (also a somewhat predictive account) with NO appeal to object “concept”

• Dynamic model better explains data than traditional (e.g., effects of waiting time, hidden object location)

• Remember the outfielder problem? • Eliminate complex mental representational structures by

emphasizing role of perception-action in an environment rich with dynamic information

https://www.youtube.com/watch?v=kGzJJpI0u3Y

12/4/2019

6

Dynamic Research

• Since processes are dynamic, need to have dynamic (continuous) measurements • Have to go beyond simple RT – need to look at time-course of

process • Motion tracking very useful here (eye/mouse/lip)

• Example: spoken word perception • From a dynamic standpoint – perception of spoken word is a

continuous and on-line process, moving towards attractor basin in real time

• So if you continuously measure where people “are” in a state space landscape, you should see some sort of continuous approach to the target

• Compare situation where target is determined earlier vs. later and compare trajectories of mouse movement to click on target

Spivey et al., 2005

Candle

Candy

12/4/2019

7

Energy Landscape of the Mind

• The “mental landscape” is a multidimensional state space (energy landscape)

• This space contains attractor basins of particular neural patterns that correspond to concepts

Energy Landscape of the Mind

• When concepts are more familiar, more learned, more processed, “valley” on energy landscape becomes deeper and steeper (easier to reach “stable” state) • Examples: frequency effects in word naming, prototype

categorization, canonical viewpoint object naming, blocking effects during recall

• Why “stable” in quotes? Mind is dynamic, constantly moving around energy landscape. Never settles in one valley, always moving from point to point.

12/4/2019

8

Dynamic Consciousness

• What happens when there is no changing input? Does the mind resist “settling”? • Yes!

• Visual blankout • Ganzfeld procedure • Sensory deprivation

• Interesting thought – how does this relate to deep meditation? Quieting the mind? • Note also – once meditation is

practiced, easier to reach desired state

• Is this like “balancing” between valleys?

The Representation Debate

• We’ve seen this before, largely in the form of analog vs. propositional representation

• Of course, another possibility – NO representation

• Where is the “representation” in the centrifugal governor?

Cognition without Representation

• Perception-action loop • As long as organism is in constant contact with

environment, environment does not need to be represented

• Coupling of brain, body, environment

• Outfielder problem – no calculation, no representation needed

• Gibsonian vision – no representation needed for depth

12/4/2019

9

Cognition without Representation

• How do you explain lack of representation when discussing memory and imagination?

• Describe an apple’s shape. • There is no physical apple, so

what are you describing based off of?

• Can this be explained with NO representation? How?

• Some things might just require representation to do, and some things might not

Successes and Failures of Replacement • Successes:

• Idea of perception/action loop as central component of perception and action itself is pretty solid

• Roboticists and connectionists appreciate the emphasis on interacting subsystems over time and emergent structure from simple units (lots of success when it comes to designing robots to perform perception/action style tasks)

• Fits in nicely with complex systems science, which is supposed to be a fairly “universal” tool for describing any complex system

• Failures: • Complete denial of any and all representation is a tough pill

to swallow for cognitivists, especially psycholinguists • How do we explain higher level cognition (symbolic

processing, propositional representation, language, etc.) with no concept of representation whatsoever? • Note – this is not necessarily impossible; but it is an uphill battle

and might need to be weakened

One other theoretical success of dynamic systems – link with complexity science

12/4/2019

10

Complex Systems Science

• Systems with units and interactions between units can self-organize into a larger emergent structure

• We call these things complex systems • As opposed to simple systems (prediction 100%ish)

• And chaotic or non-deterministic systems (prediction 0%ish)

• Complex systems are deterministic, but often unpredictable (prediction rate varies)

• Multiple components interacting over time – dynamic interactions

https://www.youtube.com/watch?v=g5evD6AQeCQ

Complex Systems Science

• Hallmarks of complex systems • Heterogeneous, independent agents

• Connections between agents (physical (or not), spatial, temporal)

• Self-organization (no “leader”)

• Emergence

• Fractal structure (and power law distribution)

• Criticality (teetering on the edge between simple and chaotic)

• If we see this stuff occurring in some sort of system, it’s a complex system – and this all occurs in the brain itself!

12/4/2019

11

Fractal Structure

• Fractal – self-similar recursive object • Each level is the same

structure as previous/next

• Scale-free

• A triangle made of triangles made of triangles made of…

• Sometimes extends into infinity, other times clearly defined

• Fractals are everywhere

Fractals Everywhere

• Fractals follow a general power law distribution • For every big “part”, lots

of tiny “parts” • Scales down

exponentially at each level

• When we see a process that follows a power law, it might be because of underlying fractal structure!

12/4/2019

12

Zipf’s Law

• Zipf’s law • Power law distribution of the frequency of words within

a language

• Similar idea – 80/20 rule • 80% of effects come from 20% of causes

The bad news about power law distributions… • Take something like income (or wealth) distribution

• Also follows a power law

• As more people are put into the system, tail extends further and further out

• In a free market (a complex economic system), “ultra-wealthy” and inequality are inevitable?

12/4/2019

13

A complex model of traffic

• Agents – drivers (or cars if you want)

• Connections – interactions between driver and car and other drivers/cars

• Rules – Don’t hit other cars or objects. Don’t drive on the shoulder. Don’t drive dangerously. Don’t speed. Don’t run red lights. Etc. etc.

• Goal – get home as fast as possible!

• What happens when too many cars follow these rules without enough road?

https://www.youtube.com/watch?v=iHzzSao6ypE

Another Complex Model - Segregation • What leads to cities (or microcosms) being

segregated? • Probably lots of causes in the real world, but let’s focus

on just one – personal preference to be with others “like” you

• Schelling’s segregation model • Agents – individuals (blue or orange)

• Connections – view color of other individuals

• Rules – ensure that you are always around __% of same color

12/4/2019

14

Another Complex Model - Segregation • When does segregation

emerge? • 90% preference? 75%

preference? • We can model it! Use an agent-

based model • Netlogo demonstration

• Segregated neighborhoods emerge even when individual similarity preference is around 50%

• Going back to energy landscapes – given a starting point, network settles in valley

The Brain as a Complex System

• Agents – neurons (entire nervous system)

• Connections – synapses

• “Rules” – fire when threshold potential reached

• Abilities – adjust connection weights

• Goal of the system? Keep itself going. • Need to get food and nutrients • Need to reproduce • This things require a majorly important sub-goal – perceive and

act in the world (to be successful…minimize prediction error)

• Can the complex mind be generated by simple units?

• Do we need to assume representation and algorithmic style processing?

Power Laws and the Brain

• If the brain is a complex dynamic system, we should see power law distributions manifesting in various cognitive activities • Visual search and Levy flight

• Language use and Zipf’s law

• Power law distributions of RT tasks (often times, outliers thrown out…making data appear more normal!)

• Sparse coding • Many neurons rarely spike beyond resting rate

• Looks a lot like 80/20 rule

• Some validity to the “10% of your brain” myth??? (maybe more like…20%?)

12/4/2019

15

Conclusion

Approaching the End • Evaluating embodied cognition

• Standard cognitive science has inspired much interesting research and informative theories about the nature of the mind

• Embodied cognitive science has also inspired new empirical research • Who would think to do a lot of the studies we talked about if not

for the theoretical idea that cognition can be embodied?

• Is standard cognitive science dead? • Nope.

• Embodiment has a lot of successes, but also a lot of challenges • Standard cog theories still widely accepted, research methodology

still frequently used

• But much has been learned about visual and spatial cognition through approaching problems from an embodied perspective!

Goals of this course… • Outline the basics of our visuospatial perceptual systems

• Establish and evaluate the two main competing theories of representation and cognition – symbolic and non- symbolic / embodied.

• Investigate cognitive processes and how they relate to the embodied experiences of visual and spatial perception

• Explore the symbol grounding problem and the nature of “understanding” something

• Discuss the current and future state of cognitive science as a discipline

• Expand your mind dude

12/4/2019

16

Where are we now? • Clear role for embodiment in many cognitive processes,

but specific nature of the relation is still hotly debated • Cognition utilizes sensorimotor resources and

knowledge to conceptualize, think, and act in the world • Visual and spatial perception are part of many

additional cognitive processes • There is no easy answer to the question of if the brain

is primarily a symbolic processor or something else, but interesting work is coming from both camps

• Empirical studies can help to disentangle complex philosophical arguments about the nature of the mind, and new theories/models guide new research

• Lots of interesting process has been made to understand cognition, yet there is still much work to be done!

Thank you!