psychology essay

Minfu Liang
04_RepresentationandInformationProcessing.pdf

10/14/2019

1

Information Processing in the Brain

Information Processing

• Sticking with the computer metaphor provides us with concepts such as: • Processing (ease/difficulty of processing)

• Algorithm

• Modularity

• Serial / parallel processing

• Local / distributed representation

Algorithms and Heuristics • Algorithms

• Strict step-by-step rules, solution guaranteed given correct algorithm • Might not always be the “best” method, in terms of processing

power and speed • Not good for tasks that are highly variable, or tasks where the

solution is unclear (such as creative tasks)

• Heuristics • Rules of thumb • Solution is likely • Typically more efficient than algorithms

• Computers – algorithms • How computer programming traditionally works (Line-by-line

coding; if/then functions; logic gates) • Traditional computers don’t like heuristics – need to explicitly tell

computer EXACTLY what you want it to do, and it will only do exactly what it has been told

• Humans – heuristics primarily, when appropriate

10/14/2019

2

Example: Solving a Maze

Modularity • Fodor (1983) “The Modularity of Mind”

• Basically – fast, automatic, predictable, discrete, and isolated

• What (if any) mental processes are modular?

Serial and Parallel Processing

• A serial process occurs in a linear order, with the next operation occurring after the previous has finished.

• A parallel process occurs simultaneously, with all operations occurring at the same time.

10/14/2019

3

Local and Distributed Representation • Local:

• Concepts are stored and represented as one coherent unit.

• “Grandmother cell” – one neuron (or local group of neurons) responds when seeing your grandmother (and nothing else)

• Distributed: • Concepts are defined by their

pattern of activation across many smaller units.

• Barcode

Local Representation

Is a dog Is a cat Is a fish Is Netflix

Dog x

Cat x

Fish x

Netflix x

Dog 1000 Cat 0100 Fish 0010 Netflix 0001

Distributed Representation

Is alive Has fur Eaten in Santa Cruz

Fun to interact with

Dog x x x

Cat x x

Fish x x

Netflix x

Dog 1101 Cat 1100 Fish 1010 Netflix 0001

10/14/2019

4

Connectionism

• PDP – Parallel distributed processing

• Inspired by neuroanatomy of the cortex • Nodes – units of info – analogous to neurons

• Connections between nodes – excitatory or inhibitory

• Concepts are built from distributed representations of smaller units activated in parallel

• Networks learn to associate input with desired (or taught) output

• Moving away from “classic” computer metaphor (but still computational)

Pandemonium Model • Earliest “connectionist” model (Selfridge, 1959)

10/14/2019

5

Challenges to the Computer Metaphor • Parallel Distributed Processing / Connectionism

• Embodied Cognition • Abstract concepts and processes are understood

through sensory and bodily experience

• Situated Cognition • What happens inside the head is only one piece of the

story – environment is part of cognition (calculators)

• Dynamic Systems • Emphasis on time, context, and interacting subsystems

• Mental processes as continuous, not discrete

Representation

Mmm…donut…

->

“Donut” (/ˈdoʊnət/) Edible and tasty Source of delicious nutrition Mmm…donut…

Physical object Mental representation

10/14/2019

6

Perception

Representation

Action

Representations

• What is a representation? • Internal model linked to external stimuli or information

• Two possibilities for mental representation • Analog representation – sensory (for visual information,

depictive)

• Propositional representation - symbolic

• Philosophical divide in cognitive psychology • Emphasis on propositional representation – symbolic

processing, computer metaphor, “classic” cog

• Emphasis on analog (or even lack of) representation – embodied cognition

Analog Representations

• Representation has same relationships as properties of referent

• Tied to sensory modality (in this case, visual)

• Non-symbolic representation

A B

D

C

Referent

A B

D

C

Analog Representation

10/14/2019

7

Propositional Representations

• Abstract assertions that maintain relationship of referent

• Not tied to a particular sensory modality

• Symbolic representation

A B

D

C

Referent Propositional Representation

“A” is in lower left “B” is right of “A” “D” is above “A” “C” is further away “C” is highest Etc. etc. etc.

How do I add?

1 red apple 1 green apple

2 apples

Also just

1+ 1 = 2

Propositional additionAnalog addition

Analog Advantages

• Principle of least commitment • Analog representation does not presuppose that certain

aspects of a stimulus are not important

• When encoding propositional representation, only encode what is considered relevant

• When recalling depictive image, “extra” information can be gathered

• Counterfactual simulation • By manipulating and combining analog representations,

can evaluate new scenarios that have never been directly experienced or learned

10/14/2019

8

Propositional Advantages

• Communication and language • Since language is a symbolic system, we can

communicate propositional information to each other through words

• How can we “communicate” analog representation directly?

• Parsimony • Flipside of principle of least commitment – propositional

representation can selectively represent important information, neglecting irrelevant information (perhaps saving on processing power and space)

Heterogenous Representation

• Does it have to be either/or?

• Can humans represent information both in analog and propositionally?

• Did the development of language allow for humans to utilize propositional representation?

• When do we represent information propositionally and when do we represent it in analog?

• Can we determine this experimentally?

Memorize this Map!

10/14/2019

9

Kosslyn, Ball, & Reiser, 1978

Mental Rotation

• Are the two objects identical or mirror images?

• Measure response time and degree of rotation for identical items

10/14/2019

10

Mental Rotation

• Response time varies linearly as a function of degree of rotation!

Shepard & Metzler, 1971

Imagery

• Relative size and imagery • Imagine a rabbit standing next to an elephant

• Does the rabbit have two front paws?

• Imagine a rabbit standing next to a fly • Does the rabbit have two ears?

Kosslyn, 1975

10/14/2019

11

Imagination and Perception

• Evidence leads us towards analog representation, but to what degree do perception and imagination overlap?

• Mental images are not generated from typical external stimulation • “Pure” form of mental representation?

• How does this process compare to normal perception?

• Compare fMRI during imagery and during perception of same object

Dijkstra, Bosch, & van Gerven, 2017

Back to philosophy… • Realism

• We perceive the actual object • My conscious awareness of an object is of that direct

object

• Representationalism • We perceive the mental representation cued by the

object • My conscious awareness of an object is my

representation of that object (indirect realism)

• In both cases, we are perceiving a “real” object – they differ in the specifics of representation • Yes, this is a tough question. • If we believe in representation, do we need to be

representationalists?

10/14/2019

12

Cognitivism and Representationalism

• Let’s take the side that mental representation is real and can be propositional

• Perception -> representation -> action

• We “reside” in a representation of the world? • Not STRICTLY reality

• But not arbitrary (partially subjective)

• Representation cued from reality, and we do a great job of it (mostly)

Symbolic Processing

• If representation is propositional, it can be expressed as a series of symbols

• Language is a system of symbols – so if we can explain it using language, it can be converted to OTHER symbol systems • Mathematical representation

• Computational representation

• IF the brain works using symbolic processing, then the computer metaphor is good and we can build an AI using traditional programming methods

Why might we want to avoid assuming symbolic

representation?

10/14/2019

13

The Chinese Room

• Searle – Where does meaning arise in an information processor?

• Symbol manipulation is not sufficient to explain cognition.

The Symbol-Grounding Problem

• Basic idea behind Chinese room – symbols must be “grounded” in what the actually mean • Without grounding, there can be no understanding

Describing color to a blind person

10/14/2019

14

The Symbol-Grounding Problem

• Symbols can be grounded through perceptual experience • For someone without visual perception, the word “blue”

does not have the same meaning as someone who has experienced blue

• Perception is the “language” of the mind – and language symbolically refers to perceptual experience

• Note that since we perceive our own actions (and generate them obviously), action can also be included in this non-symbolic “language” of the mind

• What about abstract concepts that don’t directly relate to sensorimotor experience?