Cog Sci - Module 4 - The Computational theory of mind

Lesion Deficit method

- Broca’s Area, Wernicke’s Area 

fMRI 

• FFA (Fusiform Face Area) 

• Computational modeling 

Split Brain studies

• Language production highly left lateralized 

• But there is some language in the right so much that he can read the words on the page with the right brain/using the right brain 

Limits of nueroscience 

• What really want to know: 

  • What information does the region represent? 

  • What is the causal role of that region’s response in behavior 
    

Activity in brain

• Carries information about what we are seeing 

• Ex: tells whether its a face or tree or car (ex: Fusiform face area) 

• We still don’t know the internal details of it though 

“Big Data Alone aren’t enough.” 

• We still cant understand the mind fully just by looking at the brain 

• we may just be overanalyzing things/activity that is not even responsible for certain activity/actions in the brain 

• Even physicalists 

• We have to ask new questions, learn how to categorize things in brain 

Scientists did a test on a microprocessor that supported three games:

Games:

• Space invaders 

• Pitfall 

• Donkey kong 

Test/example: zapped specific transistors on microprocessor 

• Space invaders still plays, Pitfall still plays

• Donkey kong no longer plays  

• We might say that that transistor supports/does donkey kong 

• We are mistaken because thats like saying Broca’s area does/is solely  responsible for language 

We did tests on the microprocessor, TMS, and other stuff, but still couldn’t understand it from that 

• But on the other hand we understood the system perfectly because we designed it/knew the game logic

Even using                         and                     from neuroscience we still fall short from understanding the brain completely 

methods and reasoning 

Sameness Ex: car accident

Everyone does the same thing when responding to a car accident/emergency  even though it looks different across everyone’s unique perspective/experience 

Ex:  where they grabbed their phone from, what they said when they called, etc.

But If you try to just boil it down to physicalizations the small stuff get lost, but we need the details 

• Similar to how well study the mind 

Computation in a word

 perception, memory, language, emotions, creativity, reason, decisions

What is a computation? 

• We count something as a computer because, and only when, its inputs and outputs can be usefully and systematically interpreted as representing the ordered pairs of some function that interests us

Pancomputationalism 

• All sufficiently complex physical systems perform computations 

  • Hurricanes, solar systems, even rocks 

Physical Symbol System/Formal Systems

takes symbols, combines them into expressions, and manipulates them using processes.

Formal systems example: Logic

• Symbols -  p and q 

• Expression -  p, q, p → q

• In formal logic expressions are things that can be true or false 

• Process - of deduction, that allows us to conclude q 
  

Formal systems example: Chess

• Symbols - pieces 

• Expressions- board state, how pisces move 

• Processes- moves you can make to take u from one board state to the next 

A Formal system is … 

• Independent from any particular instantiation 

• Chess: Independent from the media that you’re playing 

• Can play online, wood, plastic, glass (and it would still be Chess) 

A Formal system is … also

• More abstract than a physical instantiation 

  • Level of abstraction - not a particular physical thing, something that can be manifested across different physical things 

Andy Clark POV

 The mind is not like a computer. The mind is a computer 

The mind is a computer and we’ll study it according to those systems 

David Clark POV 

- The mind is very complicated, and we need to study it in complicated ways 

In Cognitive science 

• We need to study the mind at different levels

• Different parts of it 

• To understand it as a whole 

Levels of analysis

Level number 1 - most abstract level 

Computational level 

Level 2 - basic/intermediate level 

Algorithmic level

Level 3 - most specific

Implementational level 

Computational level

• What is the goal of computation? 

  • What is it taking in and putting out/and why 

• Why is it appropriate? 

• What is the logic of the strategy- rules of play? 

Algorithmic level 

• How are the input and output represented? 

• What specific steps are taken in the process from input to output?

Implementational level 

• How is the computation realized in a physical system? 

• What physical structures and processes are involved? 

• Whats the relevant level of description? 

Cash register: Implementation level 

asks where?

 (Basically looking at parts of the machine- where do processes take place)

• Makeup of cash register 

• Internal tape 

• Material the keys are made of, material the screen is made of 

Cash register: Algorithmic level 

asks How? 

(What does the machine do? Add!)

• Goal of system: Adding - its an adding machine   

• Adding machine 

• Representation - arabic numerals (1, 2, 3, 4…) 

• Algorithm - Steps like adding the least significant digits first and carrying the 1 if the sum exceeds 9 

Cash register: Computational level level 

asks what? and  why? 

Why addition? - our real goal is to combine prices and addition is perfect for the job 

Commutativity 

•  What - (3 + 4) and (4 + 3) both equal 7 

• Why - It shouldn’t matter what order your goods are presented to the register 
  
  

Associativity 

•  What - 3 + (4 + 5) is the same as (3 + 4) + 5 

• Why - Arranging goods into different piles shouldn’t affect how much you pay

Algorithmic level - theory of addition

Commutativity 

• 3 + 4) and (4 + 3) are both equal to 7 

• Doesn't matter what order you put numbers in

Associativity 

• 3 + (4 + 5) is the same as (3+4) + 5

Cash Register - full levels of analysis

What? Why? -Computational Level:

• What - addition 

• Why - well suited for combining prices

How? Algorithmic level: 

• Rep -  Arabic numerals 

• Algo - Least digit first, carry the 1, etc 

Where? Implementation Level:

• The cash register itself

Why we need all three levels of analysis

• we need to know how we get from talking about an algorithm to actually completing that algorithm/analysis 

Computational description 

• Adding 

Physical description/Implementation

• Describes us doing the adding 

We want to know whats in between, how were doing all this stuff 

• We need all three of these levels to figure it out  

• Explain and connect across levels

Marr’s argument

• in order to understand the brain of anything you must look at all levels 

• Cannot know what its doing if you just look at beginning, middle, and end, 

  • Need to understand and consider all 3 

•  in order to understand the brain of anything you need to take the computational method system 

Computational Level 

• What information is computed and why?

anthropology, philosophy, linguistics 

Algorithmic Level

 How information is represented and computed

Comp sci, psychology

Implementation Level

• The physical substrate that performs the computation 

neuroscience 

Levels of analysis: Language - computational level 

Computational level - past tense = verb + “ed”

• Formal linguistics 

Levels of analysis: Language - algorithmic level 

Algorithmic level -  1. Kick, 2. Kick + -ed, 3. Kicked 

• Psycholinguistics 

2  possible algorithms 

• We have the word in our vocabulary as a whole 

• We form it 

Levels of analysis: Language - implementation level 

Implementation Level

• Neurolinguistics 

• Actually doing studies and testing the parts of the brain 

Algorithmic level - 2 choices

Representation

Algorithm 

• Different representational format = different algorithms and different things being done 

  • Algorithmic choice : choice in execution of algorithms 

    • Ex: 11 + 14, many ways to add it, different steps to take 

software vs hardware

The mind - a software 

The brain- hardware that runs that software 

Simulation - pizza

 for everyday objects, physical make-up does matter 

• A virtual pizza isn't a pizza

Real pizza

• Has to be a physical thing, made out of the right materials, and edible 

Simulation pizza

- Something is missing from making it a real pizza 

- similar buy not quite the real thing

Simulation - chess

Virtual chess and irl chess 

• Different formal system - but just a different instantiations

• Same overall thing/game - the same game of chess 

• Different instantiations are just different formal systems 

• Still the same system/more than one formal system can support chess 

Different instantiations are just different formal systems 

For formal systems, physical makeup doesn’t matter

Doom game - on computer vs calculator

• Same formal system but different instantiations!

• Commonality across all physical instantiations 

Multiple realizability

• Any given mental kind can be realized, implemented, instantiated by more than one physical kind

Ex: Color Vision/perception 

• Humans can see color, animals can see color, robots can see color 

Cuttlefish 

• Behavior, changer color to mate 

• For camouflage

•  Not doing the same, thing, don’t have multiple cones like humans

Levels of analysis 

• Implementational level - human 

• Computational Level - color perception 

• Algorithmic level - cones in the eyes 

• Input + Output 

  Input- colors 

  Output- behavior, naming the colors

The multiple realizability hypothesis

P1-  All mental kinds are multiple realizable by distinct physical kinds 

P2 - if a given mental kind is multiple realizable by distinct physical kinds, then it cannot be identical to any specific kind

C1 - No mental kind is identical to any physical kind

The multiple realizability hypothesis ex

• P1- Mental kinds such as seeing red can be realized by humans, fish, and robots 

• P2- So seeing red cannot be identical, and just realized by a specific species or thing 

C1- Therefore seeing red/any mental kind is not specific to any certain thing 

Functionalism

• The mind is what the brain does 

• Mental states are defined by the function they play 

• They can be in different physical systems- human dog robot 

• Functions are just the right level of abstraction- the crucial link between brain and behavior.

Mental states are defined in terms of their causes and effects 

In functionalism we need 

all three levels of analysis 

Explain the current limits of neuroscience in studying the mind.

• what information does the region represent? What is casual role of that regions response in behavior? “Big data alone arent enough”

What is computation?

Computation - something is a computer because its inputs and outputs can be usefully and systematically interpreted as representing the ordered pairs of some function that interests us 

Why do we care about the computational theory of mind in cognitive science? What is it good for?

• Computational Theory of Mind - describing the physical chain of events misses something - e.g desiring to get help, deciding to dial 911, performing that action

  • Provides framework to understand mind as info processing system 

What are Marr’s levels of analysis, and why are they helpful in studying the mind?

1. Computational Level - behavior

• Why, goal, rules, and logic if computation

1. Algorithmic Level - Mind

What is the relevant level of description + how is it realized in physical system

I. Implementational Level - Brain

What are the physical processes involved in the task

• We need 3 levels of analysis because - 

• We need to know how we get from talking about algorithm to actually completing the algorithm or analysis

• We want to know what is in between, how we are doing all of this 

  • Need all three levels to figure it out

  • Explain and connect across levels

Give an example of a system that you could explain with Marr’s levels.

• Cash register

1. Computational Level - Asks what? And why?

   1. Why addition? Our real goal is to combine prices, and addition is perfect for the job

   2. Commutativity 

      1. What - (3+4) and (4+3) both equal 7

      2. Why - it shouldnt matter what order your goods are presented to the register

   3. Associativity 

      1. What - 3+(4+5) is the same as (3+4)+5

      2. Why - Arranging goods into different piles shouldn’t affect how much you pay 

2. Algorithmic Level - Asks How? (What does the machine do? Add!!!)

   1. Goal of system - Adding - it is a adding machine 

   2. Representation - arabic numerals (1,2,3,4…)

   3. Algorithm - steps like adding the least significant digits first and carrying the 1 if the sum exceeds 9 

3. Implementational Level - Asks Where? (Basically looking at the parts of machine - where do processes take place)

   1. Makeup of cash register

      1. Internal tape

      2. Material keys are made of, material screen is made of 

Explain multiple realizability, and give an example.

• Multiple realizability - any given mental kind can be be realized, implemented, instantiated by more than one physical kind

  • Ex: Color vision/perception

  • Humans can see color, animals and robots can see color

  • Cuttlefish: behavior changes color to mate for camo

    • Not same as humans as they dont have multiple cones

Why do we care about multiple realizability in cognitive science?

• Shows that single mental function (eg. feeling pain) can be preformed by diff physical systems, like human brain or robot neural network

• Crucial for defining theories like functionalism

According to functionalism, what makes something a particular kind of mental state?

Their causes and effects

The kind of functions they play