Exam 1

levels of analysis

experimental way of viewing and explaining computer/machine systems

• types:

  • computational

  • algorithmic

  • implementational

computational level

[top level of analysis] what is the system doing/why is it doing it, defines what the computation is

• Focuses on

  • The goal of the system

  • What will happen after the system is finished

  • What will the output of the computation look like

• ex: music

  • goal of the orchestra (to play a set of specific songs)

  • what will happen when the process is over (validation of good performance via audience applause)

  • logic of their strategy (practicing the songs as often as possible until ready)

algorithmic level

[middle level of analysis] specifies details in the representations used (how)

• Focuses on

  • Steps taken to accomplish the goal

  • Kinds of representations created

  • Kinds of operations formed

• ex: music

  • examining the conductor’s score and player sheet music

  • count for all applicable musical instruments

  • figure out different tempos, rhythms, volumes, musical notes, lyrics (if applicable), etc

implementational level

[last level of analysis] how the algorithm is executed (where)

• Focuses on:

  • Physical system executing steps

  • How steps are executed

  • Active brain regions during cognitive tasks

  • Physical processes

• ex: music

  • observing the performers and conductor while they play and looking for things like

    • note-fingering

    • breathing patterns

    • arm strokes

    • note annunciation

    • posture

  • comparing it to the sheet music for the song being played

representational systems

uses symbols and concepts to stand in for things in the world, especially in the human mind

• Indirect relation between the signal (thing from environment being represented) and system

  • Something stands in for a signal 

  • Stand-ins are part of a larger representational scheme

  • Needs to be distance between environment and system behavior (what stand-in is for)

• Ex: pictures

  • Signal = face 

  • System = phone 

  • Stand-in = photo

mental representations

Representation systems in the mind used to interpret the world; allow for:

• Picturing objects and situations

• Reasoning about objects and abstract ideas

• Planning/carrying out appropriate actions

• Engaging in social cooperation

Types:

• Imagistic

• Propositional

• Symbolic

referent

(mental representations) the thing in the external world the representation stands in for 

• if someone is looking at some food, the physical food is what this is

content

(mental representations) converts the live picture into something the brain can read/understand

• if someone is looking at some food, the mental picture of that food is what this is

intentionality

(mental representations) mental content has a relationship to the real-world referent

• if the person is looking at the food with a want to eat it, that want is what this is

imagistic representation

the mental image of something

• 1-to-1 correspondence between the representation and the real-life thing (isomorphism)

  • Multiple possible but 1/1 is ideal

propositional representation

the remembered name for something

• Can be true or false, often represented by true facts

• Objects of propositional attitudes that specify some way the world could be 

• Can represent counterfactual situations (ex: mom loves dad = dad loves mom)

symbolic representation

the arbitrary way of describing something

• Digital

• Discrete

• Categorical

• Analog (ex: handed clock)

• On a continuous spectrum (ex: color spectrum)

mind vs brain debate

“How are the mind and brain related?”

• Main theories:

  • Dualism

  • Monism

  • Functionalism

dualism

(mind vs. brain) the mind and brain are different bodily parts responsible for separate outputs

• subsets:

  • substance _______

  • property _______

• conflicts:

  • does not say what the mind is not:

    • mind is not the brain

    • mind is not the body

    • mind is different and separate from physical world

substance dualism

(mind vs. body) the mind and body are physically made of different materials

• Res extensa: tangible, physical matter (not just bodies)

• Res cogitans: things the mind is made of

property dualism

mind and body fall under a similar category but displaying different characteristics

• Biological naturalism: mental states are not identical to brain states, but are casually reducible to brain states

• Epiphenomenalism: mental states are caused by brain states, but they do not cause anything (causal dead ends)

  • Ex: just because someone is in a good mood doesn’t mean they’ll have a good day

  • Similar to illusionism

• Panpsychism: mental states are an inherent property of matter/organization of matter (matter made of consciousness)

reductionism

(mind vs body) everything has to be explained by physical natural laws

• If an analysis leaves something out, the problem is falsely posed

• Useless to base the defense of materialism on any analysis of mental phenomena that fails to explicitly deal with their subjective characters

  • Ex: imagining what its like for yourself to be a bat vs. wanting to know what its like for a bat to be a bat

qualia

(mind vs. body) unfamiliar term for things that can’t be familiar to anyone

• Inability for science to explain one’s subjective life experiences

  • Ex: Mary’s room (she doesn’t learn anything, she just has a new experience)

multiple realizability

(mind vs body) mental states can be felt and expressed across species

• Ex: chess (able to be played online, virtually, mentally)

• Relation to the mind:

  • Whether the mind can be physically instantiated in more than one way

methodological behaviorism

(behaviorism; Watson, Hull, Tolman) focuses on observable, measurable entities and behaviors

radical behaviorism

(behaviorism, Skinner) internal mental states exist, but are unimportant and unexplanatory

analytic behaviorism

(behaviorism, Ryle) mental states do not exist

• behavior processes are solely what determines outward emotional expression

Tolman’s rat experiment

(behaviorism) 1937 experiment that proved mental content/representations and forced behaviorists to acknowledge mental content exists

• Put three groups of rats in a maze for 2.5 weeks

  • First group got rewards every day

    • Significant improvement in learning (70% error rate to 20% error rate)

  • Second group didn't get rewards on the first week and did get rewards on the second week

    • Sharp drop in mistakes on day 11

    • Best performance out of all the rat groups

    • Indicates that the rat was learning its way around the maze without needing to be rewarded until the middle

    • Proved that the stimulus response is not the only way that learning happens

  • Third group didn’t get rewards at all

    • Rats barely improve/learn (75% error rate to 50% error rate)

latent learning

the subconscious retention of information without outside reinforcement or motivation

• what Tolman’s second group of rats (third in the picture) were doing

• types:

  • place learning

  • response learning

place learning

a type of latent learning that happens by remembering spatial layout

• Tolmans’ rat example

  • Food is always located in the same place on the east side

  • Rats have to interchange between right and left turns

• requires a mental map

response learning

a type of latent learning that happens by remembering sequences

• Tolman’s rat example:

  • Food is moved east to west, right and left

  • Rat always has to move right

simulation

for an information-processing system, information is all that matters

• _____ and instantiation are the same thing when making informational systems

  • recreating what it does

instantiation

for non-informational system, physical realization of what it is matters

• Physical objects cannot be simulated

• Need more than just information to create instantiation-al objects

Turing machine

• Composition:

  • Infinite tape

  • read/write head

  • Alphabet of symbols

  • Set of instructions (machine table)

• Impact:

  • Opened the possibility that the mind can be a machine

    • Gave the first definition of algorithms and computations

  • Inspired Church-Turing thesis

    • Anything that is mathematically computable can be turned into a Turing machine

framing problem

how AI handles uncertainty and irrelevance

• Relevance = knowing which consequences of the environment are relevant or irrelevant for one’s behavior

  • Set of possible consequences can be infinite

Chinese room

(comp. theory of mind, Searle) questions the sufficiency of symbol manipulation to account for human cognition in AI

• Person speaking a character language slipping a paper into a room things that there is a machine that understands their language, but the ‘machine’ is another person looking up correct sentences in a dictionary of the language

computational theory of mind

algorithms are what matter, and a symbolic processor can compute anything

connectionism

the brain is nothing like a Turing machine, we need a different architecture

• networks of nodes and weighted links

• sub-symbolic informational representation 

  • info distributed among nodes

  • any given node is part of the representation

• Provides an attractive account of concepts

  • Ex: chairs

    • wooden, four legs, can sit on comfortably (able to present them separately as forms of chairs)

weaknesses of connectionism

• Systematicity: rigid focus on X and Y versus generalizing output

  • AI can’t recreate things that it hasn’t seen before

• One-off learning: AI cannot learn from a single instance

• Productivity: humans can have a continuum of processes that they have seen before and come up with new sequences by repeating things; AI can do neither

  • One will only entertain a finite number of thoughts, but there are infinitely many thoughts one may think

• Lack of constitution: AI thought doesn’t realize that A and B are constituents of A&B

  • Ex: “Billy’s dog is small” + “Billy’s dog is black” = “Billy’s dog is small and black”

artificial neural networks

the neurons, connections and activation levels in AI that replace the thoughts, images, memories and language in human cognition

• Node types:

  • Input

  • Output 

  • Hidden

• Connections: weights

• Activation function: sums up input from all connected nodes

graceful degradation

(connectionism) unit destruction in pictures 

• reconstruction of noisy input into comprehensive pictures/sound

backpropagation

the addition of hidden layers to artificial neural networks

• got cognitive science out of the first AI winter

• created another AI winter because there was no hardware for the system to work with

corpus callosum

nerve fibers that run laterally from one hemisphere to another

• responsible for maintaining inter-hemispherical functions

contralateral organization

right brain controls the left side of the body, left brain controls the right side of the body

• connected by the corpus callosum

• when cut, affected people become split-brain patients

  • physical function post-surgery is normal, but orientation is permanently damaged

  • split in self-awareness (right recognizes left and vice versa, but they can’t recognize themselves combined)

fMRI

(functional magnetic resonance imaging) measures blood oxygen levels to see where brain functions are being performed

• doesn’t tell:

  • if the brain regions are necessary

  • anything about cognitive processes

  • how the brain supports cognitive processes

EEG

measures cognitive processes (via neuronal activity) in real-time

• Measures:

  • Action potentials (within a neuron, faster)

  • Postsynaptic potential (between neurons, slower)

    • Chemical and electrical

  • Event-related potentials (measures brain reactions to stimuli)

    • Amplitude and latency in brain waves

representational trade-offs

when one type of representation makes something clear, other previously used information is neglected to oblivion (Marr)

aphantasia

the inability to create mental imagery in one’s mind.

• people with this disorder are still able to accurately and intricately draw out things that others ask them to

• challenges the idea that people solely rely on mental imagery and representations to interpret the world around them.

analog representations

a continuous spectrum of representation

• ex: the solar system

  • infinite planets and starts = infinite spectrum

digital representations

a discrete, categorical type of representation

• ex: color systems in old cameras that only used light colors to take pictures (versus modern cameras)

platonic dualism

(dualism) there are two worlds, the material world and the ideal world (a perfect world) latter of which is only accessible via the mind

Cartesian dualism

(dualism) two different cognitive substances are:

• res extensa (tangible, physical matter)

• res cogitans (things the mind is made of)

monism

(mind vs. brain) the mind and the brain are one and the same

• “What is the mind if not the brain”

  • Brain functions controlling the mind’s reaction

  • Mind is a byproduct of the brain

• Brain is mind, mind is thing that is (vs) mind is brain, brain is what is 

  • Separateness of the brain’s physical output of the brain itself (ex: art from artist)

idealism

(monism) the mind is a product of the universe, which itself is mentally constructed

physicalism

(monism) everything in the universe is physical, including the mind

1. Atom theory: because it is a product of the universe, the mind is made of atoms

2. Identity theory: mental states are brain states

3. Aristotelian physicalism: the distinction between the mind and the body is a distinction between form and matter

Church-Turing thesis

“no computational procedure can be considered as an algorithm unless it can be represented as a Turing machine (Church), and a function is effectively calculable if its values can be found by some purely mechanical process (Turing).”

• If something is mathematically calculable at all, a Turing machine can compute it

• Turing machine’s possibility of mathematically calculating a human mind