psych 354 - text definitions (chat)

cognitive science

interdsciplinary study of mind+ cognition

• ex. using psychology + CS + linguisitics + philosophy to stuy memory

• brings different fields together to understadn cognition

information processing

cognition as processing information , the rule governed manipulation of symbols

• ex. when a computer uses built in rules to manipulate or rearrange symbols in its memory, we say the compiter is performing information processing.

• important since researchers assume human cognition results form symbol manipulation

classical approach

views the mind as a computer/symbol manipulator

• ex. solving math problems using rules

• focus on symbols, rules, planning

connectionist approach

views the mind as a network of interconnected unites

• ex. neural network learning torecognize cats

• contributed the idea of models learning and pattern recognition

embodied approach

views cognition as dependent on body and environment

• ex. using your body-environment to navigate

• showed that cognition isnt necceearily just inside the head

methodological sollipism

study the mind without referring to the outside world

• ex. explaining a though only through internal repreentations

• introdiced important assumption of classical cognitive science

sense-think-act

classical approach to cognition in that we perceive→ think/plan→ act

• ex. see a ball, decide to catch, then move hand

• provided a framework for classical model of cognition

sense-act

embodied approach to cogntition in that we perceive then act, without requiring consious planning

• ex. robot sees obstacle then immedietly moves around it

• important to embodied cogntiion

affordance

an enviornment offers possibilities for action

• ex. a chair affors sitting

• this idea connected percpetion direclty to action

feedback

actions change the environmet, which changes future actions

• ex. move toward ball, ball moves, adjust movemnt

• this idea was important to embodied/dynamical systems

extended minf

cognition can extend beyong the brain/body

• ex. using ur phone as an external memory

• this cahllenges the idea that cognition is only inside the skull

cognitive scaffolding

external tool support cognition

• ex. using a notebook to remember things

• shows how the environment can enhance thinking

c

cognitive atchitecture

the basic strucute/mechanism of cognition

• ex. symbolic coputer vs. neural network

• major source of disagreement between approaches

computational level

what problem is the mind solving

• ex. recognizing whether a picture contains a face

• identifies what the cognitive system need to accomplis

algorithmic level

what steps/processes are used to solve the problem

• ex. comparing featrues of a face to stored information

• explains how information is processed

architectrual level

what basic components are used to carry out the process?

• ex. memory and procesing units

• explains the basic stucutre that supports cognition

implementational level

what physical echanisms make the system work

• ex. neurons and brain activity

• connects cognitive processes to their physical/biological basis

  • 
    

algorithim

a set of steps for solving a problem

• ex. steps used to recognize a face

• helps exlaon how cognition operates

architecture

the basic building blocks of a cognitive system

• ex. different processing units working together

• explians what cognitive processes are built from

implementation

the physical form that creates the architecture

• ex. brian neurons creating processing units

• connects cognitive theories to physical mechnaisms

multiple relaization

the sam function can be done by different physical systems

• ex. 2 different types of machines performing the same circulation

• shows that the function and physical material can be studied seperately

input-ouput mapping

how a system turns an input into an ouput

• ex. sees a dog → identify it as a dog

• helps researhcers desribe what a cognitive system does

binary logic

uses 2 states: 0 or 1, or TRUE or FALSE

• ex. a switch is ON =1, OFF= 0

• helped connect logic and thinking to physical computers

logicism

the idea that thinking is based on logic

• ex. using logical rules to solve a problem

• helped inspire the idea that thinking could be modeled computationally

homunculus

imaginary littler person inside the mind doing the thnking

• ex. the little person i n ur brain decides what to do

• shows wy cognitive explantions need to explain the actual mechanism rather than sumng another thinker

behaviour by design

behaviour that happens because a system was designed to produce it

• ex. a calculatoe is designed to calculate

• helps explain behaviour in terms of a systems intended funciton

behaviour by artifact

behaviour that comes form the way a system i physically built

• ex. a circuit produces n output because of how its parts work

• shows why we need to ecamine the hysical system behind behvaiour

disembodied mind

idea that the mind is purely logical, thinking system that works independently of the physical body or sensory-motor experiences

• ex. playing a game of online text-based chess, where thinking happens through abstract logic iwhtout needing physical movement or body interaction

• established early classical belief that cog sci could study mental processing as software without worrying abt bodily hardware

physical symbol systems (PSS)

a physical machine that takes physical tokens (symbols), arranges them into structures, and manipulates them according to rules to produce inteligent behaviour

• ex. an app or digital calculator that takes inputs (numbers/letters) and manipulates them using programmed rules to display an answer

• proved that physical hardware can display general intelligence, providing a materialist foundation for cog sci

recursion

a rule or process that refers back to itself, allowing finite rules to build infinitely complex or nested structures

• ex. russian nesting dolls, or adding endless descriptive clauses in everyday conversation

• showed that human minds do not need an infinite memory bank to produce endless novel thoughts or sentences, disproving simple behvaiourist models

context-free grammar & rewrite rules

rules that break down or rewrite categories into smaller parts regardless of surrounding context

• ex. a recipe step that says ‘make sandwhich’, which expands into ‘bread+filling’, where ‘filling’ expands into ‘cheese+tomato’

• gave cognition scientists a formal tool to model how human brains construct the hierarchial grammar of natural language

universal turing machine

a theoretical device that can run any program or execute any computable task when given the right instructions

• ex. ur laptop, it isnt just a typewriter, a calculator, or a video plater; it becomes whichever one u need depending on the app software u open

• provided the foundational framework for the computational theory of mind, establishing the idea that human cognition is a form of information processin

componentiality (locality)

rules that act locally on individual symbol components rather than processing an entire complex expression as one giant whole

• ex. spell-checking a word letter by letter rather than evaluating thvisual shape of the entire page

• showed how complex, intelligent opertions can be built out of mindless, step-vy-step local rules without needing an intelligent homunculus insisde

syntax (formal operations)

manipulating symbols strictly by their physical form, shape, or rules, without knowing what those symbols mean

• ex. a vending machine accpeting a coin because of its exact weight, thicness, and mental shape, without understanding what ‘25 cents’ or ‘money’ actually means

• concept explained how a purely physical brain can perform logical calculations based on symbol shapes alone

designation (symbol grounding)

the connection between an internal symbol and a real object or outcome in the world

• ex. preseeing the button labeled ‘brew’ on coffee machinne, which causes the physical machine to pour real coffee into ur cup

• helped explain how abstract mental symbols connect to real-world objects and actions

intentional stance

predicting what something will do by assuming it has beliefs, desires, and goals, and that it will act rationally

• ex. saying ur robot vacuum cleaner ‘wants to return to its dock because it thinks its battery is low’

• validated using everyday mental terms (beliefs, desires) as illegitimate tools for scientific prediction

classical cognitive architecture

a clean separation between stored data/symbols (structure) and the active rules that manipulate them (process)

• ex. a cookbook sitting passively on the ocunter (structure) versus the chef actively reading and executing the steps (process)

• defined the blueprint of classical digitla arhcitecture, contrasting them with neural networks where memory and processinf are merged

production system

a system controlled by working memory and a set of confition-action (if…then…) rules

• ex. a smart home setup: if motion is detected and if it is dark outsidde, then tur on the porch light

• provided a dominant model for stimulaitng human rule-folowing, decision-making, and problem-solving

the classical sandwhich (sense-think-act cycle)

traditional view that perception (sensing) and action are separate, linked only by central planning/thinking phase

• ex. driving a car, seeing a red light (sense), calculating hwo quickly to brake (think), and then stepping on the brake pedal (act)

• served as the foundational contol model for early artificial intelligence and classical cognitive psychology

weak equivalence vs. strong equivalence

weak: 2 systems get the exact same final answer, but use different internal steps. strong: Two systems get the same answer and use the exact same internal algorithm and architecture to get it

• ex. a human doing long division on paper vs a digital pcoket calculator both arriving at 42.If a computer simulation solves a math puzzle using the exact same step-by-step methods, pauses, and mistakes as a human student, it is strongly equivalent

• defined the ultimate goal of classical cog.sci→ creating computer models that are true thoeries of human thought, not just output copiers

second-order effects (artifacts)

unintended behavioural side effects- like response time delays or msitake patterns- that reveal how an internal algorithim works

• ex. taking longer to mentally multiply17×24ehan 2×3 revveals ur mental algorithim works in mulitple step bystep column seperations

• provide psychologists with experimental tools (like measuring reaction time) to reverse engineer hidden mental processes

cognitive impernetrability

a mental function that cannot be altered or turned off by ur consious beliefs, goals, or knowledge

• ex. optical illuisons → even when u measure 2 lines with a ruler and know for a fact they are the exact same length, one still looks longer to ur visual system

• helped cognitive scientists separate hardwired biological ‘hardware’ primitves from flexible ‘software; beliefs

empiricism

view that we are born as a blank slate and knowledge is acquired through expereinces and interacting with the world

• ex. infant larning to categorize speech sounds or human faces over mmonths of sensory exposure rather than relying on pre-programed linguistic rules

• shifted cog sci away from nativism toward neural networks that are able to learn directly from environmental input

associationism

framework where mental processing occurs through connected units, learning happens by adjusting numerical strengths (weights) of those links based on experience

• ex. classical conditioning, a dog associationg a bell tone wit food because neural units representing the food and sound repeatedly fire together, strenghtening their connection weights

• replaced hand-coded symbolic rules with biologically plausible model of memory stored across adjustable weights

perceptron

early single-layer ANN that multipies inputs by connection weigths and uses threshold activation fucniton to output binary decision

• ex. Mark I percpetron classifies visual shapes or a simple neural netwolr deciding whether a musical chord is major or minor based on note inputs

• this demonstrated a physical system could learn perceptual classifications directly from experience, serving as foundation for early connectionism

linear sperability

property of a problem wheer all inputs in one category can be separated from another category by drawing a single straight line through the pattern space

• ex. the logical AND operation, where a single line isolates the single ‘true’ pair, from 3 ‘false’ pairs

• it defined the maximum computational limit of single layer perceptron

linearly nonseparable problems

classification problems wheere categories cannot be separated by a single straight line, requiring at least 2 cuts or curved boundaries

• ex. the logical XOR operation, where ‘true’ outputs sit opposite diagonals from ‘false’ outputs

• perceptron failure on XOR drove scientists to invent multilayer networks with hidden units, better model for human cognition as it is able to solve more complex problems

connectionist sandwhich

a network architecture where sensory input layers and motor output layers are separated by intermediate representational hidden layers

• ex. human visual object recognition, light on retina (input/sense) passes thorugh visual cortex to process features (hidden/think), leading to grabbing object (output/act)

• proved that connectionist networks are fundamentally representational systems that conform to a version of the classical sense-think-act cycle

hidden units

intermediate processors between input and output layers that re-represent data to detect complex, higher-order features

• ex. in amusical chord network input takes raw notes , while hidden units combine them to detect abstract pitch relationships like major thirds

• overcame the limits of single-layer perceptrons, allowing networks to solve linearly nonseperable problems and discover internal representations

backpropagation

learning tule that measures output error and sends error signals backward through the network to adjust hidden connection weights

• ex. language network outputting ‘runned’, receiveing an error signal agast target ‘ran’, and passing feedback backward to tweak internal conenction strengths

• solved the credit assignment problem , by letting multilater networks automatically learn complex representations

supervised learning vs unsupervised

supervised modifies weights using explicit target answers and external error feedback, unsupervised organizes weghts based on statistical patterns in raw data without a teacher

• ex. supervised: a student learning letters with a teacer correcting mistakes, unsupervised:infants brain mapping tones into a circle of fifths by listneing to ambient sound

• demosntrated how cog systems acquire structure either through direct feedback or through self-organized perceptual categorization

distributed representation and course coding

stong concepts across activation patterns over many units, individuals are broadly turned on their own, but pooling their overlapping repsonses yields accurate representations

• ex. trichromatic vision, 3 broadly tined cone types respond to wide light range, pooling overlapping signals allow pervpetion of millions of distinct colors

• expanded cog sci beyond localized symbolic memory addresses, allows networks to genralize across similiar inout patterns

methodological sollipism

classical assumption that mental states should be studied purely as internal represetnations inside an individuals hhead- ignoring the external environment and body

• ex. modeling a chess computers deciison-making by looking only at internal board-evaluating algorithim, without considering the physical board

• served as the core philosphical premise of classical cog sci, became primary target of critique for embodied cognitive scientists that argue the mind cannot be isolated form te body and the world

sense-act cycle

a control framework that directly connects perception (sensing) to behaviour (acting), completley bypassing central representational planning (thinking)

• ex. a reflex like wall-followinf robot that automatically steers away when its bumper sensor hits an obstacle, rather than mapping out a 3D floor plan in memory first

• replaced the traditional sense-think-act cycle(classical sandwhich) in embodied cog sci, demosntrating that complex behvaiour can emerge without central mental models

situatedness

an agents capacity to perceive and interact with its immediate environment in real time through its specific sesory apparatus

• ex. bat navigating a dark cave usinig echolocation, where its perceptron depend s==s continously on acoustis signals bouncing off the walls

• shifted focus from offline, abstract reasoning toward real-time perceptual enagegment with surrounding environment

emobodiment

having aphysical body whose specific physical structure (limbs, sensors, weight, shape) constrains and enables how an agent can affect and be affected by the world

• ex. a passive dynamic walking robot that walks down a slope without any motors or computers, relying only on gravity, joint angles, and leg length

• showed that physical body mechanics can solve problems that classical AI assumed required complex internal software calculationns

feedback

2 way loop where agent’s actions change the environment, and the environmetnal changes alter the agents next perceptions and actions

• ex. a driver steering a car- turning the wheel, alters cars position on the road, which changes visual input, which guides the drivers next steering adjustment

• intorduced cybernetic feedback principles to explain how intricate behaviours emerge dynamically without needing central executive planner

affordance

concept of potential action that an object or environment offers to a specific agent bassed on agents physical body

• ex. flat, rigid surface at knee heirght offers ‘sitability; to an adult human, but not to an elephant or an ant

• grounded percepetion directly in motor action, showing that organisms perceive possibilities for action rather than abstract physical dimensions

stigmergic control (stigmergy)

indirect coordination where an agents action alters the physical environment and that environmental chnage triggers the next action by the agent or others

• ex. termites building complex, nest pillars by depositing mud pellets wherever dirt concentrations are already hhighest, without any central blueprint or leader

• explained how decentralized, highly organized group behaviours emerge naturally from simple environmental cues

forward engineering (synthetic psychology)

bottom-up methodology of building simple physical agents and placing them in environments to see what complex behaviours emerge, rather than reverse-engineering intact minds top-down

• ex. building a simple wheeled robot with light sensors wired to wheels and observing it exhibit ‘fear’ or ‘aggression’ toward light sources

• demonstrated that simple internal mechanisms interacting with complex environments can generate behaviours tha tappear deceptively complex

cognitive scaffolding

using physical structures, tools, language, or cultural artifacts in the external world to reduce internal cognitve workload

• ex. An expert dairy worker calculating a milk order by looking directly at partially filled milk crates rather than performing mental arithmetic in working memory

• expanded cognitive theories beyond the ‘naked brain; to include physical environments as active partners in problem-solving

extended mind hypothesis

when external tools are reliably integrated into our thinking processes, those tools literally become part of the cognitive system itself

• ex. person with alzheimers using a notebook as their LTM store; notebook funcctions identically to internal memory retrieval

• challenegd the skull boundary of mind, asserting that cognitive processing extend into the physical world