COGST 1101: Finals Terms

General Intelligence (g)

1. Spearman 1904:

   1. correlation psychology: old psychology relies too much on introspection and needs to move towards objectively determining facts

   2. does studies on g

   3. determines that there is one central ability that shows up across different tasks, different subjects draw on g to different degrees

   4. intelligence  = general factor (g) + specific factor (s)

      1. intelligence = academic performance + common sense ratings

      2. s = hearing, light, touch discrimination

      3. uses “correction formulas” to control for bias

         1. random errors, prior experience, sex and age differences, socio-cultural and educational effects

         2. sex, age, and cultural disparity are irrelevant factors that can reverse the correlation pattern

   5. sensory tests correlation strong but not perfect with each other, showing that humans have general and specific factor intelligence

2. Gottfredson 1998:

   1. the shared cognitive capacity underlying performance across all mental tasks

   2. reflects a real, measurable ability

   3. predicts almost everything

   4. used in education, employment, policy and society

   5. hierarchical model

      1. general factor (g): core mental capacity

      2. group factors: broad domains (verbal, spatial, etc)

      3. specific skills (s): task dependent abilities

   6. every cognitive test measures g and specific abilities

      1. so IQ approximates g but is contaminated by specific skills

   7. measures:

      1. abstract reasoning: recognizing patterns and relationships

      2. problem solving: breaking complex challenges into manageable steps

      3. learning speed: grasping new concepts quickly and applying them effectively

3. Davis et al 2011:

   1. not at odds with MI theory

   2. just narrow and fails to capture the broad range of human cognition

Gardner’s Theory of Multiple Intelligences

1. Davis et al 2011:

   1. intelligence is better as multiple and content-specific than unitary and general

   2. everyone has all 8 kinds of intelligence, just different levels of each

   3. grounded in empirical studies, gains or loses credibility over time

   4. cannot be observed in isolation

   5. 8 kinds

      1. linguistic

      2. logical-mathematical

      3. spatial

      4. bodily-kinesthetic

      5. musical

      6. naturalist

      7. interpersonal

      8. intrapersonal

   6. bad candidates:

      1. attention and absorption: components of sensory systems

      2. artistic: all intelligences can be used artistically

      3. moral: normative

      4. humor and cooking: mix of intelligences

2. Visser et al 2006:

   1. tests MI theory

      1. 200 adults, ages 17-66, 2-3 hour experiment

      2. 16 tasks, 2 for each intelligence

      3. and the Wonderlic Personnel Test (WPT)

   2. reasoning ability is the common thread linking performance across tests

   3. scored high on WPT = scored high on linguistic, logic/math, spatial, interpersonal, and naturalistic

   4. controls for language (Gardner does not)

      1. finds that correlation is not language, it is g

   5. WPT:

      1. external measure for g

   6. linguistic: ability to use language effectively

      1. vocab quiz, antonym test

   7. logical-mathematical: reasoning, problem solving patterns

      1. arithmetic calculation, word-problem reasoning

   8. spatial: ability to visualize and mentally manipulate objects/spaces

      1. map navigation, paper folding hole punch

   9. musical: skill with sound, rhythm, tone, pitch

      1. pair melodies

   10. bodily-kinesthetic: bodily expression and performance skills, movement

       1. balance, fine motor skill

   11. interpersonal - communication with others

       1. cartoon predictions, social translation (tone/context changes)

   12. intrapersonal - understanding of oneself

       1. self-report consistence, self-assessment accuracy

   13. naturalistic - relates to nature and the environment

       1. categorization, diagramming relationships

Dual process/dual system theories

1. Frankish 2010:

   1. type 1: unconscious, fast, irrational

   2. type 2: conscious, slow, rational

   3. different dual process theories:

      1. learning: implicit learning (learning independent of conscious attempts)

         1. artificial grammar leaning: implicitly extract rules of grammar

      2. reasoning

         1. Wason selection task: green card and even number, alcohol and age

         2. heuristic-analytic theory of reasoning (bias in human reasoning)

      3. social cognition

         1. type 1: default process reliant on simple associations and context

         2. type 2: more cognitively demanding process involving assessment of the message’s content

      4. decision making

         1. conjunction fallacy

   4. dual system theories: the two dual processes are due to two distinct cognitive systems, with different structures, functions, and evolutionary histories (at the extreme: we have two minds)

      1. reflexive associative system: draws inferences from statistical regularities in the environment

      2. deliberate rule-based system: operates on symbolic structures and aims to describe the underlying logical and causal structure

      3. implicit system: non-conscious or pre-conscious, rapid, parallel, low effort, high capacity, domain-specific

      4. explicit: conscious, slow, serial, high effort, limited capacity, responsive to verbal instruction

      5. there is interaction between the systems

   5. differences in g are differences in the capacity of system 2

   6. potential type 3: process that is responsible for initiating type 2 processing and for resolving conflicts between autonomous and analytic processes

      1. have ultimate control of behavior

2. Pennycook and Thompson 2016:

   1. type 1: autonomous, fast, high capacity

      1. processes automatically generate a quick “first answer”

   2. type 2: reflective, slow, resource demanding

      1. processes can accept, reject, or correct the type 1 response

   3. people tend to rely on type 1 processing first, rather than careful statistical reasoning

Conjunction Fallacy

1. Frankish 2010:

   1. a conjunction of things will always be less likely than the thing by itself

   2. Linda and if she is a bank teller and a feminist

Representativeness Heuristic

1. Tversky and Kahneman 1974:

   1. insensitivity to prior probability outcomes

      1. when a description feels representative, we ignore real statistics

      2. eg: librarians vs farmers

   2. insensitivity to sample size

      1. people treat small sample sizes as if they are just as stable as large

      2. eg: hospitals where over 60% of babies born will be boys

   3. misconceptions of chance

      1. people expect sequences to be random, even short

      2. eg: gambler’s fallacy

   4. insensitivity to predictability

      1. people judge outcomes based on how closely a description resembles them, even if they hold little predictive value

      2. ignore how reliable the info is, and if the outcome is predictable

   5. the illusion of validity

      1. people feel confident in a prediction when the outcome seems to match the given prediction

      2. confidence is high even when the info is incomplete, unreliable, loosely related to the outcome

   6. misconceptions of regression

      1. regression toward the mean: an extreme outcome is usually followed by results that are closer to the average outcome

      2. people ignore this and expect another extreme outcome

      3. eg: plane landing

Availability Heuristic

1. Tversky and Kahneman 1974:

   1. a mental shortcut where people estimate likelihood, frequency, or probability based on how easily examples come to mind, not actual statistics

   2. eg: words that start with t are easier to generate

   3. frequent, recent, vivid/emotional, personal experience, and media exposure

   4. bias 1: effectiveness of search set

      1. we overestimate categories that are easier to search in memory

      2. eg: there are more words that have a k as the 3rd letter than start with a k

   5. bias 2: retrievability of instances

      1. classes with more memorable examples feel larger

      2. eg: list of famous women vs random men

   6. bias 3: imaginability

      1. people judge the likelihood or frequency of an event by imagining how easily they can mentally construct examples

      2. eg: when mentally listing committee sizes, it is easier to list 2 person committees then 8

   7. bias 4: illusory correlation

      1. people falsely believe two things are related because they are strongly associated in their memory or stereotypes

      2. eg: drawings of paranoia involve eyes

Adjustment & Anchoring

1. Tversky and Kahneman 1974:

   1. phenomenon anchoring

      1. estimates start off from an initial value, which are then adjusted to yield a final answer

   2. insufficient adjustment 

      1. anchoring given problem:

         1. estimates can be influenced by the framing/information provided in the problem

         2. eg: given an initial value to estimate what % of African countries were in the UN

      2. anchoring given partial computation:

         1. partial computation can greatly influence an estimate

         2. eg: multiplication 1-8 or 8-1

   3. conjunctive vs disjunctive events

      1. conjunctive (and): series of successful events

         1. eg: planning: successfully completing a project or plan requires a series of events to successfully occur: leads to unwarranted optimism

      2. disjunctive (or): one success out of a series of events

         1. eg: risk evaluation: complex systems, such as nuclear reactors or human bodies, often fail if an essential component fails: leads to an underestimation of a possible failure in the overall system

      3. more people prefer to bet on conjunctive events than simple events, and simple events over disjunctive events

   4. subjective probability distributions

      1. with different distribution calculation methods and problem framing, anchoring is clear in the resulting odds

      2. eg: experts express their opinions about something (eg: Dow Jones average)

Base Rate Neglect

1. Pennycook and Thompson 2016:

   1. base-rate: the statistically calculated probability of event occurring given no other conditions

   2. the human tendency to ignore the base-rate of an even happening

      1. are likely to be ignored in favor of stereotypes

   3. eg: is Paul more likely to be a doctor or a nurse

   4. eg: breast cancer screenings

Base Rate Sensitivity

1. Pennycook and Thompson 2016:

   1. people are sensitive to the base rate when:

      1. the causal link between the base-rate and judged case are explicit

         1. eg: green and blue cabs, motor a and motor b

      2. frequency formats are more easily understood by participants because they are consistent with the sequential way that information is acquired with natural sampling

         1. eg: breast cancer screenings written in frequency format

      3. given problems where the base-rates come after individuating information (eg: stereotypes) (order of presentation)

         1. congruency: given base-rate and stereotype agree

         2. incongruency: given base-rate and stereotype disagrees

            1. when this happens, participants no longer integrate base-rates

The Gambler’s Fallacy

1. Tversky and Kahneman 1974:

   1. misconceptions of chance: people expect sequences to look random, even if both sequences are actually equally likely

Regression to the Mean

1. Tversky and Kahneman 1974:

   1. an extreme outcome is usually followed by results that are closer to the average outcome

   2. people miss this pattern and expect another extreme outcome

   3. eg: plane landings

Utility/Utility Function

1. Shafir and Tversky 1995:

   1. a function that captures each individual person’s subjective, experience-based value for money

   2. also known as “subjective value”

Expected Value

1. Shafir and Tversky 1995:

   1. a predicted outcome calculated by summing the product of the probability of each potential outcome with its value

   2. objective

   3. eg: 50% chance to win $200 or 100% chance to get $100 → options are identically rational

Expected Utility

1. Shafir and Tversky 1995:

   1. a concept used to model decision-making in uncertain circumstances, calculated by summing the product of the probability and utility of each potential outcome

   2. subjective

Risk Aversion

1. Shafir and Tversky 1995:

   1. a preference for a sure outcome over a risky prospect that has higher or equal expected worth

   2. losses “hurt” more than equivalent gains feel good

      1. the psychological value curve of losing $50 is steeper than gaining $50

Risk Seeking

1. Shafir and Tversky 1995:

   1. a preference for a risky prospect over a sure outcome that has higher or equal expected worth

Prospect Theory

1. Shafir and Tversky 1995:

   1. a part of descriptive choice analysis that accounts for observed regularities in risky choice

   2. 3 parts

      1. convex for losses and concave for gains

         1. risk aversion with gains and risk seeking with losses

      2. based on gains and losses not total wealth

         1. people normally treat outcomes as gains and losses relative to a neutral reference point rather than total wealth

         2. framing impact decision-making

      3. steeper for losses than gains

         1. losses loom larger than respective gains

         2. aka: loss aversion

Framing Effects

1. Shafir and Tversky 1995:

   1. the same problem presented in different ways can lead to different preferences

   2. eg: you are $300 richer than today:

      1. gain $100 OR 50% chance to gain $200 and 50% to gain nothing

   3. eg: you are $500 richer than today

      1. lose $100 OR 50% to lose nothing and 50% chance to lose $200

   4. problem 1: risk averse

   5. problem 2: risk seeking

   6. dominance principle: if one option is better than the other on one attribute and at least as good on all the rest, then it should be chosen

      1. eg: gain $240 OR 25% chance to gain $1000 and 75% chance to gain nothing

      2. eg: a sure loss of $750 OR 75% chance to lose $100 and 25% chance to lose nothing

      3. problem 1: risk averse

      4. problem 2: risk seeking

   7. description invariance: equivalent representations of a choice problem that contain the same information should result in the same preferences

      1. framing effect may lead to a violation of this

Endowment Effect

1. Shafir and Tversky 1995:

   1. owning something makes giving it up feel like a loss, and loss aversion makes that loss feel big

   2. eg: half of the participants are given a mug, sellers are asked how much they’d sell for it, choosers are asked how much they’d pay for it

   3. eg: keeping vs selling a concert ticket

Compatibility-Based Attribute Weighting

1. Shafir and Tversky 1995:

   1. if you are selecting something, you are looking at its positive attributes

   2. if you are denying something, you are looking at its negative attributes

   3. eg: roommates

   4. eg: custody case (Sharif 1993)

      1. parent a (impoverished option): average income, health, working hours (neutral)

      2. parent b (enriched option): positive features: above-average income, extremely close relationship with child

         1. negative features: minor health problems, very active social life, frequent travel

      3. asked if you should award or deny custody

      4. most people choose and reject parent b

   5. procedure invariance: preferences should be the same whether you ask people to choose or reject

Mental Model

1. Markman and Gentner 2001:

   1. a representation of some domain or situation that supports understanding, reasoning, and prediction

   2. two main approaches to study:

      1. characterize the knowledge and processes that support understanding and reasoning in knowledge-rich domains

      2. focus on mental models as working-memory constructs that support logical reasoning

Turing Test

1. Mays 1952:

   1. created by Alan Turing

      1. measures a machine’s ability to exhibit intelligent behavior equal to or indistinguishable from a human

      2. if you can’t tell the difference between a machine and a human, then the machine is thinking

   2. Mays is critical of his work

   3. claims that for a machine to be thinking it must have

      1. a private life of its own

      2. daydreams when it does some tasks

      3. has intelligence, consciousness, and free will

   4. symbol manipulation is not cognition, binary translation is limited

   5. logical atomism: the belief that everything can be broken down into small discrete parts and analyzed individually rather than as a whole

      1. atomic system: can be broken down, eg: computer

      2. organic system: cannot be understood by breaking down into parts, eg: human brain

Weak vs. Strong Methods of Reasoning

1. Markman and Gentner 2001:

   1. first proposed by Newell and Simon (1972)

   2. weak methods: general strategies that can operate without special knowledge of a domain

      1. general

   3. strong methods: make intensive use of represented knowledge, as in reasoning by example

      1. typically superior when the appropriate knowledge is present

Mental Simulations

1. Markman and Gentner 2001:

   1. qualitative

      1. people reasoning about relative properties of physical systems

      2. people do not estimate values of specific quantities

   2. mental imagery

      1. mental simulation often involves the use of imagery

         1. eg: moving gears

   3. Wason rule discovery test

      1. rule that generates a series of 3 numbers

      2. in order to test the correct series, a person would need to guess series that do not fit

         1. however they don’t normally do this

Weak vs. Strong AI

1. Bory et al 2024:

   1. strong AI narrative: AI is sentient/conscious

      1. human-like or superhuman abilities

      2. common in sci-fi and future speculation

   2. weak AI narrative: focus on existing systems

      1. descriptions of how current AI actually works

      2. no claims of sentience

   3. eg: people think the kid-AI metaphor (LaMDA) is strong AI, but it is not

The “Animism Trap”

1. Mays 1952:

   1. we can’t trust our intuition on whether machines are thinking

   2. against the Turing test

Searle’s Chinese Room Argument

1. Greve 2025:

   1. this is "biological naturalism”

      1. humans mindedness requires being a living organism, not just any physical system

Autopoiesis

1. Greve 2025:

   1. “self-producing/organizing”

   2. the belief that being alive is sufficient for cognition

Biophilia

1. Greve 2025:

   1. biophilic prejudice: the emotional bias that leads humans to favor living beings and instinctively doubt that non-living systems could ever have real minds

   2. we do not know if machines can be alive

   3. we do not know if life is required for human-like mindedness