group cognition Flashcards
1
Q
the madness of crowds
A
- Charles Mackey
- began with stock market ideas
- stock market activity and economic bubbles:
–> investors buy up cheap shares and this snowballs increasing the prices of the shares until they no longer reflect the value of the company
–> bubble bursts and the shares are worth very little - idea that we make worse decisions due to a group
2
Q
wisdom of the crowds - brief example
A
- people asked to estimate weight of an ox
- guesses made individually but the mean the guesses was very close to the actual weight of the ox
- idea that we can make better decisions and estimates as a group
- “any investigation into the trustworthiness and peculiarities of popular judgements is of interest”
3
Q
real world example
A
- northern rock bank
- got a bad rep
- people wanted to take money out
- the mass withdrawal of money actually caused the bank to crash
- the group decision led to the downfall of the bank
–> may not have happened if mass withdrawal didn’t occir
4
Q
small group decision making
A
- typical paradigm to test group cognition is as follows:
–> 3 - 6 people
–> short tasks
–> common aims
5
Q
example of a short decision task
A
- short task with a definitive answer
- Robin is looking at Charlie, Charlie is looking at Jules
- Robin is married but Jules is not married
- is a person who is married looking at a person who is not married?
- answers:
–> A = yes
–> B = no
–> C = we cannot tell
6
Q
example of a short decision task = typical answers
A
- individual guesses:
–> yes = 5%
–> no = 2%
–> can’t tell = 93% - group guesses
–> yes = 24%
–> no = 3%
–> can’t tell = 69% - more people get the answer right in groups
7
Q
right answer to the simple decision task
A
- yes
- if Charlie is married (m), Robin (m) is looking at Charlie (m) and Charlie is looking at Jules (nm)
- if Charlie is not married (nm), Robin (m) is looking at Charlie (m) and Charlie (m) is looking at Jules (nm)
- in either scenario a married person is looking at a non-married person
8
Q
Wason’s selection task
A
- four cards all have a letter on one side and a number on the other side
- example: E X 1 6
- rule: all cards with a vowel on one side have an even number on the other side
- which cards would you have to turn over to decide whether this statement is true or false?
–> E and 1 - 1 is an odd number so if a vowel is on this side, the rule has been broken
- most people select E and 6
–> this is confirmation bias
9
Q
confirmation bias
A
- A preference for seeking information that can only confirm your existing beliefs, rather than contradict it
- About active search for information, not just whether you believe information when you encounter it
10
Q
confirmation bias in Wason’s task
A
- You should turn over E, since it might have an odd number on the other side
- You need to also turn over 1, since it might have a vowel on the other side
- Turning over 6 doesn’t tell you anything
–> it would just confirm the rule but doesn’t PROVE it is correct
11
Q
Wason selection task in groups
A
- can be used in a small group decision task
- approx 80% of groups arrive at the correct answer
–> complete flip from individuals where approx 70-80% arrive at the wrong answer - few mins of discussion can change the wrong answer into a correct response
- allows researchers to look at the process of reasoning in groups to come to the correct answer
12
Q
Wason task controls
A
- things that don’t help:
–> Motivation / reward
–> Changing the wording
–> University education - something which does help:
–> Making the task less abstract
–> Working within a group (perhaps)
13
Q
Wason task - social rule version
A
- there are four people in a pub, each has a drink
- these cards each have the age (on one side) and the drink (on the other) of someone in the bar
- beer, cola, 17, 25
- rule: people with an alcoholic drink must be older than 18
- which cards inspect the rule is broken?
–> beer and 17
–> make sure beer isn’t below 18
–> make sure 17 isn’t alcoholic
14
Q
conclude Wason task
A
group cognition can improve on individual reasoning
15
Q
are groups better than individuals?
A
- ‘process loss’
–> where group decisions are worse than individual (madness of crowds) - ‘process gain”
–> where group decisions are better than individual (wisdom of crowds) - Most of the time groups performed at the accuracy of second best member of the group
–> group cognition tends to avoid the individual worst answer but also the best answer
16
Q
why comparing groups is hard:
A
- need to be able to define four factors:
1. task type
2. standards of comparison
3. coordination methods
4. individual differences
17
Q
task types
A
- intellective
–> have a definite answer - judgement tasks
–> estimations / opinions - also what does the task depend on:
–> require insight?
–> require background knowledge?
–> provoke strong intuitions or emotions (biases)?
18
Q
task type: individual vs group?
A
- evidence suggests that:
–> Given time and discussion, groups perform as well as best individual on Intellective tasks
–> Evidence that best members outperform groups on judgement tasks
–> When the task does not have a clear answer then groups tend to be perform at the level of the average members
19
Q
standards of comparison
A
- performance is split into:
–> worst individual
–> average
–> best individual
20
Q
coordination methods:
A
- refers to how the group functions
–> level of discussion
–> anonymity
–> revision - no discussion
–> averaging individual answers - iterative, anonymous answers, no discussion
–> ‘Delphi’ method revises answers to reach consensus - Discussion group choses the best individual to answer
–> “dictator method” - Discussion
–> come to group agreement “consensus method” - Discussion with revision
–> given collective mean, discuss and revise “dialectic methods”
21
Q
evidence for different coordination methods
A
- best improvement in dictator group
- then delphi
- then dialectic
- least improvement was the consensus group
- However, none outperformed the best individual members
- Note = in the dictator group the best performers often adjusted their response towards the collective mean
22
Q
individual differences
A
- In sources of information
–> access to cues - In ability
–> i.e., better memory - In other capacities
–> e.g. ability/willingness to coordinate
23
Q
achieving group consensus
A
- Sniezek and Henry (1990)
- suggested that consensus is achieved through revision and weighting
- revision occurs within the individual
- weighting (the combination of multiple judgements) occurs within the group
24
Q
lens model of group decision and making consensus
A
- based on ideas of revising in individuals and weighting in groups
- model informs how groups may arrive at consensus judgements
- accuracy of group decision making relies on accuracy of individual judgments
- group judgements that are highly related to the criteria in the environment then this would be an accurate decision
- can be influenced by systematic bias or persuasive individuals (unequal weighting)
- weighting towards individuals and information can affect accuracy of the group judgement
25
Q
lens model conclusion
A
- Lens Model provides a framework to help us think systematically about the different factors which might affect group cognition
- However, it can be difficult to study as there is often limited access to the internal thought processes in discussions
26
Q
conditions for wisdom of crowds
A
- wisdom of crowds does not work all of the time
- the conditions when it does work:
–> independent estimates = uncorrelated errors
–> no systematic biases
–> no coordination between group members
27
Q
uncorrelated errors
A
- there is a true value
- there is noise (differences)
- with uncorrelated errors, Ps sample this true value with noise
–> spread out estimates
–> some below, some above and some close to the true value
28
Q
correlated errors
A
- Ps sample the true value with noise AND bias
- all tend to group together
–> tend to be above the true value
–> over estimate - Due to limited information
- Due to shared (individual) biases
- Due to group conformity
–> all reduce the wisdom of crowds effect
29
Q
things that affect group cognition
A
- groupthink
- diversity
30
Q
groupthink
A
- polarisation in group decision making
–> attitudes expressed in the group move away from the average of individual’s opinion
–> move towards a more extreme position - groupthink
–> highly cohesive groups exhibit premature consensus seeking
–> i.e. premature closure on the group level (quick decision making)
–> leads to poor decision making - impacted by:
–> overconfidence
–> blindness to errors
–> conformity
31
Q
what inspired groupthink?
A
- JFK
- Cuban Missile Crisis
- Bay of Pigs
- huge mistake
- USA sent planes and tried to bomb Cuba
- massive error
- his group decision making with his close advisors was highly criticised
32
Q
example of groupthink: space shuttle challenger
A
- launched on January 28th 1986
- several delays before
- freezing temperatures
- some engineers concerned about the seals in the cold
- ended up in a huge disaster
- Rogers Commission found that issues with NASA’S organisational culture and decision-making processes had been key to the accident
33
Q
criticisms of groupthink
A
- Not a distinct phenomenon?
- Does it add anything to the literature of group reasoning?
- Has it thwarted understanding of group reasoning?
- Doesn’t happen?
–> lack of empirical evidence for all of the constructs associated with groupthink - Focus on when group decisions have led to negative outcomes, restricts the understanding of group decision making process
- BUT has been useful in focusing attention on potential flaws of group decision making
34
Q
diversity as an antidote to bias
A
- more diverse editing teams produce higher quality Wikipedia articles
- they spend longer in more complex discussion
35
Q
social accounts of reason
A
- interactionist account
- argumentative theory of reasoning
36
Q
interactionist account
A
- reason evolved to produce and evaluate arguments
- not for individuals to solve problems (the individualist account)
37
Q
interactionist approach - Wason task
A
- Individually, 80% of people fail
–> a strong bias against getting the right answer
–> simple aggregation should compound this effect - Result: 80% of groups get the answer right
–> majority failure converted to majority success
–> a “truth wins” scenario
38
Q
argumentative theory of reasoning
A
- analysis of transcripts shows exchange of arguments is key
–> argumentative theory or reasoning - Mercier & Sperber argue that confirmation bias is an individual failing
–> but here a collective strength
39
Q
how do groups reason?
A
- exchange of arguments:
–> “Groups typically co-constructed a structure of arguments qualitatively more sophisticated than that generated by most individuals” - arguments change people’s problem representation
40
Q
predicting collective intelligence - collective intelligence definition
A
the ability of a group to perform a wide variety of tasks
41
Q
collective intelligence
A
- collective intelligence is the general ability of a particular group to perform well across a wide range of different tasks
- Inspired by the idea of general intelligence performance of an individual across a range of different kinds of cognitive tasks, encapsulated as a common statistical factor calledgor general intelligence
- does general intelligence of group predict collective intelligence?
42
Q
general intelligence or social sensitivity?
A
- collective intelligence is not strongly correlated (only moderate relationship) with the average or maximum individual intelligence of group members
- collective intelligence is correlated with:
–> the average social sensitivity of group members
–> the equality in distribution of conversational turn-taking
–> the proportion of females in the group
diversity within the group (can hinder or help depending on the task
–> cognitive diversity (e.g., thinking styles) - takes more than a group of smart individuals to make a smart group
- in creative innovative tasks, diversity assists
- if efficacy is important, diversity is not helpful
43
Q
Wooley et al (2015)
A
- 272 participants
- 68 groups
–> 34 groups online
–> 34 groups face-to-face
44
Q
Wooley et al (2015) - experiment tasks
A
- reading the mind in the eyes
–> measure of social cognition and ability (theory of mind)
–> guess someone’s mental states by looking at their eyes - choosing intelligence task
–> sudoku was used as one measure of intelligence and was conducted either online (where communication was via text) or face-to-face - Raven’s Advanced Progressive Matrices test
- generating task
–> intelligence task
45
Q
Wooley et al (2015) - results
A
- dominant collective intelligence factor
- explained variance in results more than any other factor
- social intelligence factor
–> online and offline average RMET scores predicted average collective intelligence
–> i.e. social sensitivity is important to group functioning - strong empirical evidence for the emergence of collective intelligence in online groups
- theory of mind abilities are a significant determinant of group collective intelligence
–> even when group is online and has limited communication channels
46
Q
predictors of group intelligence
A
- average social sensitivity
- amount of communication
- distribution of communication
- strongly suggests that the coordination problem of group work often outweighs the intellectual challenges
47
Q
summarise group cognition
A
- Rather than “groups are smart” vs “groups are dumb” consider:
–> what factors affect when groups are smart
–> individuals bring information (but also biases to groups)
–> how the group functions can average out (uncorrelated biases) or overcome them (through argument)
–> communication can be as important as knowledge or intelligence
