SPRING Reasoning and judgement Flashcards
describe the sally clarke problem
both babies die from sudden infant death - very unlikely
argued that prob so low, more likely to be murder
but lawyers used poor statistical reasoning - two events were not independent of eachother and the prob of two independent events not the same as two linked ones (prosecutors fallacy)
ignores probabilities of two murders in same family
contraceptive probabiities
98% vs 99% effective
over time, 98% sig less effective than 99%
how do we make subjective probability judgements
decision making based on SEU
heuristic strategies to judge the likelihood of event occurance
describe the availability heuristic
judge probability based on ease comes to mind
if more freq then more available
tversky and kahenman availability heuristic
more common: r as first or third letter
most say first - easier to think of words beginning with r
slovic, fichoff and litchenstien 1979 availability heurostic
estimate freq of causes of death
overestimate those unlikely but heavily reported ie murder and underestimate those rarely reported ie emphysema
describe the representativeness heuristic
degree of correspondence between an instance and a category
ie looks and sounds like it so probably is
how much a certain set of circumstances will drive the likelihood of something based on past experience - link closely to stereotypes
tversky and kahenmen 1983 conjunction fallacy (representativeness heuristic)
liinda - bank teller, feminist bank teller, feminist
often chose feminist bank teller but statistically the least likely
two independent events make conjunction of two together even less probable - feminist bankteller is smaller subset of banktellers
describe anchoring heuristic
base estimated probability on the first piece of information we are given
tversky and kahenman 1974 anchoring
estimates under time pressure 1. 1x2x3x4x5x6x7... 2. 7x6x5x4x3x2x1 = ~40000 BUT estimate 1 lower than 2 based on first numbers
lichtenstein et al 1978 anchoring
achoring on first given value means fail to adjust appropriately
estimate freq of 40 causes of death
a. 50000 motor cycle or b.10000 electrocution
estimate higher for a
real world examples of anchoring
anchoring consumers for price reference so view as less expensive
credit card minimums
high prices as ref point for lower
what is bayesian reasoning
update probabilities based on additional information that causes to ignore the original base rate and focus on hit rate
make probabalistic inferences based on additional info and therefore forget the initial info given
kahenman and tversky 1973 base rate fallacy
30 engineers and 70 lawyers stereotypical description of engineer/lawyer or neutral neurtral tend to report 50% rate increase if stereotypical BASE RATE ALWAYS THE SAME: 42%
medical diagnosis problem base rate fallacy
casscells et al 1978
prob breast cancer 1%…given values of pos mamog if have cancer = 80% and not = 9.6%
= overall 7.8% prob
clinicians probability = 70-80% as ignore probability of 1% and focus on hit rate
gigerenzer, hofferage and ebert 1998 bayesian reasoning
positive HIV test - prob that ACUTALLY positive?
clinicians say 100% even if low risk patient
gigerenzer and hofferage 1995 format of decision
if way info is presented is more reflective of the way we make decisions in real life then more likely to interpret in correct way
ie breast cancer problem in frequency format and most people get correct
cosmides and tooby 1996frequentists hypothesis
why are frequency problems easier
inductive readoning = prob in the form of freq
easier to break down if in freuqncy form - tree
girotto and gonzalez 2001 defective frequencies
if not given enough info to complete freq tree OR the information given is all drawn from multiple samples
how might you improve bayesian reasoning problems
change structure of the info not the format
apply cumulative reasoning
what are naive probabilites
when people make judgements on probabilities they are not expert in - relate to the alternatives they consider
ie not given probabilties then assume equiprobable
girotto and gonzalez 2001 prob and freq use
prob not alwats bad ie can format as chance to improve interpretation
freq not alwats good - come harder than probabilities, dependent on the info structure
mccloy byrne and johnson laurd 2010 improveing bayesian readoning
naive individuals dont use binomal formula to calc cumulative prob and instead use conjunctions ie if not yet occred then the prob = 0/no. trials
those who recog that freq cumulates over time may still underestimate the rate of increased risk
improve by presenting problems in specific ways - ie more structured layout
johnson laird et al
sentanial resoning
when an individual understands discours and imagines a state of affairs then constructs mental models of the relevant situation
prepresents the possibilities and diff ways they might occur - represents explicitly what is true but not what is false
each model therefore represents an equiprobable alt unless info given to hold belief of the contrary
the greater the no of mental models a task elicits and the greater their complexity the worse the performance
equiprobable sentenial reasoning example
box with red or black marble or both
prob of black marble with another marble?
- tend to say 2/3 as no info given about prob - ill specified
define prosecutors fallacy
tendency to ignore that to occurances are not statistically independent from one another and therefore the probability of the two occuring is lower (more likely) than if they were independent
how did sally clarkes case fail to prosecutors fallacy
argued deaths of two newborns statistically independent = 1/72m and that this equates to the prob that guilty
BUT NOT - sig lower prob because share same environment
define bayesian reasoning
a means of quantifying uncertainty
rule for refining an hypothesis by factoring in additional evidence and background information, and leads to a number representing the degree of probability that the hypothesis is true
- ie accounting for independence of events in prosecutors fallacy
define subjective probability
the extent to which a coherent person believes that a statement is true based on the info that is available at that time
use of heuristics to form a subjective judgement
seidelmeier hertwig and gigerenzer 1998
availability heiristic
memory often good at stoing freq info
tend to measure ease info comes to mind based on no instances retrieved in given time frame and speed of retrieval
BUT - some predictions where overestimate low freq and underestimate high so not always completely accurate
describe linda
t+k conjunction fallacy
31 y/o outspoken and bright
major in philosophy
concerned with discrimination and social justice
participates in antinuclear demonstrations
diff subective prob judgements
availability heuristic
representativeness heuristic
anchoring
bayesian reasoning prob
base rate fallacy
medical diagnosis prob
hiv screening
prob vs freq vs chance
