what are descriptive models?
models of how decision processes operate, irrespective of whether the outcome of the decision is good or bad
what are normative models?
evaluate a decision in terms of the goals of the decision maker
decisions can be good or bad and a decision is good if it reaches these goals
what is rational thinking?
normative - a decision is good to the extent that it enables the decision maker to reach their goals
what are rational decisions?
maximises expected value
what is rational choice theory?
assumes agent has full access to all alternatives
knows the probability of their outcomes, knows the value of their outcomes
can integrate all this information and chooses the option with highest utility
what is bounded rationality?
limited cognitive capacity and information on outcomes
human behaviour can be impulsive, habitual/conditioned, imitating others, random
what do we do when making choices under uncertainty?
given choice between two or more alternatives, rational choice is the option with the highest value or utility
if outcomes are uncertain, multiply the utility by the probability of receiving that outcome to give expected value, EV
what is the expected value (EV)?
p(outcome) * value
option A - £100 with 50% probability, EV = £50
option B - £1000 with 2% probability, EV = £20
in RCT, A is better
what is the St Petersburg Paradox?
Bernoulli (1738) - Peter tosses a coin and continues to do so until it should land on “heads” when it comes to the ground. He agrees to give Paul one ducat if he gets heads on the first throw, two ducats if he gets on the second, four if on the third, eight if on the fourth and so on, so that with each additional throw the number of ducats he must pay is doubled
EV is infinite = 1/2 * 1 + 1/2 * 2 + 1/2 * 4 ….
RCT suggests player should be willing to pay virtually any sum of money, since EV is infinite
although EV of gamble is infinite, subjective utility is low, since change to win high value is low
case where RCT fails to make sensible predictions about “rational choice”
how did Bernoulli explain the St Petersburg paradox?
EV not equal to EU instead the utility of wealth is proportional to its logarithm
consequences of this is that each additional unit of wealth is worth less that the previous one so the utility of additional currency units decreases as the number of currency units increase
extra utility of the high winnings in the St Petersburg Paradox is no longer high enough to compensate for the very low probabilites
problem with this, in turn, is that a game could enhance the “win” function so the problem persists
what does the St Petersburg Paradox show?
inherent limitation of EV
despite its seemingly plausible or trivial nature, not general enough for many situations even under assumption that decision maker behaved entirely rational (in terms of maximising their gains)
what is expected utility theory?
normative theory
clearly defines what rational choices are
expectated utility may, however, deviate from EV in that allows EU of a particular choice outcome to be situationally dependent - on the accumulated wealth
weak ordering - for any set of choices, we must always be able to say we prefer one over the other or neither
what was Kahneman & Tversky’s (1979, 1992) experiment 1?
would you prefer a certain gain of £3000 or an 80% chance of gaining £4000 or else nothing
would you prefer an 80% chance to lose £4000 otherwise nothing or a certain loss of £3000)
most people prefer a sure gain of £3000 over 80% chance to gain £4000 although EV of risky option is higher
preferences are reversed when outcomes are losses - most people now prefer risky option over certain option
participants behaviour reflects (fairly consistently) exact opposite of what RCT has predicted in both cases
moreover, and significantly, participants don’t appear to generally favour the certain option - only favour certain option when it comes to making a gain
what was Kahneman & Tversky’s (1979, 1992) experiment 1?
would you prefer a certain gain of £5 or a 1/1000 chance of gaining £5000
would you prefer a certain loss of £5 or a 1/1000 chance of losing £5000
participants again deviate from RCT and invert their choice patterns when small probabilities are involved and more extreme gains or loses
what is prospect theory?
Daniel Kahneman & Amon Tversky
developed as a descriptive model of decision making
intended to account for deviations from RCT
two main components - utility and probability
what is the value function (Tversky & Kahneman, 1981) in prospect theory?
proposed a value function that differs for gains and losses
x-axis reflects gains to right and losses to left instead of total wealth - midpoint as person’s current reference point
y-axis is utility
function for gains and losses is asymmetric - explains why we treat losses as more serious than gains
besides the different overall slop of straight lines for gains and loses however the value - utility functions are non linear
for both gains and loses, increments near small values lead to considerably bigger increase in gains/losses whereas for large values, increase in utility is smaller
what is the π function in prospect theory?
for gains, people prefer certainty - certainty effect
for losses, prefer to gamble
according to prospect theory, don’t treat probabilities as they are stated instead distorted by π function - objective p becomes subjective π
probabilities near 0 are overweighted relative to objective probabilities
probabilities near 1 (but not 1) are underweighted
weighting functions are different for losses vs gains - subjective probability of a loss is deviating more from linearity than for gain
what is a summary of prospect theory?
assumes differential sensitivity to gains and losses, explaining loss aversion
assumes nonlinear functions of utility as function of value
takes into account that subjective probabilities are not a linear function of probability but sigmoid function which leads to overestimation of small probabilities and underestimation of high probabilities
descriptive model established on basis of empirical data and takes subjective factor, for both value and probability, into account
why is prospect theory not a complete theory?
although describes some deviations from normative expectations by means of the pi function and value function, does not give any psychological reasons for shape of these functions
what is regret theory?
we compare outcomes, particularly after decision is made
assumption is, we will regret a decision if an alternative outcome would have led to a higher payoff even when prospect was better
we rejoice if our choice led to a better outcome than other alternatives
what is an example of inter-temporal choice?
would you prefer £500 now or £1000 in 10 years
most people prefer option a which has a lower value than b but has a higher utility because the utility of b is discounted by delay
who came up with the Discounted Utility Model?
Paul Samuelson (1937)
what is the normative assumption about the discount function?
exponential constant rate
what is exponential discounting?
v = Ve^(-kd) where v = subjective discounted rate, k = discount rate, d = delay
three properties that fit empirical behaviour
if there is no delay, no discounting
as delay increases, present value decreases
as delay approach infinity, present values approach 0
-
introduction to perception46
-
the visual pathway 155
-
the visual pathway 245
-
colour perception47
-
beyond v121
-
object recognition40
-
face perception30
-
false and recovered memories32
-
encoding failures26
-
effective encoding29
-
storage failures25
-
active forgetting23
-
types of retrieval22
-
improving retrieval22
-
autobiographical memory21
-
visual word recognition 133
-
visual word recognition 267
-
embodiment24
-
shallow processing29
-
perception and bayesian inference35
-
bounded rationality35
-
heuristics27
-
attention and consciousness39
