Lecture 11&12&13 - Decision Making Flashcards
homo economicus
= rational own-payoff maximizer
- sophisticated rational beliefs, stable preferences, maximizes own payoff, disregards other peoples well-being
-> factoring out each potential bias (ex. time pressure, stress, …)
therefore, not real-life applicable …
decision making - processes
- goal-directed behavior
- reward learning & memory
- executive control
- inferece to establish causal links
- often in a social context
economic vs psychological decision making model
economic: rational (but choice bias)
psychological: context dependent choices (social, cultural, moral)
idea of neuroeconomics
= combination of psychology, economics and neuroscience
-> including social context of decision making, evaluates underlying circuits in value-based decision making, investigates which brain chemicals are involved
why studying decision making?
- phylosophically (value of decisions)
- economically (consumer decisions; nudging (leading decision making))
- clinical (maladaptove decisions)
+ gambling disorder, impulsive traits, anto-social or norm disregarding decisions
expected value
multiplication of objective value times probability (magnitude, times itr probability)
risk vs ambiguity
risk = known variance of distribution (gamble effect) ambiguity = unknown variance (outcomes); unknown probabilites, need to infer about those
reference effects
utility is a subjective value, everyone has own starting/reference point (= marginal utility)
risk aversion
preferring lower certain reward over risky higher (on average) reward
-> valuing certainty over gamble (everyone has own subjective treshold about that)
[certainty equivalent shows risk aversion level]
[for gains, the mean utility of the win and loss gamble is worse than the utility of the (mean) certain outcome]
when do people become risk seeking? (reflection effect)
in the loss domain
-> the curve in the loss domain is much steeper than in the gains domain (loss is waited much more heavily)
[for gains, the mean utility of the win and loss gamble is better than the utility of the (mean) certain outcome]
-> losses loom larger than wins
- less losses required to offset pros -> utility gains
- implications for framing
ambiguity aversion
preference for known risks over unknown risks
probability weighting
- people overestimate small chances
- people underestimate large probabilites
purpose of rewards
reinforce behavior
-> primary or secondary reinforcers
[remember law of effects, William James]
role of dopamine (+ where it is produced)
associated with rewards
dopamine - basics
- > produced in midbrain and basal ganglia (substania nigra) [also adrenaline], its a catecholamines
- > innervation of large parts of the cortex
- > changes firing probabiliy of neurons (motor and cognitive effects, therewith effects decision making)
dopamine & reinforcement learning
dopamine as teaching signal:
- reward as expected = no dopamine firing rate change
- reward better = increased dopamine firing rate
- reward worse = decreased dopamine firing
- > learning from (un)expected outcomes
- > basis of associative learning
Rescolra-Wagner learning rule & prediciton errors
associative learning is strongest when difference between prediciton and outcome is the largest (surprise effect, positive or negative)
[outcome - expectation = prediciton error]
meaning of undermining motivation
- intrinsically motivated group; reatied rewards signals
- incentivized group (externally motivated); lost reward signals [when external reward is gone, there is also no internal rewards anymore]
- > differential Dopamine releases
Dopamine as prediciton error signal
- unexpected reward occuring -> firing increases
- cue predicts reward fully -> firing increases
- oucome lower than expected -> firing rate decreases
- > information
dopamine drives learning
-> biases choice behavior towards cues where more dopamine neurons were active (dopamine coactivating receptors leading to enjoyment)
temporal difference learning
reward calculation; increase in signal when known that reward will come
-> reward signals influence processing in whole brain
which brain area is activated in the cause of unexpected events?
the basal ganglia
risk (considering ongoings in the brain)
risk activates network of regions
- insula; potential role in negative outcomes, interoceptive signal, for risk aversion (affecting decision making)
ambiguity - when information is missing (brain areas, ongoings)
- lateral and orbitofrontal cortex (higher oder areas); infering about range of outcomes -> executive planning (more active for ambiguity-seeking individuals)
- > requires strategizing (more than simple risk analysis)
delay discounting (hyperbolic delay discounting)
future rewards less values than rewards nearby
-> wating for reward incurs opportunity costs (cost of not doing anything else)
Hyperbolic DD: preference of which reward to take depends on large, delayed rewards magnitude and amount of delay
-> preference reversal; preference in small early rewards on small time scales and reversal when brought to larger timescales
-> the steeper the discount curve in the future, the less likely it is waited for the reward
pro-social behavior (its drive) - reciprocity
-> human tend to be pro-social even in a one-shot situation
stabalizes long-term social bounds (best genes survive), increases group fitness (reputation tracking)
-> benefiting from social engagements (direct or indirect (ex. donating) activates ventral stratium, just like for self-rewards)
cooperation
activates reward centers; mutual cooperation giving best results
-> same activation when reward given to oneself as when given to others (social context)
learning whom to trust
- reputation matters (update trust based on feedback)
-> capacity (ability grows with development (getting older)
Influences on trust: - vignettes: fictional background stories
- fixed trustworthiness
social discounting
generosity towards others decreases accross social dustance between actor and recipient (more generous to people we feel close to than to strangers)
- > generosity declines hyperbolically across social distance
- > TPJ, VMPFC, insula frame effect on social discounting
involved brain areas in social discounting
junction of VMPFC and TPJ
- > temporary deactivation of TPJ increases social discounting
- > can be influenced (ex. by giving noradrenaline), but also depends on the progestrone levels accross females mentrual cycle
gain vs loss frame (neural processes?)
they are different for both frames and can therfore be mapped on distinct neural networks
-> motives underlying generostiy are different
gain frame (neural mechanims and "feeling") - strong decline in generostiy across social distance
TPJ upregulates vicarious reward value signlas in VMPFC in a social-distance dependent fashion -> increase the value of generous option
-> being generous is rewarding: VMPFC and TPJ
loss frame (neural mechanisms and "feeling") -> independent of social distance
insula downregulates onw-reward value signals in VMPFC in a social-distance dependent fashion -> decrease the value of selfish option
[motivatd by social norms]
-> generosity is motivated by guilt aversion: anterior insula and VMPFC
effect of stress in social discounting
-> dampens framing effect
stressed participants are much less generous towards starngers in the loss frame, but generosity towards family/friends increases
-> increased fight or flight behavior (seeking comfort and help in stress situation)
-> stress counteract social discounting
-> insula activation; harm avoidance
5 cooperation rules - Nowak
- kin selection: when related to person with percentage r, there is incentive to promote genes
- direct reciprocity: helping each other
- indirect reciptrocity: helping someone, which helps someone else (no direct return)
- network reciprocity: find trustworthy group
- group selection: starts own group found in network reciprocity
value-based decision making brain areas
- ventro-meidal PFC; codes abstract rewards signals
- lateral PFC; executive control
heuristics
when there are too many choice dimesniosn implied, heuristics are used; using smaller sets of decision features (amount which can be evaluated)
-> bounded rationality, shortcuts (anchoring discounts and endowment effects)
choice influences
- risk/ambiguity
- social contexts
- trust and norms
- > humans are biased and make use of heuristics