w12+w13 decision making Flashcards
decision making requires the following processes:…
Goal-directed behavior
Reward Learning & Memory
Executive Control
Inference to establish causal links
Often in a social context
how economists view decision-making
Homo economicus: rational own-payoff maximizer. Sophisticated rational beliefs. Stable preferences. Maximizes own payoff. Disregards other peoples’ wellbeing
But this creature doesn’t exist because we have cognitive biases
expected value
Expected Value of an offer is § Its magnitude § Times its Probability
utility
subjective value. Valuation dependent on starting position, for example margial utility
risk
known variance of a distribution
so the situations where you know the probabilities of the outcomes
ambiguity
unknown variance of the outcomes. Situations where you don’t know the probabilities.
risk aversion
most people prefer a lower, certain reward over a risky higher (on average) reward
measured by how low can the certain offer be so that the person still chooses not to gamble. Certainty equivalent
certainty equivalent
how low can the certain offer be so that the person still chooses not to gamble
shows risk aversion level
utility curve and risk aversion for loses
CE for losing and gaining
it’s low for gaining (will agree for a small amount of money for certain instead of gambling) and high for losing (will only agree to lose a certain amount when it’s really close to the highest he could lose, otherwise would gamble to lose nothing)
reflection effect
People become riskseeking in the loss domain. Losses loom larger than wins. Less losses required to offset positive utility gains. Implications for framing (coming up)
dopamine in the brain
dopamine: role in learning
Acts as a teaching signal: Reward Prediction Error, Rescorla-Wagner learning rule: basis of associative learning
Learning from (un)- expected outcomes
Rescorla-Wagner learning rule
extrinsic motivation and dopamine
they produced a lot of dopamine (the nucleus accumbens lights so brightly on the imaging) during the first trial, but then the extrinsic reward was taken away (and for controls they were always just told that the game is interesting), and they produced no dopamine at all
what does this mean?
it means that the brain codes for expected utility (so the reward’s magnitude times probability) rather than only magnitude or only probability
temporal difference learning
and it is related to how the neurons fire. that you can see it at the graphs of the relationship between the cue, the probability and the reward occurring.
which cells is this recorded from and why?
VTA - ventral tegmental area. Because they produce dopamine for the whole mesocortical pathway.
what happens if the cue is learned to predict a negative event?
some cells in VTA fire. So dopamine also predicts negative events.
what does risk do to the brain?
activates a network of brain regions
Importantly: insula. Potential role in anticipating negative outcomes. Interoceptive signal, for risk aversion
ambiguity: the reaction of the brain
so some people prefer ambiguity over risk
y Many people are averse to ambiguity, which is defined by economists as a situation in which one must make a decision without knowing the probabilities of the possible outcomes.
(A) Compared to risk in decision making, ambiguity in decision making evokes increased activation in the orbitofrontal cortex, perhaps because of its aversiveness.
(B) Ambiguity also evokes increased activation in lateral prefrontal cortex, which may reflect the need to think about the possible probabilities and to construct a decision rule. Supporting this latter interpretation, the magnitude of activation in the lateral prefrontal cortex during decision making under ambiguous circumstances depends on whether the subject prefers or is averse to ambiguity.
how to conceptualize uncertainty?
in terms of delay discounting.
Future rewards might not materialize. Waiting incurs opportunity costs - costs of not doing anything else. So you have to somehow measure how big the reward has to be so that an individual would wait for it (large delayed reward magnitude), and also how short the waiting period should be (large delayed reward delay).
the steaper the discounting curve - the more likely is the animal to choose the immediate reward over the delayed reward
preference reversal
because of the different steepness of the curve around zero and in the remote future the preferences for the small or for the big reward reverse over time
so the real preference is inconsistent with exponential decay and consistent with hyperbolic discounting
pro-social behavior
what drives pro-social behavior?
also cooperation, like in game theory: if they cooperate, every agent wins more, prisoner’s dilemma
social vs non-social valuation
it’s about if the social value and non-social value are calculated in the same way
which part of the brain is activated by benefitting the charity? and why
ventral striatum (with nucleus accumbens), because the rewards themselves activate VS, but also because we act pro-social, so the reaction to the reward to the self activates the same VS as the reaction to the reward to the charity, the areas overlap
trust vs learning
and the participants also believed if you told them the story at the beginning. They trusted more if the bot was represented as trustworthy, and less otherwise
neural signatures of reputation
