Lecture 16: Decision-making - Uncertainty and Risks Flashcards
Heuristics and biases
¤ Heuristic: Mental shortcut or rule of thumb that can be used to get a quick and mostly accurate response in some situations but may lead to errors in others
¤ Bias: deviations from rationality (errors) that are caused by using heuristics
Three ‘categories’
¤ Biases that affect how we interpret information
¤ Biases that affect how we judge frequency (how often something happens)
¤ Biases that affect how we make predictions
Availability
¤ Availability: estimate the probability of an event based on the ease at which it can be brought to mind
¤ What is more common? Words that start with the letter “R” or words in which the third letter is “R” ?
¤ Overestimate the probability of a shark attack after watching JAWS
¤ Can explain why people are afraid of flying but not driving
Representativeness Heuristic
¤ Tom W. is of high intelligence, although lacking in true creativity. He has a need
for order and clarity, and for neat and tidy systems in which every detail finds its
appropriate place. His writing is rather dull and mechanical, occasionally enlivened by somewhat corny puns and by flashes of imagination of the sci-fi type. He has a strong drive for competence. He seems to feel little sympathy for other people and does not enjoy interacting with others. Self-centered, he nonetheless has a deep moral sense.
¤ Is Tom studying psychology or Library Sciences?
Psychology is a big program and library science is not even a program
Representativeness Heuristic
¤ Representative Heuristic: tend to make inferences on the basis that small samples resemble the larger population they were drawn from. The idea that we judge how close they are to the stereotypical idea of that concept (how close is Tom to the stereotypical psychology student).
- Related to stereotypes, schemas, and other pre-existing knowledge structures
- People base their judgements of group membership based on how similarity
¤ This results in biases like base-rate neglect & conjunction fallacy
Base rate neglect
¤ Base Rate Neglect: When you fail to use information about the prior probability of an event to judge the likelihood of an event
¤ Imagine running an HIV test on population of 1000 people, in which only 1% are infected. The false positive rate (falsely diagnosing someone) is 5% with no false negatives. If a test comes back positive what is the likelihood someone has HIV? We need to take into account how accurate the test is but also how likely it is for someone to have the disease. Have to take the test multiple times.
* It is actually 17%
* 10 people in the population have HIV
* 50 people without HIV would be falsely identified
¤ Important for doctors diagnosing with low incidence populations
* Need to consider the base rate (prior probability of someone actually having HIV) before diagnosing ( need to take into account if the disease is very rare)
* Like in the Tom W problem where people don’t consider the prior probability of being a librarian (take into consideration how likely it is that he is a librarian - how big is the program?)
Conjunction Fallacy
¤ Conjunction fallacy: False belief that the conjunction of two conditions is more likely than either single condition
- the probability of one event happening is more probable than the 2 events happening.
¤ Linda the feminist Bank Teller
¤ Because the description was more representative of both categories people think the conjunction is the most likely label —> this is not the case tho (small probability that both circles overlap)
Anchoring and Adjustment
¤ Anchoring and adjustment are too heavily influenced by initial values (that you have).
¤ Which product is larger?
¤ People start off with one value and adjust accordingly from there
¤ Important when getting ratings from a scale
- because of the order people will look at the first number and then adjust from there how big they think this multiplication is.
- If you ask people to move the scale, we are already introducing a bias (by choosing an anchoring point). If you start it closer to the more anxious then more likely that people rate themselves as being less anxious.
Regression to the Mean
¤ Regression toward the mean: when a process is somewhat random (i.e. weak
correlation), extreme values will be closer to the mean (i.e. less extreme) when measured a second time. Correlation is a degree of strength between a relationship.
¤ Related to illusionary correlations.
¤ People tend to see causal relationships even when there are none
Value goes towards the mean - how anxious you are feeling will go to the mean of how you typically feel anxious?
Example: Regression Toward the Mean
¤ A pageant mom rewards her daughter when she preforms unexpectedly well and
wins a pageant. But the next pageant she comes in last place, the mom punishes
her and the following competition she does well. The mom concludes punishment works better than rewards.
¤ Related to our understanding of the roles of reward and punishment on learning
- Can’t always attribute changes in performance to manipulations
- Sometimes it’s just noise
- The chance is that she will get somewhere in the middle of the two. It is more about how random this process is not a direct consequence to the reward or punishment.
Bounded Rationality
¤ Why use Heuristics? People are thought to be Bounded rational (Simon 1957) meaning they are limited by both environmental constraints (e.g. time pressure) and individual constraints (e.g. working memory, attention). Things are affecting our ability to have cognition.
¤ People are Satisficers: look for solutions that are “good enough”. Spare some time and cognitive capacity to solve other problems that are more important.
¤ “Making do” with the limitations we have as humans
¤ Although heuristics sometimes provide incorrect answers and lead to biases; they also work
- heuristics is the adaptive method that we have come to.
Ecological Rationality
¤ Gigerenzer proposed an alternative view to heuristics called Ecological Rationality which sees heuristics not as a “good enough” approach to solving a problem but as the optimal approach. He says essentially it is all based on your environment - given the right environment a heuristic can be better than a more complex statistical approach.
¤ While previous views on heuristics drew a separation between how we should act and how we do act, Ecological rationality does not distinguish these two
¤ Given the right environment, a heuristic can be better than optimization or other complex
strategies
- Choosing the most optimal solution for the problem in the specific environment.
Example: Ecological Rationality
¤ Say you have some money you want to invest and a bunch of options to choose form, but limited information about how risky each one is or the past performance…
¤ Equally dividing your assets (money) among the options (1/N heuristic) has been shown to
provide better results than other more complex optimization algorithms
¤ Sometimes heuristics (give the right circumstances) can be better than complex strategies
Summary: Heuristics and Biases
¤ Heuristics and biases arise from the limitations we face but can sometimes produce correct responses
* Applying heuristics too often can lead to biases
¤ Several examples of Heuristics
* Availability
* Representativeness
* Anchoring and adjustment
* Regression towards the Mean
Decision-making
¤ Kinds of Decision-making
¤ Decision-making under uncertainty and risk
* When an given action has several possible outcomes
¤ The Framing Effect
* The difference between framing an outcome as a Loss vs Gain
¤ Prospect theory
* How should we make decisions? Vs How do we make decisions?
¤ Emotional factors affect decision making
Kinds of Decision-Making
¤ Perceptual Decision Making: objective (externally defined) criterion for making your choice
* Are the dots moving left or right?
Value-based Decision making: subjective (internally defined) criterion for making your choice. Making a decision that is subjective - based on your personal preference.
* Do I want cake or ice cream for dessert
* Depends on motivational state and goal
¤ Risk can be defined as taking an action despite the outcome being uncertain
* Specific to Value-based decision making
¤ Ambiguity can be defined when you have incomplete information about the consequences.
- ex: would not know what the probability of winning or what i would win. I dont know something about these specific consequences.
Endowment Effect
¤ Once ownership is established, people are averse to give it up
¤ (Kahneman, Knetsch, & Thaler, 1990)
- people with the mug are willing to sell the mug for a lot more than people are willing to pay for it
-once you are giving something away that is yours, you put a premium on it.
Prospect Theory
¤ Utility function: Describes how people map money to satisfaction (Maps the dollar value to the satisfaction value)
¤ The extra satisfaction earned from gaining a dollar is larger when you only have $1 vs when you have $1M (when you already have a lot of money adding 1$ does not add much happiness)
¤ Asymmetrical: Steeper for Losses than Gains $1 lost hurts more than one dollar earned. Dislike loosing things more than gaining something.
¤ Losses loom larger than gains (framing effect)
Example: Prospect Theory
¤ Let’s say you’re at the pizza parlor ready to eat pizza
¤ You can map pizza to the satisfaction you get from it (Utility function)
¤ Being given 1 pizza slice is worth more to you when you had none then when you have 12
¤ Happiness (utility) of receiving a slice of pizza is less than the sadness (disutility) after giving a slice away. Loss aversion - much more dissatisfied with loosing a slice than gaining one. .
Prospect Theory
¤ Probabilities are not treated objectively
- Extreme events tend to be rare (overestimate how unlikely rare events are and underestimate how likely common events are).
¤ What causes more deaths: dying in a car crash or dying of cancer?
¤ Extremity of event related to perceived probability
- Unlikely events are overestimated
- Likely events are underestimated
¤ Availability of an option changes the perceived frequency of occurrence. More extreme events tend to me more rare and since common events since to be less extreme in terms of affect people tend to overestimate it.
Risks as feelings: Gains, loss & emotion activation
¤ Participants made choices between two outcomes framed as gains or losses
- A risky (A gamble) outcome
- A safe (A Sure thing) outcome
¤ Increased amygdala activity for chosen safe outcomes for gains and chosen risky outcomes for losses suggests that an emotional response may underlie the framing effect
- activity in frontal cortex was higher when making decisions against the framing effect.
¤ Choices are influenced by your affect (Lowenstein et al., 2001)
Assessing risk and emotion
¤ Participants read newspaper stories designed to induce positive (happy) or negative (sad) affect
¤ Then, the participants estimated frequencies of death for various causes
- High risk
- Non-fatal risk
- Low risk life problems
¤ There were higher estimates of death frequency when people were in a negative mood compared to a positive mood.
- people who read negative stories were most likely to have negative mood and therefore chose negative effect.
Mood affect real world Gambling
¤ Prediction errors in sports outcomes and the weather have been found to affect people’s mood
¤ Positive PE increases positive affect (better mood)
¤ Negative PE increases negative affect (worse mood)
¤ Changes in mood predict risky decision-making - when people are happy they are more likely to gamble
- if there is a prediction error in the sunshine, then you have a positive affect, and lead to happier and more gambling.
- Tight relationship on how your mood effects your risky decision making
