Lecture 5-6 Flashcards
Why study normative models?
We need to understand optimization to try to approximate it. Also helps us understand decision making. Easier to understand than descriptive as they are consistent and follow rules.
Why study normative models?
We need to understand optimization to try to approximate it. Also helps us understand decision making. Easier to understand than descriptive as they are consistent and follow rules.
Objective standards
Physical, biological, economic criteria
Subjective standards
Happiness, well-being, utility, reference points, multiple goals
Constrained optimization
Decision making in the real world must be understood as constrained optimization. There are physical constraints (engineering), logical constraints (axioms of normative decision making), cognitive and emotional constraints (descriptive decisions)
Perceptions of randomness
People see patterns when there is actually randomness - confirmation bias
Coincidences seem extremely unlikely due to the fact that we tend to focus on what is the same, not what is different. Confirmation bias. We also tend to focus on the particular (such as names) instead of focusing on the coincidence of an unspecified match of the same category. Latter is much more likely - see birthday problem.
Randomness is also mistaken for patterns in the lab - correct guesses make participants superstitious and more likely to choose the alternative the next time.
Gambler’s Fallacy
Idea that probability is self correcting - after many heads, it has to be tails. Reverse Gambler’s Fallacy - superstition that something has to keep on happening. Of course, sometimes events are not independent - earthquakes become increasingly likely, for example, the longer it’s been since previous one.
Law of small numbers
In a small sample, if there are streaks then we believe that something non-random is happening even if it is random.
Why do people tend to see patterns in randomness?
Rewards for recognizing real patterns might be greater than penalties for imagining patterns.
Illusory correlation
Important to view ratios when comparing, not just the numbers themselves.
Post-hoc fallacy
If event B closely follows event A, we often think A caused B. That’s why there are people who believe vaccinations are harmful. Temporal connections are compelling; we often ignore many events between A and B. We pay more attention to what is there than what isn’t there
Positive vs. negative evidence
We tend to pay more attention to positive evidence than negative evidence. Confirmation bias
Third factor
Mistaken causation b/c ignore possible presence of a third factor related to both.
Correlation vs. causation
Correlation is just when A is related to B, can be with or without causation. There may be third variables causing correlation. For causation, one must precede the other in time.
Objective standards
Physical, biological, economic criteria
Subjective standards
Happiness, well-being, utility, reference points, multiple goals
Constrained optimization
Decision making = objective function + constraints
Decision making in the real world must be understood as constrained optimization. There are physical constraints (engineering), logical constraints (axioms of normative decision making), cognitive, emotional, and socialconstraints (descriptive decisions)
Perceptions of randomness
People see patterns when there is actually randomness - confirmation bias
Coincidences seem extremely unlikely due to the fact that we tend to focus on what is the same, not what is different. Confirmation bias. We also tend to focus on the particular (such as names) instead of focusing on the coincidence of an unspecified match of the same category. Latter is much more likely - see birthday problem.
Randomness is also mistaken for patterns in the lab - correct guesses make participants superstitious and more likely to choose the alternative the next time.
Availability heuristic
Things that are more easily available in memory, primed, vivid, or easier to think of (more examples) are judged as occurring more frequently. This is problematic in news as we can bring to mind catastrophic events more - plane travel seems riskier. Being killed by a shark seems more likely than being killed by falling airplane parts, although the latter is 30x as likely
Pros: many times, frequent events are easier to recall
Cons: Ease of recall often driven by factors other than frequency. Selective exposure.
Law of small numbers
In a small sample, if there are streaks then we believe that something non-random is happening even if it is random.
Why do people tend to see patterns in randomness?
Rewards for recognizing real patterns might be greater than penalties for imagining patterns.
Illusory correlation
Important to view ratios when comparing, not just the numbers themselves.
Post-hoc fallacy
If event B closely follows event A, we often think A caused B. That’s why there are people who believe vaccinations are harmful. Temporal connections are compelling; we often ignore many events between A and B. We pay more attention to what is there than what isn’t there
Positive vs. negative evidence
We tend to pay more attention to positive evidence than negative evidence. Confirmation bias
Third factor
Mistaken causation b/c ignore possible presence of a third factor related to both.
Correlation vs. causation
Correlation is just when A is related to B, can be with or without causation. There may be third variables causing correlation. For causation, one must precede the other in time.
Maximizing vs. satisficing
Maximizing - normative - find best outcome regardless of time and energy spent
Satisficing - descriptive - finding first outcome above certain criterion (good enough)
We satisfice to save time or when a decision is relatively unimportant
System 1 vs. System 2
System 1 - hot system - emotion, associations, memory. Fast and automatic. Associated with satisficing
System 2 - cold system - reason, calculations, abstract and logical. Slow and effortful
Cognitive heuristics
Mental shortcuts or rules of thumb that take less effort than deliberating using system 2, and that usually give us decent results.
Pros and cons of heuristics
Advantages: less complexity, functional, fast, usually provides decent results
Disadvantages: possible biases or errors, usage often unconscious, errors may persist even when we are aware
Anchoring & adjustment heuristic
When we estimate unknown quantity, we start with convenient initial value (anchor) and shift upwards or downwards (adjust). Anchor may be a known value, a primed value, a calculated value, or a given value. We do this even if the anchor is obviously extreme or irrelevant.
Pros: anchor is often a good starting point, and adjustment gets us close enough
Cons: easy to weigh anchors too heavily and adjust insufficiently. We can be manipulated. Combat by generating equally extreme alternative anchors.
Availability heuristic
Things that are more easily available in memory or easier to think of (more examples) are judged as occurring more frequently. This is problematic in news as we can bring to mind catastrophic events more - plane travel seems riskier. Being killed by a shark seems more likely than being killed by falling airplane parts, although the latter is 30x as likely
Representativeness heuristic
Probability of Event A is evaluated by how much A resembles B; if A is like B then A should be categorized in the same way as B. This is why we sometimes think probability of a compound event (that matches things we have seen in real life) greater than probability of element. Adding details increases representativeness
Pros: similarity often predicts group membership
Cons: ignores prior probabilities, similarities can be illusory, stereotyping.
