Anchoring, availability and representativeness Lecture 5
Tversky & Kahneman: Judgment under Uncertainty: Heuristics and Biases Summary in three slides
Representativeness heuristic Usually employed when people are asked to judge the probability that an object belongs to a certain class Probabilities are evaluated by the degree to which A is representative of B (resembles B). When A is highly representative of B, the probability that A originates from B is judged to be high. On the other hand, if A is not similar to B, the probability that A originates from B is judged to be low.
Availability heuristic Biases due to the retrievability Lists of men and women (in some more famous women; in some more famous men) ... easier to recall names of famous people Biases due to the effectiveness of a search mechanism Word starts with an r or its third letter is an r Biases of imaginability Committees of size 2 and 8 Illusory correlation Associative connections are strengthened when the events frequently co-occur
Anchoring Anchoring can happen with completely trivial anchors Social security number influenced the subjects’ estimate of the number of doctors on Manhattan Throwing dice influenced the judges’ estimate of the proper prison sentence length in a given case Anchoring is more effective the less we know about the target But are real-life anchors this uninformative? Wages Buying prices Internet helps us find more correct reference prices
Hammond, Keeney, Raiffa: The Hidden Traps in Decision Making 22.4.2019
Key idea Lists biases relevant to business decisions Anchoring Status quo bias Sunk cost fallacy Confirmation bias Framing Biases in forecasting
Anchoring How do you estimate the sales of a product? Often too much anchoring on past sales Anchors in negotiations Salaries, sales prices, contract terms Setting the anchor is a powerful tool What can you do? Use different opinions, perspectives, and analysts Don’t let them anchor on your initial idea 22.4.2019
Status quo bias Sin of omission is not seen as bad What can you do? Killing someone vs. failing to save ”If it isn’t broken, don’t fix it” What can you do? Think, would you really choose this from a choice set? Jump into the future after ”rocking the boat” – which alternative is better? On the other hand, sometimes status quo is indeed the smart choice “If I had asked people what they wanted, they would have said faster horses.” –Henry Ford 22.4.2019
Sunk cost fallacy Suppose you invest 1 million in developing a new product, but the prototype is a failure. What should you do now? Give up, and abandon the product Invest more to make it a success What can you do? Solicit opinions from someone who is not involved in the history (no stake) Own up to failures, and make it a good thing in the culture Look at quality of decision making process, not just outcomes Is the narrative of headstrong perseverance always right? 22.4.2019
Confirmation bias We interpret ambigous evidence favorably to our opinion Wason’s 2,4,6 example More data may just mean ”more reasons I’m right” https://www.youtube.com/watch?v=9oSJdSL8YOE What can you do? Have someone play devil’s advocate (maybe anonymously, depending on culture) Think of your goals: are you hunting for truth or arguments? Avoid people who always agree with you 22.4.2019
Framing When things are framed in losses vs. gains, decisions change (examples in Lecture 6) Different reference points can result in different frames What can you do? Consider different frames – what are you expecting? What is a ”good result” in your problem, and why? Write out all the results of a choice, don’t leave them implicit 22.4.2019
Biases in forecasting We are widely overconfident of our forecasts On the other hand, we pad forecasts ”just to be safe” Our supporting data is biased by what we remember What can you do? List reasons why you might be wrong, and revise your estimate Have a culture of honesty about estimates Keep a database to estimate accuracy Use actual unbiased data whenever possible 22.4.2019
The Systems of Thinking Mostly from: Kahneman: Maps of Bounded Rationality: Psychology for Behavioral Economics, The American Economic Review, Vol. 93, No. 5 (Dec., 2003), pp. 1449-1475 22.4.2019
Heuristics and biases Biases Heuristics ”Rule of thumb” Mental shortcut Quick and dirty solution Instead of exhaustive search Linked often to intuition Not always bad! (Lecture 7) Eg. Chess masters Pattern of deviation in judgment Compared to some objective standard Statistically systematic Not predictable on individual level Clearly visible in larger samples Hot (emotional) Cold (cognitive)
Why do heuristics & biases matter? Concern about reliance on judgmental heuristics not attributable to motivational effects These heuristics are highly economical and often effective, but they sometimes lead to systematic and predictable errors Experienced researchers (when they think intuitively) are prone to many of the same biases as laymen Learn from lifelong experience Is it sufficient to be internally consistent in probability statements? What can we do about the biases?
Dealing with biases Kahneman is not very hopeful about people being able to correct their own biases: “I have been studying this for years and my intuitions are no better than they were.” But correcting others looks more promising: “But I'm fairly good at recognizing situations in which I, or somebody else, is likely to make a mistake - though I'm better when I think about other people than when I think about myself. My suggestion is that organizations are more likely than individuals to find this kind of thinking useful.”
Conscious Effortful Deliberate WITH self-awareness WITH control System 1 System 2 Can control Unconscious Effortless Automatic WITHOUT self-awareness WITHOUT control Conscious Effortful Deliberate WITH self-awareness WITH control
22.4.2019 Dual process theory Intuition (System 1) thoughts seem to come spontaneously to mind, without conscious search or computation, and without effort 2 + 2 = Driving a car in a “easy situation” Reasoning (System 2) is done deliberately and effortfully 38 x 252 = Turning left with car Most individual’s thoughts and actions are normally intuitive What about business decisions or decisions in organizations?
22.4.2019
22.4.2019
Consciousness Unconscious behavior Unconscious processing Automatic with lack of awareness, intention and control Inaccessible to conscious mind Unconscious processing Our perceptions, memory, and social judgments are all constructed by our unconscious mind Limited data, employing context, expectations and even desires Nature has created to make us survive System 1 product of thousands of years of evolution
Perception and System 1 Our social perception is not a direct result of what we experience, But constructed by our minds, employing context, prior knowledge, and desire Visual perception The ease with which mental contents come to mind Reference dependence Mind fills in the missing parts of the information
Reference dependence
Our mind fills in the missing information
Our mind fills in the missing information A girl learned to read at age of 4. What is her likely level of education? Social factors (no siblings) Environmental factors (learnt in an orphanage)
System 2 Lazy by nature Limited capacity Reasoning may be motivated Avoid difficult decisions (organ donations) Information overload (jam tasting and purchases) Limited capacity The Magic Number Seven Reasoning may be motivated Looks at conclusion that you want and limits information search on information that support the desired conclusion
22.4.2019
Cognitive systems Source: Kahneman 2003 Perception Intuition – system 1 Reasoning – System 2 Process Fast, parallel, automatic, effortless, associative, slow-learning, emotional Slow, serial, controlled, effortful, rule-governed, flexible, neutral Content Percepts, current stimulation, stimulus-bound Conceptual representations; Past, present, and future; Can be evoked by language Source: Kahneman 2003
The boundaries of thinking The judgments that people express, the actions they take and the mistakes they commit depend on The monitoring and corrective function of System 2 The impressions and tendencies generated by System 1 System 1 has been associated with poor performance, but it can also be powerful and accurate System 2 monitors judgments lightly, on some occasions it may however detect potential error and an effort will be made to correct it
Stanovich (2009). What Intelligence Tests Miss: The Psychology of Rational Thought. 22.4.2019
How can we improve thinking? Thinking is impaired with: Time pressure Concurrent involvement in a different cognitive task Thinking is influenced by: Emotions Mood Context Reasoning is positively correlated with: Intelligence Need for cognition (whether the individual enjoys thinking) Exposure to statistical thinking