Decision Making A basic introduction to the field for MIE448 Cognitive Ergonomics
References • Arkes, H. R., and Blumer, C. The Psychology of Sunk Cost, Organizational Behavior and Human Decision Processess35, (1985), 124-140. • Kahneman, D., and Tversky, A. Subjective Probability: A Judgement of Representativeness, Cognitive Psychology3, (1972), 430-454. • Tversky, A., and Kahneman, D. Availability: A Hueristic for Judging Frequency and Probability, Cognitive Psychology5, (1973), 207-232. • Tversky, A., and Kahneman, D. Rational Choice and the Framing of Decisions, Journal of Business59, 4, pt2 (1986), S251-S277. • Wickens, C. D. Engineering Psychology and Human Performance, Second ed. HarperCollins Publishers Inc., 1992). • Wickens, C. D., and Hollands, J. G. Engineering Psychology and Human Performance, Third ed. (Upper Saddle River, New Jersey: Prentice Hall, 2000).
What this lecture hopes you walk away with ... • An understanding of what the field of decision making is and who is interested in it • A familiarity with some of the better known decision making biases and heuristics • Potentially- an understanding of how people make judgments and decisions and to design appropriately
Who is interested in the study of judgement & decision making? • Economists, mathematics, business - normative models how people should make decisions • cognitive psychologists, marketing- descriptive models of how people do make decisions • Other fields: • medical diagnosis, fault diagnosis, weather forecasting, judicial procedures (weighing evidence), military, university admission procedures
Decision making tasks refers to which … • The decision maker must select one of a number of possibilities (sometimes many options). • The time to make the decision is relatively long as compared to manual control tasks. • The decision often must be made under uncertain conditions (stimulus information may of may not be reliable)
Human Limits • People are bad at making absolute judgments, (but are very good at making relative judgments) • Have a poor understanding of extreme values (< 0.05 and > 0.95) • Can not easily extrapolate growth functions
Three causes for conservatism in extrapolation • Exponential growth functions are more complex • conservatism • true life exponential growth are rare Therefore, when possible use a predictive display to assist, or a transformation to create a linear display.
In the beginning… • Gamblers were asking statisticians how much should they be willing to pay for different gambles? (Coin flip: heads win $1, tails win $0.) • Simple answer (normative rule): compute the expected value. (50%*1 + 50%*0 = bet up to 50 cents)
Basic Steps In Rational Decision Making 1. Identify all possible options (including doing nothing). 2. Quantify the value (or cost) of consequences which may arise if each course of action is adopted 3. Assess the likelihood of each consequence actually happening. 4. Integrate across all possibilities.
Unfortunately, this is too simple • Quantifying cost and value, and computing all options is cognitively expensive. • Isn’t descriptive. It doesn’t explain how people do make decisions- or allow for the prediction of how people make decisions. • Expected value doesn’t take into account things like “subjective utility”. (Lottery: odds of winning 1/13 million, jackpot = 50 million $, therefore bet the house …)
Too simple… continued • Decision making using an expected value model would mean there would be no reason to ever buy insurance. (instead- minimize maximum loss)
Bayes’ Rule The ultimate optimal statistical model for making decisions under uncertainty- starting point for all comparisons. People tend to use only the likelihood ratio in making decisions. P(H1 | D): Conditional probability that event H1 will occur, given that D has occurred.
Bayes’ Rule Posterior Odds = Prior Odds * Likelihood Ratio • D is a data statement (an observation, a piece of data) • H1 and H2 are two exclusive and exhaustive possibilities
Heuristics & Biases • The Salience Bias • The “as if” Heuristic • The Representativeness Heuristic • The Availability Heuristic • Overconfidence • Anchoring and Adjustment • The Confirmation Bias • The Framing Effect • Sunk Costs
The Salience Bias • We are hardwired to filter sensory information based on the saliency of the stimuli in the following order: • loud sounds • bright lights • motion • spatial position • This means people are biased toward salient information even if salient cues contain less information
The “As If” Heuristic • Information extracted from the stimuli in the world may be thought of as having different levels of importance, i.e. different weights • However, people tend to treat all different data cue “as if” they had equal weights • Decisions are made based more on the total number of data cues- without regard to their reliability or importance
The Representativeness Heuristic The subjective probability of an event is determined by the degree to which it • is similar in essential characteristics to its parent population • reflects the salient features of the process by which it is generated • sample size is irrelevant and prior probabilities are ignored (cf. Bayes’ Rule)
The Representativeness Heuristic All families of six children in a city were surveyed. In 72 families the exact order of births was GBGBBG. What is your estimate of the number of families surveyed in which the exact order of births was BGBBBB?
The Representativeness Heuristic • Both birth sequences are equally likely • However, the two sequences do not appear equally representative. (5 boys and 1 girl do not reflect the proportion of boys and girls in the population.) • Median response was 30 in Kahneman and Tversky’s experiment.
The Availability Heuristic The probability of events is evaluated by the ease with which relevant instances come to mind. In general, frequent events are easier to recall or imagine than infrequent ones.
The Availability Heuristic Consider the letter K. Is K more likely to appear in The first position of a word? The third position of a word?
The Availability Heuristic It is a lot easier to think of words which start with the letter K than of words where K is in the third position. However, a typical selection of text contains twice as many words in which K is in the third position than words which start with K.
Overconfidence • People (novices & experts) are in general, much more confident about their decisions than is reasonable given the environment in which they are making their decisions. • Potential to close off the search for answers before all available evidence can be collected because of overconfidence.
Anchoring and Adjustment • Under time pressure, estimate • 8*7*6*5*4*3*2*1 • 1*2*3*4*5*6*7*8 • 2,250 vs 512 (40,320) • Strategy to make a decision and stick with it- use future information only to make slight adjustments to original judgment
Confirmation Bias • changing hypothesis requires greater cognitive effort than maintaining the same hypothesis • harder to deal with negative information than positive information • self-fulfilling prophecy (teacher influencing the student, experimenter influence data)
The Framing Effect (Survival Frame) You are a patient with lung cancer. Which of the following two option would you prefer? Surgery: Of 100 people having surgery 90 live through the post-operative period, 68 are alive at the end of the first year and 34 are alive at the end of five years. Radiation Therapy: Of 100 people having radiation therapy all live through the treatment, 77 are alive at the end of one year and 22 are alive at the end of five years.
The Framing Effect (Mortality Frame) You are a patient with lung cancer. Which of the following two option would you prefer? Surgery: Of 100 people having surgery 10 die during surgery or the post-operative period, 32 die by the end of the first year and 66 die by the end of five years. Radiation Therapy: Of 100 people having radiation therapy none die during treatment, 23 die by the end of one year and 78 die by the end of five years.
The Framing Effect • Results: respondents who favored radiation therapy rose from 18% to 44%- no difference when subjects were patients or physicians. • How a problem is worded (or framed), either in terms of cost or value effects how people make decisions. • Difference between a perceived loss or perceived gain
The Sunk Cost Effect As the president of an airline company, you have invested 10 billion dollars of the company’s money into a research project. The purpose was to build a plane that would not be detected by conventional radar, in other words, a radar-blank plane. When the project is 90% completed, another firm begins marketing a plane that cannot be detected by radar. Also, it is apparent that their plane is much faster and far more economical that the plane your company is building. The question is: should you invest the last 10% of the research funds to finish your radar-blank plane?
The Sunk Cost Effect As the president of an airline company, you have received a suggestion from one of your employees. The suggestion is to use the last 1 billion dollars of your research funds to develop a plane that would not be detected by conventional radar, in other words, a radar-blank plane. However, another firm has just begun marketing a plane that cannot be detected by radar. Also, it is apparent that their plane is much faster and far more economical that the plane your company could build. The question is: should you invest the last billion dollars of your research funds to build the radar-blank plane proposed by your employee?
The Sunk Cost Effect As the owner of a printing company, you must choose whether to modernize your operation by spending $200,000 on a new printing press or on a fleet of new delivery trucks. You chose to buy the trucks, which can deliver your products twice as fast as your old trucks at about the same cost as your old trucks. One week after your purchase of the new trucks, one of your competitors goes bankrupt. To get some cash in a hurry, he offers to sell you his computerized printing press for $10,000. This press works 50% faster than your old press at about one-half the cost. You know that you will not be able to sell your old press to raise this money since it was built specifically for your needs and cannot be modified. However, you do have $10,000 in savings. The question is should you buy the computerized press from you bankrupt competitor?
The Sunk Cost Effect As the owner of a printing company, you must choose whether to modernize your operation by spending $200,000 on a new printing press or on a fleet of new delivery trucks. You chose to buy the press, which works twice as fast as your old press at about the same cost as the old press. One week after your purchase of the new press, one of your competitors goes bankrupt. To get some cash in a hurry, he offers to sell you his computerized printing press for $10,000. This press works 50% faster than your new press at about one-half the cost. You know that you will not be able to sell your new press to raise this money since it was built specifically for your needs and cannot be modified. However, you do have $10,000 in savings. The question is should you buy the computerized press from you bankrupt competitor?
The Sunk Cost Effect • Sunk costs are irrelevant to current decisions- instead, only incremental costs should influence future decisions. • Sunk costs have already been paid- you can’t get that cost back. • Strategy used by US nuclear energy program.