- Количество слайдов: 17
Judgment and Decision Making in Information Systems Group Judgment and Decision Making Yuval Shahar M. D. , Ph. D.
Group Decision Making • Essentially, an extension of individual decision making with all faults and biases • Unique features: Social influence, mutual discussion, a group decision-making procedure • Process and mode of decision making (e. g. , interactive discussion, majority vote) are crucial • Advantages: Potential brainstorming, multiple types of expertise and points of view, facilitation of problem solving and creativity • Disadvantages: Social loafing, conformity, polarization of previous opinions, “groupthink” phenomenon might lead to unreasonable decisions
Are N+1 Heads Better than One? • Groups are often better in solving problems, especially involving unfamiliar domains – Example: The NASA moon landing survival experiment (ranking of 15 items by degree of importance): The group error is much smaller than the average of the personal errors • Review studies show that improvement occurs in several different areas: – Quantitative judgment (up to 23 to 32% accuracy increase) – Brain teasers and logic problems (better then average) – General knowledge questions (better than average member) – Creativity and problem solving (groups can pool resources) • However, it is not clear that the group interaction is useful: – The best individual usually outperforms the group as a whole in all types of judgments [Hastie, 1986] and their accuracy is even reduced! – brainstorming is more effective when ideas are generated independently and later combined, than when conducted in a group session [Hill, 1982] (might include a “social loafing” element as well)
Conformity [Asch, 1951, 1955, 1956] • Is the test line equal in length to A, B, or C? • Small (3 of 4) majorities are sufficient to elicit substantial conformity with a wrong result • Any opposition can have a major effect; even a single dissenting individual can nullify the effect! Test line A B C
Asch Results: The Pressure to Conform Condition Error rate Subject is alone 1% With 1 person who says “A” 3% With 2 persons who say “A” 13% With 3 persons who say “A” 33% With 6 persons who say “A” and 1 person who says “B” 6%
Groupthink [Janice, 1982] • “A deterioration of mental efficiency, reality testing and moral judgment that results from ingroup pressures” • Famous examples: Bay of Pigs invasion; Challenger space shuttle launch; Israel before the October 1973 war; etc. • Occurs when groups are cohesive and insulated from the outside, while inside pressures for group loyalty and conformity lead to muddled thinking
Groupthink Symptoms [Janice 1982] • Illusion of invulnerability leads to unjustified optimism and risk taking • Collective efforts to discount warnings • Unquestioned belief in group’s morality • Stereotyped views of adversaries (too evil to negotiate with, too stupid to be a threat) • Pressure directing at any dissenting group member • A shared illusion of unanimity • Self censorship of deviations from group consensus • Self-appointed “mind guards” protect group from information that might challenge the complacency
Group Polarization [Moskovici & Zavaloni, 1969] • The “risky shift” phenomenon [Stoner, 1961]: Individuals tend to take more risks after a group discussion • Can be also a “cautious shift, ” and in any judgment • Possible explanations: – The extreme majority alternative gets more discussion time – Responsibility is shared among individual members – Extreme individuals become more extreme when they discover that their opinion is not as extreme as viewed – The extreme alternative is valued higher due to group effect – Groups are risk-neutral while members are risk-averse
Social Loafing • People do not work as hard in groups as they work alone (e. g. , rope tugging, even when blindfolded) • Responsibility for the final outcome is diffused among members of the group • Multiple examples from reality and research: – – Reporting a crime viewed by many people in New York Assisting a child lost on the highway in Italy Assisting a stranger pick up coins in an elevator Intervening as a bystander to report a fire, or to assist a stranger who cries for help in a nearby room • Presence of others reduces likelihood of intervention
Group Discussion Modes • Multiple facilitation techniques are potentially applicable – – Collective: averaging without interaction Consensus: Face to face discussion until agreement Dialectic: members discuss biasing factors Dictator: Discussion leads to selection of representative (“best”) member – Delphi: Judgments provided ahead of time and consensus is established iteratively in several rounds • The Collective method was the worst, but the Dictator method was up to three times more accurate in an experiment involving sales prediction [Davis, 1973]
Group Decision Making Rules • • • How does a group reach a decision? Multiple potential methods exist, such as a simple majority, total agreement, etc. A crucial difference exists between 1. choosing among two alternative 2. Choosing among three or more alternatives • The difference is in the ability 1. to reach non-paradoxical results, such as an undesirable effect of irrelevant options 2. To avoid potential manipulation by one of the decision makers
The Condorcet  Paradox • • 1/3 of the group prefers x>y>z 1/3 of the group prefers y>z>x 1/3 of the group prefers z>x>y Thus: – an absolute 2/3 majority prefers x to y – an absolute 2/3 majority prefers y to z – But: an absolute 2/3 majority prefers z to x • Inconsistent with a transitivity assumption (!!)
The Borda Scoring Rule • Each group member ranks all K alternatives, giving the best K points and he worst 1 point • Points are summed to determine the decision • However, the assumption of independence from irrelevant alternatives (IIA) is violated • For example, removing 3 inferior alternatives from a list of 5 alternatives can change the preference order for the two main alternatives!
Example of Borda IIA Violation • Assume 5 individuals and 5 alternatives a, b, c, x, y: 1. 2. 3. 4. 5. • • x>y>a>b>c y>a>c>b>x c>x>y>a>b x>y>b>c>a y>b>a>x>c Note: x>y Note: y>x Thus, score(y) = 21 > score(x) = 17; y>x by group But: removing a, b, c leads to: – score(x) = 8 > score(y) = 7; x>y by group!
The Importance of the Comparison Order • Assume 3 individuals and 3 alternatives x, y, z 1. x>y>z 2. y>z>x 3. z>x>y (or: x>z>y) • • • If we first compare x to z, z wins and then y wins; if we first compare x to y, x wins and then z win The comparison order (e. g. , as determined by a committee head in a parliament) is crucial! The third individual can change the eventual winner by manipulating their preferences so that x wins when compared to z and then wins over y!
The Arrow [1951, 1963] Impossibility Theorem • It is impossible to find a group decision rule for choosing the preferred alternative out of three or more alternatives, without potentially leading to a paradoxical result (i. e. , more than one chosen alternative) • Shows that at least one of four given assumptions, such as independence of irrelevant alternatives, cannot be maintained • Later results show that every reasonable election method for one candidate out of three or more can be manipulated by voting against one’s own preferences
Summary: Group Decision Making • Advantages: – Multiple views and types of expertise – Social facilitation due to directed open discussion – Brainstorming might lead to creative solutions • Disadvantages: – Often fails to equal the best individual solution – Conformity bias (e. g. , the Asch (1951 -1956) 3 lines experiments) – Groupthink (overconfidence, ignorance of facts) – Polarization effect (e. g. , “risky shift” phenomenon) – Social loafing (pool, elevator experiments) – In general: No optimal decision rule exists for more than 2 options • Much depends on how the discussion is managed (e. g. , order of speaking; use of a discussion facilitator; a consensus, dictator or a Delphi methodology)