Debiasing Decisions

One continuing line of research involves investigating whether biases can be reduced by encouraging subjects to put more effort into forming judgments. Asking students to "show their work," for example, has been shown to slightly increase the chances of obtaining a correct answer (it is more helpful for pinpointing where lack of understanding may occur). In general, however, the limited success of such techniques suggests that most biases are not very sensitive to the amount of effort one applies.

Encouraging people to take an outsider's perspective has been shown to somewhat reduce the tendency for overconfidence to bias estimates ("What do you think someone not directly involved would say?"). The idea is to reduce personal biases by removing oneself mentally from the specific situation. Some studies show that the technique can improve estimates of the time it would take to complete a task and the odds of success [11, 12].

Increasing accountability for decisions has been shown to lead to better choices [13]. Likewise, enhancing accountability for opinions that people express can help in some circumstances. For example, it has been suggested that, when obtaining critical information from someone, it may be useful to take notes (or to appear to take notes). If people believe you may quote them to others, they may be more careful in what they say. Similarly, to support project-selection decisions, it is useful to have project proponents document precisely why they believe their proposed projects should be conducted. Going on record encourages managers to be more careful in their logic, and the fear of being proved wrong helps counter over-optimism.

Training (in biases) has been shown to help people address some biases. However, the effect is generally short lived and does not produce an overwhelming improvement in performance. One problem is that it is often hard to get people to appreciate that bias is something that affects them personally, not just others. Thus, in situations where time permits, it helps to demonstrate biases. For example, if you are obtaining judgments from a group of individuals and are concerned about overconfidence bias, don't just tell them about the 2/50 rule (described above). Instead, run them through an exercise that demonstrates that the rule applies to them.

Not surprisingly, improving cause-effect understanding of the relevant systems and processes has been shown to improve the quality of estimates and decisions. For example, studies show that when people are encouraged to look for common principles underlying seemingly unrelated tasks they are able to better create solutions for different tasks that rely on the same underlying principles [14].

There is evidence that groups reach better decisions when alternatives are evaluated simultaneously as opposed to having each alternative evaluated sequentially and potentially rejected. The presumed explanation is that people initially react emotionally when considering an alternative; they think mostly about how it will affect them personally. If alternatives are evaluated simultaneously side-by-side, group members are less susceptible to this reaction [15].

Strategies for Reducing Biases

The usual strategy for reducing a bias is to address the mental processing error that is believed to produce the bias. For example, in one study, researchers assumed that "hindsight bias," the tendency to exaggerate the extent to which one could have anticipated a particular outcome, results from the difficulty people have in appreciating the limited information available at the time and the distinct inferences that could be made from that information. By providing evidence that argued against the actual outcome, they found that their subjects could be made more resistant to the bias [16]. Similarly, it has been hypothesized that people's tendency to over-claim credit for a group accomplishment is due in part to the tendency to be more aware of one's own efforts. Researchers showed that when people are asked to estimate not only their own contributions but also those of others, they attribute less credit to themselves. [17].

Figure 4 (derived from Wilson and Brekke [18]) illustrates a view for how judgmental biases are created and suggests a general strategy for reducing biases. According to the figure, the first step is to create awareness of the flawed mental processes involved. The subject must be motivated to correct the bias, and understand the direction and magnitude of the errors produced. Finally, the bias must be removed or countered. The technique used to mitigate the bias of concern is often the application of a countering bias, for example, countering overconfidence by encouraging subjects to describe (and therefore anchor on) extreme possibilities. Many of the recommendations provided earlier in this section under "Advise" are based on this logic.

Decision Aids

Hundreds of decision aids have been developed and recommended to reduce the distortions in decisions caused by errors and biases. I developed the list of sample aids below in the context of a chapter for a book on aids for environmental decisions [19]. Links to definitions are provided for terms that tend to be used in project portfolio management. As indicated, there are at least 5 categories of aids: (1) checklists for promoting a quality decision process, (2) thinking aids intended mainly to improve perspective or create insights, (3) models and optimization methods for recommending choices, (4) aids for promoting group consensus, and (5) voting methods. As an example of the first category, Figure 5 is a checklist aid for scoring the decision-making process relative to the components of a quality decision-making process.

Sample Decision Aids
Checklists Thinking aids de Bono's Six Hats SWOT analysis Theory of constraints Models & analysis Comparison methods PMI (listing pluses, minuses and "interesting") Paired comparison Weighted comparison methods Analytic hierarchy process ELECTRE Force field analysis Grid analysis Goal programming Kepner-Tregoe method Multi-attribute utility analysis PROMETHEE SMART (simple multi-attribute rating technique) Cost-benefit analysis Decision analysis Social choice theory Expert systems	Optimization methods Linear programming Integer programming Quadratic programming Dynamic programming Probabilistic optimization methods Decision trees Stochastic programming Fuzzy logic Game theory methods Zero-sum games N-person games Cooperative games Non-cooperative games Group facilitation methods Delphi method Nominal group technique Voting methods Approval voting Borda count Copeland's method Plurality voting Plurality with run-off elections

Figure 5: Checklist diagram for evaluating deficiencies in the decision-making process [12].

Approaches to making good decisions differ greatly in terms of the amount of time and effort required. Figure 6 illustrates that the best approach to decision making depends on the significance of the decision, the challenges involved, and the time available for analysis. It is useful to develop a quiver of decision making strategies, and to select the approach that makes most sense for the given circumstance.

Considerations for selecting a decision aid

Figure 6: The appropriate decision aid and decision-making approach depends on the nature of the decision.

Models and analysis, in my opinion, represent the most effective way to address errors and biases in decision making. Essentially, the concept is to replace flawed intuitive reasoning with a formal, deliberative process. Much evidence has accumulated indicating the effectiveness of models and analysis. For example, in situations where past data is available on the inputs and results of a decision making process, simple models can be created using regression analysis. Such models are being used to help graduate schools to decide which students to admit, clinical psychologists to diagnose neuroses and psychoses, and credit card companies to decide whether to accept loan applications. For a very wide range of applications, researchers have found that such models produce better and more reliable decisions than those made by people, including the experts who made the original decisions from which the models were derived [21].

Another argument for using models comes from research that shows that experts appear to be less subject to biases when addressing issues that are within their areas of expertise. Models break the problem of evaluating alternatives down into individual pieces such that different experts with specialized knowledge can be selected to focus on each piece. Thus, models dissect a complex problem in a way that makes the required judgments less likely to be biased.

A Strategy for Decision Making

Approaches to making good decisions differ greatly in terms of the amount of time and effort required. Intuitive decisions can be fast, automatic and effortless, while analysis is slower and requires considerably more effort. Figure 7 illustrates that the appropriate approach to decision making depends on the significance of the decision, the challenges involved, and the time available for analysis. It is useful to develop a quiver of decision making strategies, and to select the approach that makes most sense for the given circumstance.

The best protection from bias comes from training, using formal techniques for obtaining important judgments, utilizing well-founded decision aids, and instituting rigorous decision processes that document the judgments and assumptions upon which choices are based. The subsequent parts of this paper describe such a process specific to decision making for project portfolio management. As stated by Ken Keyes, "To be successful in the business world we need to check our bright ideas against the territory. Our enthusiasm must be restrained long enough for us to analyze our ideas critically" [22].

Debiasing

Strategies for Reducing Biases

Decision Aids

A Strategy for Decision Making

References for Part 1