An Exploratory Study of Choice Rules Favored for High-Stakes Decisions

24
JOURNAL OF CONSUMER PSYCHOLOGY, 4(4), 305 - 328 Copyright O 1995, Lawrence Erlbaum Associates, Inc. An Exploratory Study of Choice Rules Favored for High-Stakes Decisions Barbara E. Kahn Department of Marketing University of Pennsylvania Jonathan Baron Department of Psychology University of Pennsylvania As information technology becomes more sophisticated, consumers will be able to access more information to help them make difficult high-stakes choices, such as medical and financial investments or career decision making. The purpose of this article is to examine how consumers think such information should be used in making decisions for which there are high stakes. Results, based on five exploratory studies, indicate that subjects do not spontaneously favor the use of compensatory decision procedures, such as multiattribute utility theory (MAUT). Explanation and structured pedagogical procedures significantly in- crease the subjects' endorsement of decision rules over no decision rules, but they do not increase the endorsement of MAUT. Further, subjects believe that they would be more likely to use compensatory models when they have more options and more information about the options, more time, less certainty about their goals, and more accountability. Paradoxically, although subjects generally do not want to use compensatory rules themselves, they are more likely to want their agents (e.g., physicians or financial or career advisors) to use these rules in making decisions. New interactive videos and software have created an explosion in the amount of information available to a consumer. For example, in the health care area interactive videos (e.g., those developed by the Foundation for Informed Med- ical Decision Making, Hanover, NH) allow patients to educate themselves Requests for reprints should be sent to Barbara E. Kahn, Department of Marketing, The Wharton School, University of Pennsylvania, Philadelphia, PA 19104.

Transcript of An Exploratory Study of Choice Rules Favored for High-Stakes Decisions

JOURNAL OF CONSUMER PSYCHOLOGY, 4(4), 305 - 328 Copyright O 1995, Lawrence Erlbaum Associates, Inc.

An Exploratory Study of Choice Rules Favored for High-Stakes

Decisions

Barbara E. Kahn Department of Marketing University of Pennsylvania

Jonathan Baron Department of Psychology University of Pennsylvania

As information technology becomes more sophisticated, consumers will be able to access more information to help them make difficult high-stakes choices, such as medical and financial investments or career decision making. The purpose of this article is to examine how consumers think such information should be used in making decisions for which there are high stakes. Results, based on five exploratory studies, indicate that subjects do not spontaneously favor the use of compensatory decision procedures, such as multiattribute utility theory (MAUT). Explanation and structured pedagogical procedures significantly in- crease the subjects' endorsement of decision rules over no decision rules, but they do not increase the endorsement of MAUT. Further, subjects believe that they would be more likely to use compensatory models when they have more options and more information about the options, more time, less certainty about their goals, and more accountability. Paradoxically, although subjects generally do not want to use compensatory rules themselves, they are more likely to want their agents (e.g., physicians or financial or career advisors) to use these rules in making decisions.

New interactive videos and software have created an explosion in the amount of information available to a consumer. For example, in the health care area interactive videos (e.g., those developed by the Foundation for Informed Med- ical Decision Making, Hanover, NH) allow patients to educate themselves

Requests for reprints should be sent to Barbara E. Kahn, Department of Marketing, The Wharton School, University of Pennsylvania, Philadelphia, PA 19104.

306 KAHN AND BARON

about the details of their medical conditions on the basis of personalized information that is entered in for each patient. In financial investments, there are new software packages that provide consumers with information on invest- ment opportunities, tax planning, and estate management.

Along with the availability of information comes an obligation to make decisions that once rested only with experts. These are decisions that have high stakes and that consumers may find emotionally difficult to make. For exam- ple, Dr. Jimmie Holland, chief of psychiatry at Memorial Sloan-Kettering Cancer Center in New York, noted,

We have created a climate which says that the good patient is the most consumer- oriented, who gets the most information and makes the most informed decision. . . . [hospital] staff is frequently consulted by distraught patients who just can't make up their minds. (Rosenthal, 1994).

This trend toward including the consumer more in the process of important, life-altering decisions brings up interesting research issues. Although there has been a great deal of descriptive work on how consumers make decisions in low-involvement situations, less is known about how consumers think they should be making high-stake decisions. In the studies that follow, we focus on consumers' attitudes toward compensatory decision rules based on multiat- tribute utility theory (MAUT) or expected utility theory (Keeney & Raiffa, 1976). Because such rules are featured in many decision aids, understanding consumers' attitudes toward these rules may help predict their reactions to various interactive normative tools for aiding high-stake decisions.

Compensatory models allow trade-offs among attributes such that a good value on one attribute can make up for bad values on others. These prescrip- tive models are applicable to general marketing and consumer decisions (Hauser & Urban, 1977, 1979), to medical decisions (Doubilet & McNeil, 1985; Pauker & McNeil, 1981), to investment decisions (D. M. Kleinmuntz, 1985), and to other tasks when a decision maker needs to choose a preferred alterna- tive (see Keeney & Raiffa, for a review).

Empirical studies have shown that when people are confronted with simple decisions MAUT may provide good approximations of behavior (Payne, 1976, 1982; Tversky, 1969). However, when the decision-making task increases in complexity,' systematic differences are observed between the compensatory model and actual behavior (Fischer, 1979). Many naive decision makers use either simple decision heuristics or no decision rules at all (Fischer, 1979;

l~omplexity or difficulty in the decision task has been defined as increases in (a) the number of alternatives or attributes, (b) the difficulty of processing of the material, (c) the uncertainty about the value of some of the attributes, and (d) the comparability of some of the alternatives in the set (see Bettman, Johnson & Payne, 1991).

CHOICE RULES FOR HIGH-STAKES DECISIONS 307

Payne, Bettman, &Johnson, 1993; Schoemaker, 1982). The problem with these simpler rules (e.g., a lexicographic rule) is that alternatives may be eliminated (or chosen) based on the value of one attribute without considering the values of other potentially compensating attributes.

We examined three research questions about consumers' attitudes toward compensatory decision rules for high-stakes decisions. First, we asked whether consumers think they would use compensatory decision rules spon- taneously. Based on past research on low-involvement decision-making, we hypothesized that even in life-altering decisions consumers may not spon- taneously advocate compensatory rules (Study la). We also explored whether consumers will respond positively to these tools when they are ex- plained (Study lb). Next, we studied the conditions under which consumers think that they would be more likely to use compensatory rules. Are these the same conditions that, according to empirical evidence, influence consum- ers' actual use of compensatory rules (Study 2)? Finally, if consumers do not themselves favor compensatory rules spontaneously, do they expect a ratio- nal agent to use these rules when making decisions on their behalf? We ad- dressed this question both in a hypothetical situation within subjects (Studies 3a and 3b) and in a realistic decision-making context between subjects (Study 3c).

STUDY 1: CONSUMERS' SPONTANEOUS USE OF COMPENSATORY DECISION RULES

Past research (mostly on low-involvement decisions) has shown that naive decision makers shy away from compensatory or trade-off rules for many reasons (see Payne et a]., 1993). First, compensatory decision rules are more difficult to implement. Second, compensatory rules force trade-offs, whereas noncompensatory rules avoid the conflict. Thus, decision makers may avoid compensatory decision rules not only because they require difficult cognitive effort but also because they require explicit resolution of difficult value trade- offs (Hogarth, 1987).

Many researchers (e.g., Payne et al., 1993; Tversky & Kahneman, 1986) posited that it is sensible for decision makers to use different decision strategies in different situations as an adaptive response. Decision makers have different goals in different situations; sometimes accuracy is the desired objective, and other times it is the desire to conserve cognitive energy. Empirical research with managers and student subjects (McAllister, Mitchell, & Beach, 1979) indicates that decision strategies will be more analytic and will result in a greater invest- ment of time and effort when decisions are significant and irreversible and when the decision maker is responsible for his or her actions. In high-stakes decisions, the desire for accuracy may be paramount. Hence, consumers

308 KAHN AND BARON

should be most likely to use (or to want to use) prescriptive tools, such as compensatory models, in such tasks.

In the first study (Study la), we presented an open-ended questionnaire to see how subjects would spontaneously handle information that is readily avail- able to the interested health-care consumer who needs to make medical deci- sions. We used four items concerning cancer treatment and one item concerning mammography. Decisions about the treatments involve trade-offs between quality-of-life, survival, and other factors. Decisions about mammo- grams were presented as mainly involving a trade-off between risk of cancer caused by the X rays and the chance of curing or preventing cancer as a result of early detection. To develop the cancer treatment stimuli, we consulted technical, patient-oriented materials such as Everyone's Guide to Cancer Ther- apy (Dollinger, Rosenbaum, & Cable, 1992) and those from the National Cancer Institute International Cancer Information Center.

Study la

Method. We presented treatment options for four types of cancer: blad- der, skin, ovarian, and prostate. The subjects were given a description of the cancer, the type of people most susceptible to it, the number of cases diagnosed in the United States in 1990, and the growth rate of the disease. In addition, they were told about the most widely used methods to treat each kind of cancer. Although the treatment options were disease specific, most of them include some variations on surgery, radiation, chemotherapy, and biological therapy. Subjects then were asked the following questions:

1. If you were making the choice among these methods, how would you think about the decision?

2. What factors would you consider? 3. What additional information would you want? 4. How would you use this information to make a decision?

For the mammogram question, subjects were told:

A mammogram is an X-ray study of the breasts. Mammograms are useful in detecting breast cancer in early stages. When cancer is detected early, it is more easily cured. Mammograms are so useful for this purpose that they are often given as screening tests to women who have no symptoms. But X rays themselves can cause cancer. Although the dose of X rays is low, it is large enough to have an effect on the cancer rate.

The subjects were then asked the following questions:

CHOICE RULES FOR HIGH-STAKES DECISIONS 309

1. If you were deciding whether or how often a mammogram should be given in the absence of symptoms, how would you think about your decision?

2. What factors would you consider? 3. What additional information would you want? 4. How would you use the information you had to make a decision?

Subjects. The subjects were undergraduate students at a northeastern university. They were highly educated; on average, about 50% had course material related to formal decision making. Hence, these subjects were more likely than the general population to consider the use of compensatory models in decision-making tasks. Of the 37 subjects who participated in Study la, 43% were men, 48% had no formal training in decision making, 52% had some training in class, 73% knew someone personally who had cancer, and 44% said they felt that they took precautions against cancer. None of these individual differences significantly affected the results.

Results: Choice among cancer treatments. Two independent coders analyzed the open-ended data. The two coders' frequency counts correlated .86 across subjects. Differences were reconciled by discussion. The analyses are pooled across the four different cancer types; a preliminary analysis indicated no significant differences by cancer type.

On average, the subjects mentioned 4.1 factors that they thought should be considered; 80% of the subjects mentioned quality-of-life factors, 69% of the subjects mentioned survival factors, and 33% mentioned cost factors. A major- ity, 73%, believed the decision would depend on certain contingencies, such as age, whether they would want children, their emotional status at the time, or whether they had someone to care for them.

Although the subjects did fairly well at identifying the critical factors in the decision, they were less adept at determining how these factors should be combined or used to make a decision (see Table 1). Only 29% implicitly or explicitly mentioned probabilities of consequences, using terms such as risk, chance, odds, or success rates. The majority of subjects, 55%, did not mention any type of decision rule. Of the remainder, 13% said they would make the decision by relying on some norm (e.g., what the doctor said, what textbooks said, and what friends said). Only 32% of the whole sample explicitly men- tioned some type of decision rule. However, most of these discussions were vague. This is consistent with past research (Bettman & Park, 1980) that found that systematic use of structured (e.g., lexicographic, conjunctive, or linear compensatory) rules was rare in protocol data. The majority of those who alluded to a decision rule (18% of the whole sample) mentioned a noncompen- satory rule. Some of these rules were very simple, focusing on one factor, such as "choose the treatment with the highest success rate" or "choose the treat-

31 0 KAHN AND BARON

TABLE 1 Comparison of Decision Rules: Study 1

Study la : Study lb: Open- Structured

Decision Rule Endedd ~ s s i s t a n c e ~

Cancer treatment alternatives Would rely on a norm Did not mention a decision rule Noncompensatory rule Tradeoff rule Agree with normative modelC

Yes Yes, with reservations

Decision to get a mammogram Mimicked a rule Noncompensatory Assessed trade-off in probabilities Gave no decision rule Agree with normative modelC

Yes 66% Yes, with reservations 21%

%I = 37. bn = 113. ' n = 56 for these questions in Study Ib. *Proportions in Study 1 b that were significantly different from corresponding proportions in

Study l a at the p < .05 level.

ment that causes the least pain." Some subjects used a form of lexicographic rule.

Some of the decision rules (14% of the whole sample) could be categorized as those that recognize trade-offs. However, only a very few (less than 2% of the sample) mentioned rules involving trade-offs and probabilities that could be coded as a rudimentary utility analysis.

Results: Decision to get a mammogram. The mammogram question also involved trade-offs, but a reasonable approach from the patient's perspective (especially if the cost is covered by insurance) is to ask whether the overall probability of (death from) cancer is increased or reduced. Thus, the probabil- ity of early detection could be (roughly) subtracted from the probability of inducing cancer from the X rays. This strategy considers trade-offs by looking at overall probability. Of course, more explicit compensatory rules (including such factors as cost or false positives) could also be used.

When answering how they would make the decision to get a mammogram or not (men were told to assume they were women), 35% of the subjects did not mention a decision rule at all (see Table 1). The most common response (41%) was to mimic a rule, such as "every woman should have one per year"

CHOICE RULES FOR HIGH-STAKES DECISIONS 31 1

or "do what the family doctor says." Only a minority (24%) assessed the trade-offs in probabilities of getting cancer. The failure to combine probabili- ties as a basis for decision making replicates in college students what Baron, Granato, Spranca, and Teubal (1993) found in adolescents.

For important medical decisions, these results mirror other findings for lower involvement decisions. People do not spontaneously advocate the use of MAUT or the formal assessment of probabilities, even though research has shown that judgments are generally better if a formula or model is used (B. Kleinmuntz, 1990). Would consumers accept compensatory rules more readily if they were guided toward using them by asking them about rele- vant factors and how they should be taken into account? We addressed this question in Study lb, using the same two medical decisions: cancer treat- ments and whether or not to get a mammogram. Because we found no sig- nificant differences in the responses among cancers in Study la, we used only bladder cancer in this study because it was not gender specific and still evoked quality-of-life issues.

Study 1 b

In this questionnaire, we provided the same description of bladder cancer and treatment alternatives used in Study la and asked progressively more specific questions. We started by asking the subjects what the single most important factor was and whether they should make the decision on the basis of that factor alone. If they indicated that other factors should be con- sidered, we asked them what they were. Next, we provided a list of factors that may be considered, and asked them to rate the importance of each. Finally, we asked them again how they would use these factors in making their decision.

Half the subjects were given a description of a compensatory rule on the third page of the questionnaire. We asked the subjects if they agreed, agreed with some reservations, or disagreed with this rule. We provided the following paragraph:

Decisions of this sort involve many factors. The two main factors in choosing a therapy are (a) effectiveness of the treatment in reducing the chance of recurrence and (b) its side effects. One way to think about effectiveness is to ask what percentage of patients who have the therapy survive for 5 years and what percentage of those who do not have the therapy survive for 5 years. The difference is a good measure of how effective the therapy is. Each patient needs to decide how much immediate pain and suffering to tolerate in order to achieve a

31 2 KAHN AND BARON

given level of effectiveness. Other factors can be considered if the deci- sion between two methods is close. Because the decision involves at least two factors, it cannot be made on the basis of one factor alone.

Essentially the same procedure was followed for the mammogram decision.

Subjects. Of the 113 subjects in this study, 57 were undergraduates who responded to an ad for a psychology experiment. The remaining 56 subjects were undergraduates enrolled in an introductory marketing course. The latter group had some classroom instruction in decision making and participated to satisfy a course requirement. The average age of the subjects was 19.7 years old, 54% were women, 79% said they personally knew someone who had cancer, and 45% said they took precautions against cancer. There were no significant differences in these demographic measures between the two sam- ples, and they were not significantly related to the dependent variables. In addition, the two groups did not differ on any of the dependent variables.

Results: Choice among cancer treatments. When asked to name the most important factor in a decision about alternative treatments for bladder cancer, 79% of the subjects said survival, 20% said quality of life, and 3% mentioned cost. Thus, although more subjects mentioned a quality-of-life factor than a survival factor in Study la, survival factors were more important when subjects were asked to prioritize. When we asked the subjects if they were likely to make the decision only on the basis of the most important factor, 66% said no, but 34% said that they would. However, the average importance ratings indicated that most subjects believed that more than one factor was important in these decisions.

When asked how they would use these factors in making a decision, 47% of the subjects mentioned either an explicit noncompensatory rule or a vague rule that could be coded as such (see Table 1). The next largest group (34%) did not mention a decision rule at all. Only 19% mentioned some kind of compensa- tory rule.

The 56 subjects who participated as part of a course requirement were given a description of a compensatory rule and asked if they agreed with it. Due to potential demand effects, one may expect a large majority of students to agree with it. Although 61% of the subjects did agree with the compensatory rule, another 30% had some reservations either because the approach seemed too structured and did not allow for feelings at the time of the situation or because there was some problem with the expression of a 5-year survival rate. Only 9% said they disagreed with the theory.

Results: Decision to get a mammogram. When asked to name the most important factor they would consider in making a decision to get a mammo-

CHOICE RULES FOR HIGH-STAKES DECISIONS 313

gram, 53% indicated that they would need to know the likelihood of their getting cancer, 25% mentioned the likelihood of the test causing cancer, 15% mentioned the likelihood of the test detecting the cancer, and 7% mentioned contingencies, such as the woman's age. However, once again, the majority of subjects did not combine these factors into a compensatory decision rule (see Table 1). Only 26% described such a rule. Many subjects (45%) suggested some type of noncompensatory rule, such as considering only the likelihood of the test causing cancer or the women's age and, thus, the likelihood of getting cancer. Less than one third (29%) gave no decision rule, and a few subjects (3%) provided a rule of thumb, like "once a year."

We asked half the subjects to read a discussion of the compensatory argu- ment and to indicate whether they agreed with it. A majority (66%) agreed with it, but another 21% had reservations, such as "too structured an approach," "should consider other factors," or "don't know what the probabilities are." The remaining 13% did not agree with the rule.

Discussion

Although the data in Studies la and l b are not from a single study with random assignment of subjects, it is useful to compare the two sets of findings to explore the effects of providing structure to the decision process. As Table 1 shows, providing structure had some significant effects on subjects' propen- sity to mention the use of decision rules. In the decision among cancer treat- ment alternatives, significantly fewer subjects (34% in Study l b compared to 55% in Study la, p < .05) failed to mention a decision rule when structure was provided. The data indicate, however, that they switched to a noncompensa- tory rule (47% in Study l b compared to 18% in Study la, p < .05). There was no significant difference between the two studies in the percentages of subjects who used compensatory rules.

The results for the mammogram decision are similar. Significantly fewer subjects failed to mention a decision rule (or a rule of thumb) in the structured task than in the open-ended task (32% in Study l b compared to 76% in Study la, p < .05). However, as mentioned previously, there was no significant difference between the two studies in the percentage of subjects who used the compensatory rule. The subjects were simply more likely to use a noncompen- satory rule when structure was provided (45% in Study l b compared to 0% in Study la, p < .05).

Although a majority of the subjects agreed with the compensatory decision process when explicitly asked, we did not get 100% agreement. We believe that demand effects may have produced the high rate of agreement. This process of suggestion, however, may be similar to what patients could encounter in their physician's office. These data suggest that the majority of subjects were not disregarding the compensatory rule because they thought it was wrong.

314 KAHN AND BARON

STUDY 2: PERCEIVED INFLUENCES ON AN OPTIMAL DECISION-MAKING METHOD

In Study 2, we investigated what factors consumers think should affect the optimal method of decision making and compared these judgments to empiri- cal findings in psychology on the factors that, in fact, do affect decision processes. Consumers may be more open to decision aids and guidance in situations when they perceive a need for compensatory processes.

Again, we used the medical decision context and added a financial invest- ment decision context. We examined the effects of number of options, informa- tion about the options, time pressure, accountability (the need to justify the decision to others), certainty about the relative strength of one's goals, and whether the decision is made just for oneself or for someone else as well. Previous research has addressed the effects of three of these factors on deci- sion-making strategies: complexity of choice task (number of options and amount of information about the options), time pressure, and accountability.

Research shows that in complex environments (i.e., when the number of choice alternatives is greater than three and/or when the number of attributes per alternative is relatively high), subjects use noncompensatory strategies to simplify the choice set. For choice sets containing less than three alternatives, subjects are more likely to use compensatory decision strategies (Lussier & Olshavksy, 1979; Olshavksy, 1979; Onken, Hastie, & Revelle, 1985; Payne, 1976; Sundstrom, 1987; Tirnmermans & Vlek, 1992). Experts' use of decision strategies are similarly contingent on the complexity of the choice set. Biggs, Bedard, Gaber, and Linsmeier (1985) found that experienced bank loan offi- cers were likely to use noncompensatory decision strategies as the choice task became more complex (i.e., as the number of loan decisions and the number of cues describing each loan increased). For simple tasks, the experts were likely to use compensatory strategies.

Researchers also find that, as time pressure increases, decision makers try to simplify the task (Wright, 1974) and use noncompensatory strategies more often (Zakay, 1985). In evaluating the effects of time pressure and training on the use of MAUT, Zakay and Wooler (1984) found that training results in more effective decision making (i.e., a greater use of MAUT) under no time pressure conditions. Under time pressure conditions, training did not improve decision making and subjects were likely to use noncompensatory strategies.

Finally, the effect of social context on the use of decision strategies has been studied extensively. Although there is evidence that a subject's accountability (the need to justify one's views to others) affects decision making (Tetlock, 1985), the direction of the effect is not consistent. Simonson (1989) found that accountability caused subjects to use choice strategies that are easy to justify to others (i.e., accountability encouraged subjects to use normatively inappro-

CHOICE RULES FOR HIGH-STAKES DECISIONS 31 5

priate decision strategies). Contrarily, Tetlock (1991) argued that accountabil- ity may make decision makers more cognitively flexible and self-critical as they try to determine the preferences of the people to whom they are accountable. This open mindedness would lead to better decisions and would make the use of compensatory strategies more likely. Thus, although accountability does make consumers likely to use decision strategies that are more justifiable, the strategies may not always be better (Payne et al., 1993).

Study 2

Method. In this study, we examined an important investment decision ($100,000) in addition to the same medical decisions in Study 1. Half the subjects considered the financial decision first, and the other half considered the medical decision first. This time we told the subjects the factors that were relevant to the decision (rather than ask them to provide these factors) and provided specific descriptions of decision rules.

For the medical decision, the factors suggested were the treatment's health risk, its side effects, its effect on reducing the chance of recurrence, and its cost. Then, we provided descriptions of three decision rules (the rule names in parentheses were not provided):

1. Ask which factor is the most important. Then, pick the treatment that is best on that factor. (One-factor rule.)

2. Ask which factor is the most important. Next, pick the treatments that are relatively good on that factor. Then, look at the next most important factor and eliminate the treatments that do poorly on it. Continue going through the factors until only one treatment is left. (Lexicographic rule.)

3. Consider all the factors together. Give each treatment a score on each factor. Multiply the score on the most important factors by a large number, and then multiply the score on the less important factors by a small number. The multiplier should reflect the importance of the factors. Then, add the results together and pick the treatment with the highest score on all the factors combined. (Compensatory rule.)

For the financial decision, the questionnaire read:

Suppose that you inherit $100,000 and that you do not plan to spend much of it right away. You need to invest it. You call your broker, who makes a few suggestions. The suggestions differ in the amount of risk and in the expected return over several years. In general, riskier investments yield greater return in the long run, but they fluctuate considerably, and it is possible even to lose money. To avoid excessive commission pay- ments, you have decided to put the money into just one investment.

31 6 KAHN AND BARON

Here are two methods for deciding which investment to get: A. Decide whether it is more important to you to minimize risk or to maximize return. If it is more important to minimize risk, then choose the investment with the least risk. If it is better to maximize return, then get the investment with the highest expected return. The idea here is that you make the decision in terms of the most important attribute. B. Consider both attributes. Make the decision as if you gave each investment a score on each attribute, multiply the score by the relative importance of the attribute, add the results across attributes for each investment, and then pick the investment with the highest sum. The idea here is that you consider both attributes, so that a large difference in one of them can outweigh a small difference in the other.

In both the medical and financial scenarios, subjects were asked how the value of the compensatory strategy should be affected by various factors, for example:

Consider the number of possible investments. You might imagine that you have a choice of two investments in 1 case or 10 in another case. Check one of the following, and assume that these two methods of decision making, A and B, are the only ones you are considering. (Check one.) - Method B would be a better method to use with 2 choices than

with 10. The number of choices should not affect the method. Method B would be a better method to use with 10 choices than with 2.

Each subject was asked about the following issues for various factors of interest: number of investments (e.g., 2 vs. lo), time to make the decision (I hr vs. 1 day), amount of information that your broker has about the investments (a little vs. a great deal), your certainty about your own goals (very sure vs. easily persuaded to change), whether you have to explain your decision to someone else (nobody vs. the trustees of the estate, and you do not know their views), and whether the decision is just for you or for someone else as well (you vs. a cousin you have never met). Comparable items were used for the medical scenario. Seventy-two subjects completed the questionnaire. Forty-four of them participated to satisfy a marketing class requirement; 28 other students filled out the questionnaire for pay. Of the subject pool, 37 were men.

Results. Table 2 shows how subjects thought strategy should be affected by each factor. Subjects thought compensatory strategies were better to use when they had more options, more time, more information about the options,

CHOICE RULES FOR HIGH-STAKES DECISIONS 31 7

TABLE 2 Percentage of Subjects Saying That the Compensatory Strategy Would Be Better: Study 2

Factor

Investment Decision Medical Decision

More Less pa More Less pd

Options 50 18 ,002 5 1 11 ,000 Time 32 13 ,022 4 1 10 ,000 Information 79 7 ,000 69 7 .000 Certainty About One's Goals 28 5 1 .034 25 46 .050 Accountability 58 17 ,000 41 17 ,012 Others Affected (vs. self only) 38 17 .025 27 23 ns

Note. "Same" or "doesn't matter" responses are not reported. n = 72. aSignificance levels for two-tailed sign tests.

less certainty about their goals, and more accountability. Compensatory strategies were considered better when others were affected for investment decisions but not significantly for medical decisions. Thus, opinions about what should affect strategy choice are consistent with what is known to affect choice for the Time and Accountability factors. However, they differ for the number of options and for the availability of information about the options. The subjects' opinions seem to follow the normative principle that more diffi- cult, complex, or demanding decisions should require compensatory rules. In contrast, prior empirical work suggests that increased task complexity reduces the use of such rules (see Bettman, Johnson, & Payne, 1991, for a review).

These results were the same for both sexes, but the groups differed on the effect of accountability. Subjects from the marketing class thought compensa- tory strategies were better to use when there was more accountability, whereas the paid subjects showed no significant difference on this factor. Note that subjects in the marketing class were introduced to the concept of compensatory decision making and were familiar with normative considerations. The differ- ence in the findings for accountability mirrors somewhat prior empirical re- sults: Accountability drives subjects to justify their decision-making strategies without always knowing which strategies are better.

Subjects thought that the compensatory strategy would be better when they were less certain about goals. This result is surprising. Hogarth (1987) sug- gested that if consumers are not certain about their goals, they may have difficulty making the trade-offs that are necessary to use a compensatory strategy and may choose to use a noncompensatory strategy to avoid conflict. Perhaps our subjects equated being certain about goals with strongly favoring one goal over the other. In that case, a noncompensatory strategy would closely approximate a compensatory one.

2 ~ h e accountability findings are consistent with Tetlock's (1991) data but not with Simon- son's (1989) findings.

31 8 KAHN A N D BARON

Discussion

We found that subjects believe that more difficult, complex, or demanding decisions are more likely to require compensatory rules in these high-stakes decisions. This contrasts with empirical findings (albeit in low-involvement categories) showing that, as the task becomes more complex, subjects adopt easier-to-use noncompensatory strategies. Because subjects felt that compen- satory rules would be required in these contexts and because experience has shown that, in complex situations, compensatory strategies are more difficult to use, it is likely that subjects would desire some kind of decision-making aid when the task becomes complex. Therefore, we hypothesized that consumers may look to decision aids or rational agents to help them employ compensa- tory strategies that may be required for decisions when the stakes are high. This hypothesis is tested in the last set of studies.

STUDY 3: ATTITUDES TOWARD DECISION RULES- SELF VERSUS AGENT

The findings of Studies 1 and 2 raise the following questions: Do consumers regard noncompensatory rules as inherently better than compensatory rules? Do they regard compensatory rules as optimal except for the effort required to use them? Because decision aids such as interactive software or trained decision counselors can reduce the effort required to use compensatory rules, it is important to know not only if consumers would make decisions themselves but also how they would evaluate various decision rules when employed by deci- sion aids or by rational agents on their behalf.

Past research has not quite addressed this question. Adelbratt and Mont- gomery ( 1 980) asked subjects to evaluate alternative rules for making decisions about jobs and apartments. This research shows that subjects could distinguish among the rules in a meaningful way; many subjects evaluated compensatory rules as superior, but many did not. However, subjects assumed that they would actually be making the decisions by themselves (without aids or advice). Such consumer decisions are also unlike the high-stake decisions we consid- ered. In typical medical, financial, or career decisions, aids and agents are usually present to help with decision making, which may increase consumers' affinity for compensatory rules.

If consumers believe that trade-offs are difficult or painful to make and that compensatory rules are difficult to implement, they may prefer that a doctor or counselor make the decision for them given their values are respected. In addition, consumers may want a doctor or counselor to inform them of the optimal, rational decision so that information can be combined with their own emotion and intuition, using non-normative decision rules. In either case, we

CHOICE RULES FOR HIGH-STAKES DECISIONS 31 9

hypothesized that, whereas consumers are less likely to feel that they should use compensatory decision rules in making high-stakes decisions, they are more likely to believe that a doctor or other rational agent should be using the compensatory rules on their behalf.

In Study 3a and 3b, we investigated this issue. In Study 3a, we used a structured approach toward guiding subjects in decision making but allowed them to distinguish between the way they would make a decision and how they expected an agent (e.g., a physician) to make the decision. We focused again on decisions involving treatments for bladder cancer and whether or not to get a mammogram.

Study 3a

Method. We provided the same descriptions of bladder cancer, the treat- ment alternatives, and the three decision rules as in Studies 1 and 2. Then, we asked the subjects the following:

1. Which rule they feel most comfortable using when making the decision for themselves.

2. Which rule they would want the physician to use in making the decision. 3. Which rule was the hardest to use. 4. Whether they would prefer to make the decision themselves or have the

doctor make it for them.

We also presented the following scenario:

Suppose that you and your doctor used the best of these methods and arrived at a treatment decision. Suppose that this treatment was only very rarely used on people of your age and with the kind of cancer you have. Medical textbooks recommended some other treatment for cases like yours. What do you think that you and your doctor should do?

We took the same approach for the mammogram question.

Subjects. The subjects were 32 undergraduates who responded to an ad for a psychology experiment. The mean age of the subjects was 21.5; 55% were women, and 75% of them knew someone who had cancer. Only 4% had ever been actively involved in decisions regarding choice among cancer treatments. These factors were not significant in explaining any variance.

Results. In choosing among alternative treatments for cancer, we found significant differences in the type of rule the subjects would feel most.comfort-

320 KAHN AND BARON

able using and those they would want the physician to use (see Table 3). Whereas only 15% of the subjects indicated they would feel comfortable using a compensatory rule, 30% felt that was the rule the physician should use (p < .lo, one-tailed). Contrarily, whereas 31% of the subjects felt comfortable using the one-factor rule themselves, only 8% thought that was the rule the physician should use (p < .05). The difference may be attributed to the fact that a majority of the subjects (68%) thought the compensatory rule was the most difficult to use.

We found parallel results for the decision to get a mammogram. Whereas only 32% of the subjects indicated that they would feel comfortable using the compensatory rule, 61% felt that was the rule the physician should use (p < .05). In contrast, 68% of the subjects felt comfortable using the one-factor rule themselves, whereas only 39% thought that was the rule the physician should use (p < .05). Again, a partial explanation could be that 83% of the subjects identified the probability dominance (compensatory) rule as the harder one to use.

Of the subjects who gave different answers about which rule was best for the two decision makers (the patient or the physician), 11 favored the compensa- tory rule for the doctor in at least one of the two cases (cancer and mammo- gram), 3 favored the compensatory rule for the patient in at least one case, and 2 favored the compensatory rule for the doctor in one case but for the patient in the other. Using a sign test, the difference between 11 and 3 is significant (p = .03).

A majority of subjects (66% for both questions) said that they would prefer to make the decision themselves versus have the physician do it. We examined how this preference to make the decision oneself was related to beliefs about

TABLE 3 Use of Decision Rules, Patient Versus Physician: Study 3a

Decision Rule Patient Physician Hardest Should Should to Use

Alternatives among cancer treatments Use one-factor rule 31%* 8% 24% Use lexicographic rule 54% 63% 8% Use compensatory rule IS%** 30% 68% Make the decision 66% 34%

Mammogram decision Use one-factor rule 68%* 39% 17% Use probability trade-offs 32%* 61% 83% Make the decision 66% 34%

-- --

*Proportions for patient are significantly different from corresponding proportions for physicians at the p < .05 level.

**Proportion for patient is significantly different from corresponding proportion for physi- cian at the p < .10 level (one-tailed).

CHOICE RULES FOR HIGH-STAKES DECISIONS 321

using compensatory rules and found no significant relation. There was no significant correlation between the preference to make the decision oneself and the preferred decision rules used either by the patient or by the doctor.

When we examined how this preference to make the decision oneself was related to the resolution of conflict, we found significant relations. When the subject advocated that the doctor and the patient use the same rule, the subject was more likely to designate the patient as the decision maker. However, when the rules advocated were different, the subject was more likely to favor the doctor as the decision maker. These results were significant for the mammo- gram question, x2(1, N = 29) = 8.02, p < .01, and directionally supportive for the cancer alternatives (see Table 4).

The response to the scenario comparing whether to use a medical textbook or to rely on the individual doctor's opinion also indicated that, in the face of conflict in action recommendations, the subject would be more likely to rely on the physician. In this scenario, we explicitly presented a conflict between the textbook and the doctor. Here, for the cancer treatment question, 78% said they would follow the physician; for the mammogram question, 86% said they would follow the physician rather than the textbook. Subjects making the decision themselves would probably rely heavily on textbooks for outside consultation, as opposed to relying on their own doctor. Hence, this strong preference for a doctor over a textbook shows again that, given a conflict, the subjects would be more likely to rely on the doctor than on themselves.

Study 3b

Next, we attempted to replicate these results in a medical context and with financial investments-another high-stakes decision context. The descriptions of the financial investment and medical decisions scenarios and the alternative decision strategies for the medical and the financial scenario were the same as in Study 2. We asked subjects to indicate which strategy (or strategies) was best

TABLE 4 Relation Between Preference to Make Decision Oneself and Resolution of Conflict:

Study 3a

Designated Decision Maker Same D~fferent n

Alternatives among cancer treatments Self 48% 52% 2 1 Doctor 3 6% 64% 11

Mammogram decisiona Self Doctor

322 KAHN AND BARON

and which was worst. Then, we asked about which decision strategy they would want either their stockbroker or doctor to use if that agent were making a decision on their behalf. Seventy-two student subjects (60% from a marketing class and 40% recruited for pay) completed the questionnaire.

In response to the question concerning self-made decisions in the investment scenario, 51% of the subjects thought that the compensatory strategy was better, and 42% thought the noncompensatory strategy was better. For the medical decision, 42% ranked the compensatory strategy as the best one (of the three strategies given), and 26% ranked it as the worst. Responses to the second question showed that subjects thought compensatory strategies were better for experts to use in making decisions on the subjects' behalf. Thus, 68% of the subjects thought the compensatory strategy was better for the stockbroker to use, and 32% thought it was worse. This difference was larger than that for the self-made decisions (51% vs. 42%; p = .O1 by a Wilcoxon test with variables coded as 1 for favoring the compensatory strategy, - 1 for disfavoring it, and 0 for neutrality). The results were similar for the medical decisions: Forty-five percent ranked the compensatory strategy as best for the doctor making the decision on the subjects' behalf, and 9% ranked it worst. This difference was also larger than that for self-made decisions (42% and 26%, respectively; p = .005, Wilcoxon test).

Discussion

The paradox found in Studies 3a and 3b is that consumers are more likely to advocate the use of compensatory decision rules when a rational agent (e.g., a physician) is making a decision for them than when they make the decision for themselves. Perhaps the subjects believe that compensatory rules are better (more rational) decision-making tools, but they do not want to use these tools themselves. The fact that the results replicate across two samples and across two decision scenarios raises our level of confidence in the stability of the findings. However, the studies were both projective in nature (i.e., subjects did not actually make the decisions) and used within-subject designs, perhaps heightening the contrast perceived between the two situations. We attempted to address the concerns in Study 3c.

Study 3c

Next, we conducted a three-cell, between-subjects experiment in which we asked the undergraduate student participants to indicate how much they would be willing to pay for assistance in choosing their course schedules. The methods of assistance differed on the utilization of various different decision rules. For undergraduates, the decision of choosing course schedules has im- portant consequences and involves many trade-offs (e.g., instructor ratings,

CHOICE RULES FOR HIGH-STAKES DECISIONS 323

course ratings, workload, grading difficulty, and time of day). In the first cell, the computerprogram cell, we told students that the univer-

sity had been selected as a beta test site for a computer program called the Course Planner. The computer program would be designed to help students plan their course schedules. The computer program would use the student's preferences for different factors in creating the ideal course schedule. In the second cell, the faculty advisor cell, we instructed students similarly, except that we substituted a faculty advisor as the planner rather than a computer pro- gram. In the third cell, the control student's natural process celi, we told stu- dents that the university was conducting a survey to find out how students actually planned their course schedules. Based on this information, the univer- sity would design a computer program to provide customized aid to students that would mimic how they actually made decisions.

Subjects in all three cells indicated which of several factors were important to them in choosing class schedules. Thereafter, we told subjects in the first two cells that the university was experimenting with two ways of combining these factors to make a decision. Similar to our other experiments, we described a lexicographic rule and a compensatory rule. The order of the two rules was counterbalanced across students. No order effects were found to be significant in the analyses. In the third cell, we asked subjects which of the two rules more closely resembled the way they actually made decisions. Then, we asked sub- jects in all three cells how much they would be willing to pay for the two different versions of the decision aid. Finally, all subjects assigned importance weights to the various factors and indicated on a scale from 1 (not very dzficult) to 9 (very dzficult) how difficult they found the importance weights assignment task.

Subjects. Seventy-one undergraduate students participated in this experi- ment as part of a course requirement. They were told that this information would be passed on to the university. This premise was quite credible as the university frequently conducts surveys of students on such topics as computing facilities, degree requirements, and other aspects of the undergraduate curricu- lum. Subjects were told after the experiment that, in this case, the programs did not actually exist. Some subjects expressed disappointment at this debriefing!

Results. Although there were no significant differences among the three cells in choice of decision rules-49% chose the compensatory rule and 51% chose the lexicographic rule-there were significant differences in how much the students were willing to pay for the comparative services.

First, as Table 5 shows, students indicated that they were willing to pay on average the most for a faculty advisor's aid, the second most for a computer program that would mimic their own decision process; they would pay the least for a computer program that was designed as a normative decision aid. This

324 KAHN AND BARON

TABLE 5 Willingness to Pay for Different Decision Rules, Self Versus Agent:

Study 3b

Average Average Amount Paid Amount Paid

Condition

for Lexicographic for Compensatory Rule Rule A

A B A-B -

Faculty advisor $9.37 $12.38 + $3.21a Computer program $3.09 $ 6.76 + $3.67b Self $8.25 $ 5.25 - $3.00~2~

aDifference between F for advisor and self is significant at the .05 level. b~ifference between 6 for computer program and self is significant at the .05 level

result is reminiscent of the finding in Study 3a in which subjects indicated that they would rely on a doctor rather than a textbook when they conflicted. Study 3c's subjects indicated that they would rather rely on a faculty advisor than a computer program even if the decision process was identical.

Second, consistent with the results in past studies, students indicated that they favored the lexicographic rule when they were making their own deci- sions; but when decisions were being made by a rational agent, they favored the compensatory rule. Analysis of the importance weights assigned to the various factors showed that on average the second and third most important factors were, respectively, 80% and 57% as important as the most important factor. Subjects also indicated that assigning the importance weights to the second and third factor had a mean difficulty level of 4.4 and 4.5, respectively. Thus, assigning the weights did not appear to have been overly burdensome and did not seem to have influenced the subjects' choice of lexicographic rules for their own decision making.

GENERAL DISCUSSION

Decision analysts generally agree that explicit use of compensatory decision rules based on utility theory (MAUT and expected utility) approximates the normative rule for decision making. Past research, mostly on low-involvement decisions, has shown that many naive decision makers either use noncompen- satory rules or no decision rules at all. Little research has examined whether decision makers think that they should be using such rules, particularly in high-stakes, decision-making environments.

Our research shows that many consumers advocate the use of noncompen- satory rules even in high-stakes contexts where life-altering decisions are being made. Although subjects knew the factors that should be considered in making

CHOICE RULES FOR HIGH-STAKES DECISIONS 325

these types of decisions, they did not explain how to combine these factors and form a decision. In Study la, with no assistance in the decision-making pro- cess, more than half the subjects did not even articulate a decision rule. In Study Ib, when we provided some structure, subjects were significantly more likely to advocate a decision rule, but these rules were mostly noncompensa- tory. Providing structure did not significantly increase the advocacy of com- pensatory decision rules.

A noncompensatory rule, such as the lexicographic rule (which subjects said they were most likely to use), would eliminate alternatives if these alternatives did not have the highest importance weight on the most important attribute. In the medical decisions, our subjects seemed to indicate that the most impor- tant factors were those related to survival (e.g., the treatment's effect on reducing the recurrence of cancer). However, other research (Pauker & McNeil, 1981) has shown that, although attitudes toward morbidity are impor- tant, survival is not the patient's only consideration. Thus, the use of such noncompensatory rules may lead patients away from considering treatment options that may have greater holistic value.

In Study 2, we found that subjects believe that more difficult, complex, or demanding decisions are more likely to require compensatory rules. This coun- ters the behavior that subjects exhibit in empirical studies of decision making. For instance, studies examining many low-involvement decisions have shown that subjects, in fact, are more likely to use noncompensatory strategies. The finding that the subjects think that they should be using more compensatory strategies in a complex, high-stakes environment suggests that, in such envi- ronments, consumers may be more receptive to decision aids and guidance that help them use more normative strategies.

Studies 3a, 3b, and 3c produced a consistent but paradoxical result. We found that although subjects were unlikely to use a compensatory rule when making decisions themselves, they were significantly more likely to advocate the use of compensatory decision rules when an agent was to make the decision on their behalf.

Results of these studies also showed the stronger preference that subjects had for advisors over impersonal decision aids to help them with decision making. Subjects even seemed to favor the advisor's assistance over their own opinion. Study 3a indicated that, if a conflict existed between how the patient (or even a medical textbook) advocated a decision should be made and how the physician recommended the decision be made, subjects were inclined to follow the physician's recommendation. In Study 3c, subjects were willing to pay a higher price for a faculty advisor's assistance than for computer programs that were designed either to make normative decisions or to mimic their own decision-making process.

Although in most of our studies we did not experiment with actual consum- ers making medical or financial decisions, resistance to compensatory rules

326 KAHN AND BARON

may be more extreme when such decisions are really faced. Research on decision making under stressful conditions (Janis & Mann, 1977; Kahneman, 1973; Keinan, 1987; Mano, 1992) shows that people under stress tend to employ simpler decision rules, are more polarized in evaluation, and tend to focus more on attributes considered essential, excluding attributes of lesser importance. The findings were supported in Study 3c with a realistic choice setting. Subjects leaned toward the simpler rules for their own decision making and advocated that agents use the more complicated methods on their behalf.

Note that our subject pool is highly educated and not representative of the typical consumer. Actual consumers, perhaps less educated about normative decision theories, may be even more likely to move away from compensatory decision rules in making decisions for themselves and to rely on agents. Past empirical studies seem to support this conclusion. In one study on medical decision making, Strull, Lo, & Charles (1984) found that clinicians underesti- mated the patients' desire for information and discussion but overestimated the patients' desire to make the decisions themselves. In a study of financial decision making, Formisano, Olshavsky, and Tapp (1982) found that consum- ers choosing life insurance did not engage in any prepurchase information acquisition themselves but depended largely on recommendations from others. Our studies should be extended and the results examined with other actual consumers in high-stakes contexts. However, our results indicate that consum- ers will need a lot of assistance in making decisions as they become more and more involved in the decision-making process.

REFERENCES

Adelbratt, T., & Montgomery, E. (1980). Attractiveness of decision rules. Acta Psychologica, 45, 177-85.

Baron, J., Granato, L., Spranca, M., & Teubal, E. (1993). Decision making biases in children and early adolescents: Exploratory studies. Merrill-Palmer Quarterly, 39. 23-47.

Bettman, J. R., Johnson, E. J., & Payne, J. W. (1991). Consumer decision making. In T. S. Robertson & H. H. Kassarjian (Eds.), Handbook of consumer research (pp. 50-83). Englewood Cliffs, NJ: Prentice-Hall.

Bettman, J. R., & Park, C. W. (1980). Effects of prior knowledge and experience and phase of the choice process on consumer decision processes: A protocol analysis. Journal of Consumer Re- search, 7 , 234-248.

Biggs, S. F., Bedard, J. C., Gaber, B. G., & Linsmeier, T. J. (1985). The effects of task size and similarity on the decision behavior of bank loan officers. Management Science, 36, 887-899.

Dollinger, M., Rosenbaum, E., &Cable, G. (1992). Everyone's guide to cancer therapy: How cancer is diagnosed, treated, and managed day to day. New York: Universal Press Syndicate.

Doubilet, P., & McNeil, B. J. (1985). Clinical decision making. Medical Care, 23, 648-662. Fisher, G. W. (1979). Utility models for multiple objective decisions: Do they accurately represent

human preferences? Decision Sciences, 10, 45 1-479. Formisano, R., Olshavsky, R. W., & Tapp, S. (1982). Choice strategy in a difficult task environ-

ment. Journal of Consumer Research, 8, 474-479.

CHOICE RULES FOR HIGH-STAKES DECISIONS 327

Hauser, J., & Urban, G. (1977). A normative methodology of modeling consumer response to innovation. Operations Research, 25, 579-619.

Hauser, J., & Urban, G. (1979). Assessment of attribute importances and consumer utility func- tions: Von Neumann Morgenstern theory applied to consumer behavior. Journal of Consumer Research, 5 , 25 1-262.

Hogarth, R. M. (1987). Judgement and choice (2nd ed.). New York: Wiley. Janis, I. L., & Mann, L. (1977). Decision making: A psychological analysis of conflict, choice and

commitment. New York: Free Press. Kahneman, D. (1973). Attention and effort. Englewood Cliffs, NJ: Prentice-Hall. Keeney, R. L., & Raiffa, H. (1976). Decisions with multiple objectives: Preferences and value

tradeoff New York: Wiley. Keinan, G. (1987). Decision making under stress: Scanning of alternatives under controllable and

uncontrollable threats. Journal of Personality and Social Psychology, 52, 639-644. Kleinmuntz, B. (1990). Why we still use our heads instead of formulas: Toward an integrative

approach. Psychological Bulletin. 107, 296-310. Kleinmuntz, D. M. (1985). Cognitive heuristics and feedback in a dynamic decision environment.

Management Science, 31, 680-702. Lussier, D. A,, & Olshavsky, R. W. (1979). Task complexity and contingent processing in brand

choice. Journal of Consumer Research, 6, 154-165. Mano, H. (1992). Judgments under distress: Assessing the role of unpleasantness and arousal in

judgment formation. Organizational Behavior and Human Decision Processes, 52, 216-245. McAllister, D. W., Mitchell, T. R., & Beach, L. R. (1979). The contingency model for selection of

decision strategies: An empirical test of the effects of significance, accountability, and reversibil- ity. Organizational Behavior and Human Performance, 24, 228-244.

Olshavsky, R. W. (1979). Task complexity and contingent processing in decision making: A replication and extension. Organizational Behavior and Human Performance, 24, 300-316.

Onken, J., Hastie, R., & Revelle, W. (1985). Individual differences in the use of simplification strategies in a complex decision-making task. Journal of Experimental Psychology: Human Perception and Performance, 11, 14-27.

Pauker, S. G., & McNeil, B. J. (1981). Impact of patient preferences on the selection of therapy. Journal of Chronic Diseases, 34, 77-86.

Payne, J. W. (1976). Task complexity and contingent processing in decision making: An informa- tion search and protocol analysis. Organizational Behavior and Human Performance, 16, 366- 387.

Payne, J. W. (1982). Contingent decision behavior. Psychological Bulletin, 92, 382-402. Payne, J. W., Bettman, J. R., & Johnson, E. J. (1993). The adaptive decision maker. New York:

Cambridge University Press. Rosenthal, E. (1994). Hardest medical choices shift to patients. New York Times, January 27, Sec.

A, 1, 2. Schoemaker, P. H. J. (1982). The expect utility model: Its variants, purposes, evidence and limita-

tions. Journal of Economic Literature, 20. 529-563. Simonson, I. (1989). Choice based on reasons: The case of attraction and compromise effects.

Journal of Consumer Research, 16, 158- 174. Strull, W. M., Lo, B., & Charles, G. (1984). Do patients want to participate in medical decision?

Journal of American Medical Association, 252, 2990-2994. Sundstrom, G. A. (1987). Information search and decision making: The effects of information

displays. Acta Psychologica, 65, 165-179. Tetlock, P. E. (1985). Accountability: The neglected social context of judgement and choice.

Research in Organizational Behavior, 7, 297-332. Tetlock, P. E. (1991). An alternative metaphor in the study of judgment and choice: People as

politicians. Journal of Theory and Psychology, 1, 451-475.

328 KAHN AND BARON

Timmermans, D., & Vlek, C. (1992). Multi-attribute decision support and complexity: An evalua- tion and process analysis of aided versus unaided decision-making. Acta Psychologica, 80, 49-65.

Tversky, A. (1969). Intransitivity of preferences. Psychological Review, 76, 31-48. Tversky, A,, & Kahneman, D. (1986). Rational choice and the framing of decision. Journal of

Business, 59, S25 1 -S278. Wright, P. L. (1974). The harassed decision-maker: Time pressures, distractions, and the use of

evidence. Journal of Applied Psychology, 59, 555-56 1 . Zakay, D. (1985). Post-decisional confidence and conflict experienced in a choice process. Acra

Psychologica, 58, 75-80. Zakay, D., & Wooler, S. (1984). Time pressure, training and decision effectiveness. Ergonomics, 27,

273-284.

Accepted by Dipankar Chakravarti.