Decision processes brown bag, 2000-2001

12-1:30 PM, Mondays, 204 SHDH

Other relevant events on campus

Sally Blount

The scarcity of time is a central concern in modern industrialized societies. Yet, little is truly known of the character of time. Objectively, time is a non-replenishable, person-specific asset: it can only be used in limited, pre-specified quantities. Its value is realized through the individual's investment to achieve desired material and psychological outcomes. But as philosophers and anthropologists have long noted, and subsequent research has demonstrated, human beings do not track time systematically. The experience of time is subjectively perceived and valued. Individuals need clocks, watches, schedules and plans; and communities need markets to create the perception that time is a measurable commodity. With this perspective in mind, our research offers an alternative framework for understanding how individuals perceive and value time, and how temporal valuation translates into social outcomes.

Our work begins with the assumption that people have preferences not only for outcomes, but also the pace at which they attain them; i.e., people have preferences for how long getting and doing things takes. These pacing goals affect how time is valued and how people behave.

In the laboratory, we have begun to examine how people account for pace individually and in social interactions. We have identified the in-synch preference, which we define as an individual-level tendency to prefer the experience of feeling in-pace to that of feeling out-of-pace in social interaction. Some of the implications of the in-synch preference that we have shown in the laboratory include the finding that in-pace negotiators create value more effectively than out-of-pace negotiators, and are significantly more likely to feel that they collaborated well together and to like their negotiating partners better.

Thus, we suggest that individual pacing goals and how they are aligned or misaligned across parties offers an alternative model for understanding temporal evaluation in social contexts, and how such ``socio-temporal'' evaluation affects social outcomes

Daniel Gilbert

One of psychology's most fundamental assumption is that people strive to achieve positive hedonic states, and that to do so they must do two things well. First, they must be able to predict how they will feel in a variety of possible futures, and second, they must act to bring about the best of these and avoid the worst. Although finding a target would seem to be much easier than hitting it, I will describe recent experimental evidence suggesting that people make fundamental and systematic errors when predicting their hedonic reactions to future events. I will then describe research on the psychological mechanisms that may be responsible for these errors.

J. Keith Murnighan, Deepak Malhotra, and J. Mark Weber

Trust has been defined as a psychological state comprising the intention to accept vulnerability based upon positive expectations of the intentions or behavior of others (Mayer, Davis, and Schoorman, 1995; Rousseau, Sitkin, Burt and Camerer, 1998). Common wisdom acknowledges that the fragility of relationships cannot be overcome contractually and that, as a result, trust becomes a necessary element in many interpersonal interactions. Early data suggests that trust tends to accumulate gradually but can be broken quickly (Gamson, 1968; Murnighan and Leung, 1976; Rapaport and Chammah, 1965; Roth and Murnighan, 1978). Also, once trust is broken, the data suggest that it is very difficult to reestablish.

This paper addresses some converse, counterintuitive findings with regard to trust. The paper reviews several recent research projects, which show that a linear, rational model of trust and trusting behavior does not provide a clear basis for behavioral expectations. For instance, findings suggest that a breakdown in trust can, in fact, be repaired with talk, but that stronger repair comes from personal, substantive penance (Bottom, Gibson, Daniels, and Murnighan, 2000). Other data on one version of the Trust Game suggest that when contracts are used to promote or mandate cooperation, interacting parties are likely to attribute their resulting cooperation to the existence of the contract and, as a result, trust has little opportunity to develop (Malhotra and Murnighan, 2000a). A third study documents the effects of unilateral trusting actions. The findings show that trustors should not expect strict reciprocity and that recipients of trust often expect to be trusted fully (Pillutla, Malhotra, and Murnighan, 2000). That is, actions that convey "partial" trust are not particularly effective in generating reciprocity and that in certain situations it is best to trust fully or not at all. A fourth study documents the difficulty of reestablishing trust in competitive interactions, even when communication is possible, future interaction is expected, and cooperation is mutually beneficial (Malhotra and Murnighan, 2000b). Interestingly, relationships based on mutually shared expectations of distrust provide a more likely basis for future trust than relationships based on asymmetric expectations, i.e., one party trusting, one not.

The research we review suggests that trust results from complex attributional processes and that understanding the development of trust requires us to focus on the cognitions of the trusting and the trusted parties. A central finding is that people often take risks to establish trust rather than establishing trust so that they can take risks. This reverses our original, working conceptualization of trust. It also suggests the need for an attributional model of trust (which we begin to develop) that has a variety of theoretical and practical implications.

Dan Ariely and Jonathan Levav

Many individual decisions take place in a group context wherein group members voice their choices sequentially. In this paper we examine the impact of this dynamic decision process on individuals' choices and satisfaction with their outcomes. We propose that choices reflect a balancing of two classes of goals: goals that are strictly individual and goals that are triggered by the existence of the group. The latter sometimes result in choices that undermine personal satisfaction and increase regret. We find support for goal balancing in three studies where we tracked consumers' order of dishes and drinks. In the Lunch Study we found that real groups (tables) choose more varied dishes than would be expected by random sampling of the population of all individual choices across all tables. The Beer Study demonstrated that this group level variety-seeking is attributable to the interaction-implicit or explicit-among group members, and can be dissipated when the group is forced to "disband" and its members make strictly individual choices. Finally, the Wine Study demonstrated that individual choices in a group context are also aimed at satisfying goals of information gathering and self presentation in the form of uniqueness.

Jim Sherman

A feature-matching model of preference was developed that had many similarities to Tversky's contrast model of similarity judgments. This model emphasizes the importance of the direction of comparison in the judgment, and proposes a focus on the unique features of the starting point of the comparison. For preferences, it is further assumed that shared features are cancelled and only features unique to each candidate option play a role in preference and in choice satisfaction. The model has successfully predicted which objects are chosen, the degree of pre-decision conflict, and the amount of post-decision satisfaction and regret. More recent work demonstrates that initial judgments of similarity or dissimilarity between candidate objects alter subsequent preference judgments because of changes in the weights of the features. A recent model of similarity judgments, reference point reasoning, differs from Tversky's contrast model, and calls into question whether similarity and preference judgments are based on the same process. Data will address this question.

Eric Eisenstein and Wes Hutchinson

Experts are frequently disparaged by BDT researchers, who bash experts' poor performance in forecasting, calibration, and other outcome measures. But, in real-world environments, problem solving and learning are goal-directed. Experts are frequently paid to decide on a course of action from an action-space consisting of discrete alternatives rather than to forecast, judge, or predict a continuous outcome. We propose that the psychological processes of learning and reasoning in these types of discrete, goal-directed, environments differ from the underlying psychology in the more commonly studied continuous forecasting and prediction domains. This is because 1) in discrete action domains the appropriate criterion for success changes; 2) the change in the criterion of success also changes what normatively must be learned; and 3) mental effort and attention can be more efficiently expended when choosing among discrete actions.

We present the results of two experiments that demonstrate that what is learned from identical data is influenced by the goal during training. Subjects appear to efficiently allocate their attention to goal-relevant information and to selectively ignore information that appears not to be decision-relevant. In addition to confirming our basic hypothesis, we find a counter-intuitive result that thinking harder may actually hurt decision-making performance across multiple criteria.

Despite the seemingly self-evident truth that you can not make someone worse off by adding options to an already existing array of options, I will argue that there are several psychological reasons why increasing the choices available to people can in fact diminish their psychological satisfaction with choices even when it results in choices that are in some sense "better" objectively. Multiple options can result in regret, escalated expectations about how good a choice will be, and increased self-blame when the choice, inevitably, fails to live up to expectations. I will suggest that what Kahneman calls the "satisfaction treadmill" and the rise in clinical depression we are witnessing in modern America may both stem in part from what I'm calling the "choice problem."

I will also present some preliminary data that the choice problem is not a problem for everybody. We have developed an instrument that distinguishes "maximizers" from "satisficers," and we have evidence that it is maximizers especially who suffer from the tyranny of choice.

I present a model in which people use categories to think about the world around them. Faced with data, they first pick a category which best matches it. To make predictions, they ask how representative an outcome would be of the chosen category. This simple model unifies many of the experimentally documented biases: the law of small numbers, the hot hand, representativeness, and the conjunction fallacy. Moreover, the model provides enough structure that it provides readily testable out of sample predictions regarding these biases.

The thesis of this Article is that the assessment of legal policies should depend exclusively on their effects on individuals' welfare.In particular, in the evaluation of legal policies, no independent weight should be accorded to conceptions of fairness, such as corrective justice and desert in punishment. (However, the logic leading to this conclusion does not apply to concern about equity in the distribution of income, which is often discussed under the rubric of fairness.)

Our analysis begins with the argument that, when the choice of legal rules is based even in part on notions of fairness, individuals tend to be made worse off. Indeed, if any notion of fairness is ascribed evaluative weight, everyone will necessarily be made worse off in some situations. Moreover, when we examine principles of fairness and the literature that advances them, we find it difficult to identify reasons that, on reflection, justify granting importance to these principles at the expense of individuals' well-being.

Nevertheless, policy analysts and the population at large obviously find notions of fairness appealing. We conjecture that the notions' attractiveness is rooted in several factors. Namely, individuals who believe in ideas of fairness tend to behave better toward others; the notions may serve as proxy goals for instrumental objectives; and individuals may have a taste for satisfaction of the notions. Furthermore, each of these factors is a reason that notions of fairness are relevant under a welfare-oriented normative approach to social decisionmaking. As we explain, however, none of these factors warrants treating notions of fairness as independent evaluative principles.

We develop our thesis through consideration of specific conceptions of fairness that are employed in major areas of the law: torts, contracts, legal procedure, and law enforcement. We also discuss the implications of our analysis for our primary audience, legal academics and other legal policy analysts, as well as for government officials, notably, legislators, regulators, and judges.

For relevant articles, see Olin Center at Harvard Law School

The dominant view in the study of language use makes strong normative assumptions. It assumes that when people understand language they should use their common ground with the speaker, and that they indeed do just that. Research from my laboratory suggests that language users do not do what they "should." Instead of relying on common ground, people often process language egocentrically, they anchor on their own perspective and make insufficient adjustment to the other's perspective, and they overestimate their ability to communicate their intention because they experience an illusion of transparency. Such processing suggests that miscommunication is partially systematic.

We present a theory in which parameters of general learning models applied to games are functions of experience, so that only one paramter (a response sensitivity) needs to be estimated to fit data. The theory can be used to predicdt across games (e.g. mixed-strategy equilibra, coordination, dominance-solvable games) in which players seem to use differnet learning rules. The theory is more parsimonious than other theories but fits better, so controversy about how to trade off complexity with predictive accuracy is avoided. We also show that "bionic" players using the theory do better than the typical subject (other theories sometimes do *worse*).

The presentation begins with a quick survey of our "butterfly collection" of interesting specimens of web sites purporting to aid consumer choice. Then we examine the choice implications of various attribute-oriented strategies (e.g., lexicographic ordering, elimination-by-aspects, and take the best) and option-oriented strategies (e.g., most confirming dimensions, weighted adding, and multiattribute utility theory) and how those strategies are implemented in web sites. Next we will consider some mathematical results demonstrating the importance of interattribute correlations--especially whether they are positive or negative. After a quick look at the values of interattribute correlations for a variety of consumer products and seeing there is a very wide range, we will examine empirical results from web-based choice studies in which we've manipulated the correlational structure of the attributes.

The basic thesis of this talk is that images, associated with positive and negative affective feelings, guide judgment and decision-making. Specifically, it is proposed that people use an affect heuristic to make judgments. That is, representatives of objects and events in people's minds are tagged to varying degrees with affect. People consult or refer to an "affective pool" in the process of making judgments and decisions. Using an overall, readily available affective impression can be far easier - more efficient - than weighing the pros and cons or retrieving from memory many relevant examples, especially when the required judgment or decision is complex or mental resources are limited. Reliance on the affect heuristic has important implications for the rationality of all manner of decisions including those involving risk and environment.

We consider a decision problem where a group of individuals evaluate multi-attribute alternatives. We explore minimal required agreements that are sufficient to specify the group utility function. A surprising result is that, under some conditions, a bilateral agreement among pairs of individuals on a single attribute is sufficient to derive the multi-attribute group utility. We consider both the additive and multiplicative utility functions. The usefulness of our procedure for group decision making in a managerial setting is discussed.

We present results from two new experiments on the relative importance of, and subjects' differential sensitivity to, vagueness on both probabilities and outcomes. Subjects in these studies made Certainty Equivalence (CE) judgments for precise and vague gambles. In the first study subjects responded to gain gambles only; in the second they judged gain and loss gambles. Model free analyses of the results indicate (a) a higher concern for the precision of the outcomes than that of the probabilities, (b) vagueness seeking for positive outcomes (gains) and (c) vagueness avoidance for losses. The greater salience of the outcomes can be explained by the nature of the response mode (CEs). The reflection of attitudes towards vagueness in the two domains can be explained by the distinct goals of the DMs in the two cases that cause them to focus on the highest (most desirable) possible gain or the largest (most dreaded) conceivable loss. We propose and fit a new model for of decision making with vaguely specified attributes that generalizes the Prospect Theory model for the precise case. The new generalized model combines the two sub-models (preference among precise lotteries and effects of vagueness) and allows estimation of the vagueness parameters. These estimated parameters are consistent with, and confirm, the patterns uncovered by the qualitative analysis.