Barbara Mellers - US grants

This proposal outlines a series of experiments that would explore important questions of judgment and choice under uncertainty and contribute to the understanding of basic decision processes. Central to the study is the development and testing of a theory of response mode effects that will provide new insights regarding alternative explanations for preference reversals in choice, including regret theory and contingent weighting. This topic is receiving a great deal of attention from researchers because of the implications for the study of decision making if systematically different rank orders are obtained depending on the choice between formally equivalent assessment procedures. Thus the project also has general theoretical importance, for the study of contextual effects on decision processes, as well as practical implications for the development of improved decision- aiding tools. Three studies are proposed that extend earlier work of the principal investigator to provide new information on the stability and consistency of preferences and decisions, based on experiments conducted under a variety of conditions and in a variety of contextual settings.

This special award will give Ph.D student Lisa Ordonez additional flexibility in pursuing her graduate studies and research initiation in cognitive research. This award will strengthen minority participation in research in this area.

This project studies the effects of similarity on individual judgement and choice. Three sets of experiments are proposed. The first set of experiments investigates multiattribute stimuli with similar levels along one dimension and dissimilar levels on another. The hypothesis that similarity along one dimension will enhance differences along another dimension is tested. The second set of experiments investigates the effect of the stimulus range on attribute importance. The hypothesis that similarity along a stimulus dimension will decrease the importance of that dimension is tested. The third set of studies examines similarity between stimulus and response scales and tests the conjecture that when a stimulus dimension is more closely related to the response scale, that stimulus will receive greater weight to the extent that it is perceived to be more valid. When the validity of the stimulus is unconfounded with response scale similarity, the more valid cue is weighted more heavily. By contributing to a better understanding of how individuals evaluate multiattribute items, this project may lead to more realistic models of individual valuation for use in decision making, marketing and voting. ***//

The PI proposes a series of experiments which will investigate the effect of decisions on emotions. The outcomes of various gambles are hypothesized to influence participants' emotions dependent on such factors as the range of potential outcomes and the "closeness" with which the outcome occurs. In addition, the PI proposes to investigate how accurately an anticipated emotion corresponds to an experienced emotion after the awaited event actually occurs. The PI also will examine the influence of emotions on subsequent outcomes and the hedonic experience of various outcomes in social and "real-world" contexts.

`Evaluating the Outcomes of Risky Options`. Rational choice theory assumes that the utility derived from receiving an outcome as a result of a decision depends only upon the value of that outcome - winning $10 is the same whether the alternative possibility was winning $1000 or losing $1000. As such, emotions such as regret or rejoicing, disappointment or elation, which inherently involve comparisons of what is with what might have been, play no role in choice. In prior research, the PI has examined how such emotions influence individuals' feelings regarding the outcomes they receive. The present proposal seeks funds to continue this research. Specifically, three sets of experiments are outlined. The first test implications of the model of emotion developed by the PI based on earlier research. The second set of experiments examine the extent to which anticipated emotions may not only influence ex post evaluations of happiness or dissatisfaction with outcomes but also influence the decisions people make. The final set of experiments examine the influence of anticipated emotions in richer and more diverse contexts.

Anyone who has ever made an important decision knows that anticipated emotions play a role. There are effects of the feelings and moods we experience at the time of a choice and effects of anticipated feelings about future consequences of choice. We have developed a descriptive theory of the anticipated pleasure of outcomes called decision affect theory. The theory rests on the assumption that pleasure and pain are evaluated relative to the status quo and other salient reference points, such as counterfactual outcomes, personal goals, and social comparisons. All of these comparisons have asymmetric effects; the incremental displeasure of obtaining the worse of two outcomes is greater in magnitude than the incremental pleasure of getting the better of two outcomes. Finally, beliefs about the likelihood of outcomes moderate the effects of comparisons. Surprising outcomes lead to stronger emotional responses than expected outcomes.

In past research, we have applied our theory to risky choice. We assume that people anticipate the pleasure or displeasure of each outcome of an option (as predicted by decision affect theory), weight those feelings by the chances they will occur, and sum over outcomes. This process yields an average anticipated pleasure for an option. People then compare the average pleasure of all options and select the one that maximizes expected pleasure.

Our theory of choice is called subjective expected pleasure theory. This proposal builds on past work and has three primary goals. First, we will extend our theory to riskless choice and test it by conducting seven studies. Four of them examine endowment effects and the status quo bias. Three additional studies investigate other well-known phenomena in riskless choice, such as asymmetric dominance, preference reversals, and advantages vs. disadvantages. Our second goal is to examine the effects of moods on anticipated emotions, beliefs, and choice. Although we expect to find mood-congruent effects, it is quite possible that moods can have opposing effects on judgments and choice. Our third goal is to examine the accuracy of hedonic forecasts immediately after the decision has been resolved and some time later. Accuracy is essential for emotion-based choice; decisions based on faulty predictions will surely be sub-optimal. Research suggests that people focus on attributes that are salient at the moment, but not necessarily important later one. That is, they are insensitive to adaptation. We will explore affective forecasts in three real world studies and three laboratory studies. If deviations in forecasts occur, we will explore methods for improving accuracy.

This award provides funding for a 50-station Experimental Social Sciences Laboratory with the primary role of studying individual and group behavior. The main intellectual merit of the Lab is that it facilitates efforts to synthesize models of behavior that cut across traditional disciplinary lines-that is, the mission of the Lab is to foster interdisciplinary experiments drawn from across the social sciences. Previously, a division of labor has arisen between researchers developing theory in the area of individual and group behavior and those performing experiments. This division is anachronistic. One of the intellectual merits of the proposal is that it seeks to abolish this division by dramatically lowering the barriers to conducting experiments. The design of the Lab emphasizes maximum flexibility at minimum hassle so as to accommodate highly varied approaches to conducting experiments. The physical configuration of the lab stations will be highly mobile, and the computing environment will make it a Lab without wires. The Lab will be staffed with a full-time lab manager and part-time programmer so that the creativity of the researcher, and not administration tasks, is the binding constraint to research productivity. Together, this design approach provides a highly customizable setting for researchers across disciplines. Similarly, the Lab will develop web-based and highly flexible software for the implementation of experiments without the need for programming expertise on the part of researchers. This software development effort, which is already underway, will be done in collaboration with experts at HP Labs.

The Lab benefits the research lives of faculty at UC Berkeley as well as attracting visitors from other institutions from across all fields of the social sciences. An important part of these efforts to attract outside scholars centers on our development of a program to train young scholars in the field of behavioral economics in conjunction with a successful program in this area run out ofUC Berkeley under the auspices of the Russell Sage Foundation. In addition to research, the broader impact of the Lab will occur in the areas of education and application. The Lab would contribute to the education of graduate and undergraduate students by augmenting discussion of important findings relating to behavior with hand-on experience,' thus enriching their educational experience and giving them an additional tool when entering the labor force upon graduation. The results and fmdings of the Lab are likely to have important market and policy implications which might come to be incorporated in practice.

UC Berkeley is uniquely positioned to exploit synergies arising from an interdisciplinary approach to experiments in individual and group behavior. The proposal has sparked strong interest from researchers at UC Berkeley in economics, psychology, political science, as well as in fields as far-reaching as computer science, engineering, information systems, and business. Finally, the Lab is in a position to exploit Berkeley's large and diverse student population, making it particularly amenable to studies involving gender and historically underrepresented minorities.

IARPA (Intelligence Advanced Research Projects Activity) recently sponsored a series of geopolitical forecasting tournaments to discover the most accurate possible methods for crowd-sourcing and aggregating probabilistic predictions of geopolitical events. Our group, the Good Judgment Project, competed and won the tournament each year from 2011 to 2015. In a 2015 study based on data from this project, Mellers, Arkes, and Chen examined changes in actively open-minded thinking (i.e., the tendency to look for evidence that contradicts one's favored views) between the time participants entered and exited the tournament. Forecasters were significantly more open-minded at the end of the tournament. Even opinionated forecasters appeared to become more circumspect when making judgments that were publicly scored for accuracy. These incentives may have encouraged self-critical and flexible patterns of thinking because the quest for accuracy may have trumped the desire to maintain partisan positions. The boost in open-mindedness (perhaps driven by the self-discipline needed to be accurate) may well be the most important discovery of the IARPA tournaments. This project moves this research forward with designs for new tournaments to pry open closed minds still further. The research team examines three behavioral interventions that might increase open-mindedness and reduce polarization that characterizes many political debates. The discovery of methods to change opinions and cognitive styles may reduce divisiveness and ultimately improve decision making in a world that seems more fractured and divided than ever.

This project explores whether forecasting tournaments can be designed to increase open-mindedness. The research team uses an existing public tournament to explore three factors. The first is accountability to a heterogeneous forecasting team with variability in political orientations. The previous IARPA tournaments showed that team forecasts were more accurate than independent forecasts. The researchers, however, never explored team composition, and this factor may be critical. The second factor is group pre-commitment to cognitive flexibility. Before the outcomes of events are known, forecasters will state whether they would be willing to change their minds about various debates depending on the outcomes. In "before" surveys, participants will be asked: If outcome A occurs (or does not occur), how much should your political party (or the other party) change its mind about issue X, "After" surveys will measure changes in cognitive styles and political views. The third factor is perspective taking. When forecasting questions are released, the researchers will ask liberals to write a paragraph to express the conservative position, and vice versa. Participants will be told that their work will be evaluated based on how well it captures the essence of the other side's position. The research team will use Pre and post measures of open-mindedness and political viewpoints to evaluate the effectiveness and changes that occur from the interventions.