Schedule - Parallel Session 1 - Information
Engineering F1.10 (note the change of room from F105-106 to F1.10) - 11:00 - 12:30The Misery of Knowing: When Information is Aversive
Emily Ho; David Hagmann; George Loewenstein
Abstract
The Information Aversion Scale is a unified treatment of two primary dimensions: that of domain specificity and of psychological causal mechanism. We hypothesize that individuals are more inclined to avoid information with negative valence and that this inclination is moderated by domain-specific self-serving bias. For example, a smoker will tend to avoid information about cancer screenings but not necessarily avoid looking at credit card statements. Avoiding information serves to maintain subjective social reality (Bless, Fiedler, & Strack, 2004), whereby an individual’s perception of reality helps guide cognition and reaffirm sense of self. We develop a scale that profiles people based on their propensity to avoid information in health, finance, and intra-personal domains. Information within these fields is germane to everyday decisions and, if avoided, also can be extremely costly at both the individual and institutional (e.g., insurance companies that incur sustained costs for preventable illnesses) levels. For example, we ask participants whether they would open immediately a medical bill that they suspect will be small, but worry might be large. There may be many causes for avoiding information. For example, people may be regret averse and would rather not find out how well a mutual fund they did not invest in is performing. By avoiding this information, they deprive themselves of the opportunity to change their investment allocations in favor of a potentially better fund. To develop our scale, we make use of a large dataset provided by a survey company in which millions of people answer randomly selected questions during their regular browsing activity. In addition to previously provided responses, they answer either part or the full scale, allowing us to impute missing data. This allows us to examine correlations with a diverse set of other scales and miscellaneous questions, establishing convergent and divergent validity of the theoretical information avoidance construct. For example, we may expect a positive relationship between a tendency to avoid information and pessimism, but a negative relationship between avoiding information and curiosity. We may also expect, for an unhealthy individual, a positive relationship with health-specific information aversion but no relationship with finance-specific information aversion. The Information Aversion Scale furthers the understanding of the valuation of information under risk, with broad implications for welfare-enhancing behaviors such as interventions to lower credit card debt or increasing likelihood of obtaining test results. We anticipate that the scale can also help shape policy design.
Weighing Experts, Weighing Sources: The Diversity Value
Klaus Nehring; Ani Guerdjikova
Abstract
A decision maker has to come up with a probability, preference or other judgment based on the judgments of a number of different information sources. To do so, he needs to assign weights to each source reflecting their assessed reliability. We argue that, crucially, reliability is to be understood as an attribute of sets of sources, not of sources in isolation. Specifically, we propose to view reliability as “valued diversity”, reflecting both individual source quality and similarity among sources. Intuitively, larger weight should be assigned to sources of greater quality and greater dissimilarity from the others. The main contribution of this paper is to propose and axiomatize a particular weighting rule, the Diversity Value, that captures these desiderata. The Diversity Value is defined by a logarithmic scoring criterion and can be characterized as a weighted Shapley value, in which source weights are determined endogenously. Due to the central role of source similarity, the Diversity Value frequently violates Reinforcement and exhibits the No-Show Paradox.
Repeated Search in Variable Environments and the Role of Post-Decision Search
Kinneret Teodorescu; Ke Sang; Peter Todd
Abstract
People often search for information about alternatives after they have already chosen an option, even if the choice is irreversible. While previous studies on post-decision search focused on “one shot” decisions and highlighted its irrational aspects, here we explore the possible long-term benefits of post-decision search. We focus on two (rational) motivations to conduct post-decision search in repeated settings: (1) to better understand the payoff distribution when future decisions are expected in the same environment, and (2) to improve one’s search strategy: even when future decisions are not expected in the same environment, post-decision search could be used to obtain feedback on one’s search strategy- e.g. learning that searching stopped too soon can help set a better stopping rule for future choices. To test these potential benefits, we used a simple repeated search task. In each round, a deck of cards is presented and one can choose whether to flip over the next card in the deck, or to stop the search and select the current card. In the first experiment, post-decision search was voluntary: After selecting a card, participants could continue flipping cards and see what cards would have been available if they had continued to search. Two variables were manipulated within subject: information about the distribution of values (Full Information/No Information) and repetition of the same deck in later rounds (Repeated/Unique decks). Results show people search less when provided with information about the distribution of values, highlighting the first motivation for post-decision search “collecting additional information about the environment. Repetition, on the other hand, had no significant effect, and post-decision search was significantly above zero in all conditions, which both support the second motivation “to obtain feedback about one’s own search strategy. Using a median split on the length of pre-decision search, we found that post-decision search in early rounds was correlated with improved performance later on only for “Low Searchers”: only people who initially searched little but conducted post-decision search learned to adjust their strategy appropriately. In two follow-up experiments, post-decision search was manipulated directly between subjects (Mandatory search/No search) in a no-repetition no-information environment. Results show larger improvements for subjects who were forced to conduct post-decision search, particularly for Low Searchers. When learning about the distribution was separated from setting stopping thresholds, post-decision search helped Low Searchers, mainly by increasing their learning about the distribution before generating thresholds. To conclude, although post-decision search might appear irrational, especially if one does not expect to encounter the same environment in the future, our results suggest that it can help Low Searchers to modify their data collection behaviors and improve performance.
Trust and Verify: the Informational Value of Early Trust
Michael Yu; Cleotilde Gonzalez
Abstract
Knowing who to trust helps us to capture the benefits of trusting while avoiding the costs of betrayal. Learning who to trust, however, is often confounded with the decision to trust. When interacting with a new person, concerns of betrayal could lead us to avoid trust and forgo information on whether the person is trustworthy. In relationships defined by repeated interactions, early avoidance of trust could lead to delays in trust development. Across two studies, we examine the impact of initial trust behaviors in a repeated Trust Game using simulated partners who are either always return points or always keep points and who are paired with photographs that are pretested as either appearing trustworthy or untrustworthy. In Study 1, participants required to engage in initial trust made participants more sensitive to their partner’s behavior early in the repeated interactions than participants who were free to choose. These findings were driven by participants matched with partners who returned points but had an untrustworthy appearance. Study 2 focused on partners who returned points but had an untrustworthy appearance. Participants asked to “find out” about such partners’ behavior in their initial interactions were more likely to send points to their partners early in repeated interactions than participants who were free-to-choose. However, these participants were no more likely to expect their partners to return points. In contrast, an intervention in which participants were asked to take the perspective of their partner resulted in less significant increases in the likelihood to send points and did significantly increase expectations of their partner returning points. Our results suggest that a simple information-seeking intervention can promote trust development without changing expectations of trustworthiness. This form of intervention might be useful in environments where new trust-based partnerships may need to be formed and could better reflect the goals of developing trust rather than simply increasing it.