Idea: Information that does not help to improve future decisions can still be desirable: it reduces uncertainty and makes us savor future rewards.
Abstract: Curiosity—our desire to know—is a fundamental drive in human behaviour, but its mechanisms are poorly understood. A classical question concerns the curiosity motives. What drives individuals to become curious about some but not other sources of information? Here we show that curiosity about probabilistic events depends on multiple aspects of the distribution of these events. Participants (n = 257) performed a task in which they could demand advance information about only one of two randomly selected monetary prizes that contributed to their income. Individuals differed markedly in the extent to which they requested information as a function of the ex ante uncertainty or ex ante value of an individual prize. This heterogeneity was not captured by theoretical models describing curiosity as a desire to learn about the total rewards of a situation. Instead, it could be explained by an extended model that allowed for attribute-specific anticipatory utility—the savouring of individual components of the eventual reward—and postulates that this utility increased nonlinearly with the certainty of receiving the reward. Parameter values fitting individual choices were consistent for information about gains or losses, suggesting that attribute-specific anticipatory utility captures fundamental heterogeneity in the determinants of curiosity.
Paper available on Nature Human Behavior.
Idea: We reviewed 100+ neuromarketing studies that used EEG (electroencephalography). We summarize the best practice and highlight interesting directions for future research.
Abstract: In the past decade, marketing studies have greatly benefited from the adoption of neuroscience techniques to explore conscious and unconscious drivers of consumer behavior. Electroencephalography (EEG) is one of the most frequently applied neuroscientific techniques for marketing studies, thanks to its low cost and high temporal resolution. We present an overview of EEG applications in consumer neuroscience. The aim of this review is to facilitate future research and to highlight reliable approaches for deriving research and managerial implications. We conducted a systematic review by querying five databases for the titles of articles published up to June 2020 with the terms [EEG] AND [neuromarketing] OR [consumer neuroscience]. We screened 264 abstracts and analyzed 113 articles, classified based on research topics (e.g., product characteristics, pricing, advertising attention and memorization, rational, and emotional messages) and characteristics of the experimental design (tasks, stimuli, participants, additional techniques). This review highlights the main applications of EEG to consumer neuroscience research and suggests several ways EEG technique can complement traditional experimental paradigms. Further research areas, including consumer profiling and social consumer neuroscience, have not been sufficiently explored yet and would benefit from EEG techniques to address unanswered questions.
Paper available on Frontiers in Neuroscience.
Job Market Paper
Coarse and Precise Information in Food Labeling
Public authorities and companies often adopt simple categorical labels to convey information and promote the purchase of healthy, ethical, or environmentally-friendly products. Why are these "coarse" labels favored over more detailed ones which should allow the consumer to make better decisions? This paper investigates whether precise labels can be more effective and informative than coarse ones. In a preregistered online study conducted on a representative US sample, I manipulate front-of-package labels about foods' calorie content. I find that coarse-categorical labels generate a larger reduction in calories per serving compared to detailed-numerical labels despite providing less information (-3% and -1% calories, respectively). Results also show that participants prefer coarse labels. Choices violate the predictions of Bayesian decision theory, suggesting that consumers are less responsive to detailed information. A bounded rationality model with precision overload can capture the main experimental results: detailed labels are more complex and harder to understand, and consumers face a tradeoff between simplicity and precision. Some information helps, but too much detail can be confusing, and lead to less healthy food choices.
Idea: Sometimes more information can make us more polarized because we ask different questions.
Abstract: We show that rational but inattentive agents can become polarized, even in expectation. This is driven by agents’ choice of not only how much information to acquire, but also what type of information. We present how optimal information acquisition, and subsequent belief formation, depends crucially on the agent-specific status quo valuation. Beliefs can systematically update away from the realized truth and even agents with the same initial beliefs might become polarized. We design a laboratory experiment to test the model’s predictions; the results confirm our predictions about the mechanism (rational information acquisition) and its effect on beliefs (systematic polarization).
Idea: Contemplating safe alternatives makes us more risk averse, but the opposite does not happen with risky ones.
Abstract: We study how risk preferences may be subject to context effect specific to the risk domain—the amount of avoidable risk at any given time. Avoidable risk is captured by the riskiness of the safest option in a choice set, which induces set-dependent risk preferences. In a laboratory experiment, we find that adding safer options systematically increases risk aversion, even when the added options are not themselves chosen. By contrast, adding riskier options does not result in a detectable change in risk attitude. Our finding suggests that context effect specialized to the risk domain may overwhelm those that are more generally applicable (such as the compromise effect) when it comes to studying context-dependent risk preferences.
Idea: After a successful match, we would like to declare different preferences: we would lie more, but without ending up with a worse match.
Abstract: We use a lab experiment to study the patterns and effects of learning in two classic centralized matching mechanisms widely used in school choice and other real-world settings. We use the Deferred Acceptance (DA, strategyproof but not efficient) and Immediate Acceptance (IA, efficient but not strategyproof) algorithms. Each matching problem (round) is repeated for several periods: after being informed about the outcome of the previous match, subjects are asked again the order in which they would apply to the same schools. Between one period and the other, students gradually learn about the environment (how difficult it is to get in the high-rank schools) and their opponent behavior (if they aim for the top). We observe that subjects achieve higher payoffs under the IA mechanism and are more truthful under the DA mechanism. Subjects become gradually less truthful within the round but payoffs do not decrease. Deviations from truthfulness that do not harm the players suggest that costless mistakes are compatible with extensive experience.
Working paper available upon request.
Work in Progress
Idea: When we can collect information from multiple sources, we tend to search in-depth from the same place.
Abstract: We introduce an experiment on dynamic allocation of attention that combines features from stopping and bandit problems. The decision maker refines their own belief about the uncertain state by allocating attention dynamically over different sources of information, that are biased toward alternative actions. Participants decide freely how much time to spend collecting information (as in a stopping problem with costly information acquisition), as well as how to consult the sources of information (similarly to a bandit problem). Participants systematically depart from the optimal behavior both in the amount and direction of information collected. On average, participants collect less information than optimal (low search amount), they collect information skewed towards the prior information (confirmation bias), and do not update enough the beliefs after collecting contradictory evidence (anchoring bias).
Idea: We collect information mostly when we know we are going to need it soon.
Abstract: Rational Expectations models assume that every agent knows everything that is possible to know at all times. There is abundant evidence that all three underlying assumptions do not hold, but typically is difficult to isolate them using traditional datasets. In this paper we focus on the timing of information acquisition: when do agents decide that they want to know (updating their current beliefs)? In an incentivized experiment, we use a simple forecasting task to observe information acquisition and actions. In order to isolate the role of expectations, we exogenously manipulate whether agents know in advance when they will face high-stake decisions. Expectations play a major role: even when there is no benefit of acquiring information in advance, knowing the timing of high-stake decisions drives the information acquisition behavior, and therefore the forecast accuracy. In order to capture the main features of the observed behavior, we introduce a noisy attention model. When agents do not perfectly remember the information available, timing of decision drives search. As a consequence, they are more likely to bunch information acquisition around the important decisions (instead of smoothing). This is at odds with the predictions of a traditional Bayesian agent, that would predict information smoothing.
Noisy Integration of Value Differences in Multi-Attribute Choice Problems
Idea: Multi-attribute choices are difficult. Comparing values instead of remembering them makes it easier.
Abstract: Many everyday decisions require the evaluation and comparison of alternatives across multiple dimensions. Recent experimental evidence suggests that humans suffer from systematic biases even in simple averaging tasks. Noisy perception models (in psychophysics) and random utility models (in economics) usually assume that the evaluation of each observation is implemented in isolation, and this assumption is at odds with extensive field and experimental evidence of context effect. In order to study human information integration, we use an averaging task with binary choice between compound lotteries. Participants observe two sequences of simple lotteries, whose winning probabilities are depicted using bars with different heights, and choose which compound lottery to implement. After presenting the main features of the choice data, we show that models that assume integration of differences fit the data better than their counterparts with separate evaluation of individual values. Finally, we compare the behavior in two treatments with different processes generating the trials values. The same distribution of value differences is generated by two distinct single-value distributions. The two treatments show significant differences, suggesting that the way differences are generated is also relevant. This result is consistent with a model of salience with focusing weight in favor of observations that are unusual or surprising relative to the reference. We discuss implications of noisy integration of value differences in a buyer-seller obfuscation setting, and as a possible explanation for context effect and violation of stochastic transitivity.
Idea: Fast news are more likely to be accurate, but inaccurate news are more likely to be fast.
Abstract: We live in the Information Age, an epoque characterized by easy access to an unprecedented amount of knowledge, but also a series of new problems related to information and its effect on society. In this paper we focus on the issue of inaccurate news (fake news and unreliable sources). Why do we observe a high amount of inaccurate information? We examine how reputation can provide an explanation for this phenomenon. In a fast-paced world, media outlets demonstrate their reputability through the speed and accuracy of their reports, and face a tradeoff between these two channels. We present a sender-receiver model in which media (senders) face career concerns and choose between generating fake news or waiting for a reliable source. Readers (receivers) evaluate the media based on the accuracy and speed of their news. The model predicts that in equilibrium 1) receivers assign higher reputation to early news, and fake news 2) decrease in the quality of the sender and 3) increase in the time available to send a report. We bring these predictions to the lab with an experiment that investigates the behavior of both types of players. Do readers assign higher reputation to fast reporters? Do low-quality media generate fast-fake news? Our experiment will answer these and other open questions.
We are grateful for IFREE's generous support for this project. We plan to conduct the experiment in Spring 2022.