Periodic Reporting for period 4 - MOREV (Motivated Reading of Evidence)
Período documentado: 2024-08-01 hasta 2026-01-31
The project lies in the field of behavioral economics and aims to examine when and how agents interpret evidence in a motivated way, that is, in a manner that allows them to form and maintain the beliefs they prefer. The focus is placed on situations in which the information available to agents is hard, meaning that it is necessarily true, as opposed to soft information or cheap talk.
The project is composed of two parts. In the first part, the main novelty lies in the study of belief formation about other individuals rather than about one’s own personal characteristics. In the second part, the main novelty lies in the consideration of evidence that cannot be avoided and that may be strategically transmitted by agents with their own objectives. In both parts, experiments will be designed and conducted, and game-theoretic models will be developed.
The first part investigates how agents interpret hard information about other individuals, thereby extending the analysis of motivated beliefs to the social domain. More precisely, this part examines how economic and strategic goals affect agents’ perceptions of the characteristics of others. One subpart is devoted to the experimental study of motivated social distance, that is, when and how agents distance themselves from others mentally, for instance by believing that they share fewer common traits (such as gender or political partisanship) than they actually do. Another subpart examines the role of self-perception when individuals communicate strategically about their personal attributes to algorithms.
The second part studies how agents interpret evidence when this evidence is strategically transmitted by another party and when agents attach intrinsic value to beliefs. A concrete example is that of consumers interpreting evidence about products transmitted by firms pursuing their own selling objectives. Consumers may intrinsically prefer to believe that their purchases are not detrimental to health or to the environment. One subpart considers communication games and examines how receivers interpret vague or absent evidence when the conclusions they reach directly affect their well-being. Building on this idea, this subpart investigates motivated depth of reasoning, that is, whether agents deliberately engage in fewer steps of reasoning in some contexts than in others. This part also proposes models incorporating the fact that evidence imposes new constraints on the way beliefs can be shaped internally by a single, multi-self agent. One channel of particular interest is memory, as evidence, even when it cannot initially be avoided, may later be selectively forgotten.
The project is composed of two parts. In the first part, the main novelty lies in the study of belief formation about other individuals rather than about one’s own personal characteristics. In the second part, the main novelty lies in the consideration of evidence that cannot be avoided and that may be strategically transmitted by agents with their own objectives. In both parts, experiments will be designed and conducted, and game-theoretic models will be developed.
The first part investigates how agents interpret hard information about other individuals, thereby extending the analysis of motivated beliefs to the social domain. More precisely, this part examines how economic and strategic goals affect agents’ perceptions of the characteristics of others. One subpart is devoted to the experimental study of motivated social distance, that is, when and how agents distance themselves from others mentally, for instance by believing that they share fewer common traits (such as gender or political partisanship) than they actually do. Another subpart examines the role of self-perception when individuals communicate strategically about their personal attributes to algorithms.
The second part studies how agents interpret evidence when this evidence is strategically transmitted by another party and when agents attach intrinsic value to beliefs. A concrete example is that of consumers interpreting evidence about products transmitted by firms pursuing their own selling objectives. Consumers may intrinsically prefer to believe that their purchases are not detrimental to health or to the environment. One subpart considers communication games and examines how receivers interpret vague or absent evidence when the conclusions they reach directly affect their well-being. Building on this idea, this subpart investigates motivated depth of reasoning, that is, whether agents deliberately engage in fewer steps of reasoning in some contexts than in others. This part also proposes models incorporating the fact that evidence imposes new constraints on the way beliefs can be shaped internally by a single, multi-self agent. One channel of particular interest is memory, as evidence, even when it cannot initially be avoided, may later be selectively forgotten.
As announced, the project resulted in the production of several research articles (10 in total), including studies based on controlled experiments, theoretical work, and projects combining theory and experiments. The main achievements of each part can be summarized as follows.
The first part of the project resulted in two main contributions. First, the article “Competition, Cooperation and Social Perceptions” shows that social perceptions are indeed motivated. More precisely, it demonstrates experimentally that the same evidence about an individual can be interpreted differently when that individual is presented as a competitor or as a co-worker, all else being equal. In this way, this work is one of the first to bring the motivated interpretation of evidence into the social realm. Further work will examine how individuals strategically construct images of others in order to feel better or to motivate their own effort.
A second contribution lies in showing that the way individuals perceive themselves in relation to others affects how they strategically disclose information to algorithms. This is the object of the article “Strategic Information Disclosure to Classification Algorithms: an Experiment”. In this work, experimental subjects must manage their private data to prevent an algorithm from guessing who they are. Explaining to subjects how the algorithm works (as typically required by the transparency provisions of the General Data Protection Regulation) is not enough for them to play optimally; they must also understand how common their own characteristics are in the algorithm’s training data. Given that people have motivated and biased views of who they are, it may be particularly difficult to design policies that truly help them game algorithms.
In the second part of the project, the main achievement was obtained in “Motivated Skepticism”. The experimental literature on the formation of motivated beliefs usually studies the interpretation of information when this information is provided by a neutral experimenter. However, in many economic contexts, the information interpreted by individuals is itself transmitted strategically by another agent. For instance, firms transmit information about products to consumers in a strategic way. In the proposed experiment, individuals exercise skepticism as predicted by theory, that is, they interpret absent or vague information in the way least favorable to the sender, as long as these skeptical inferences do not lead to disturbing conclusions. Think, for instance, of consumers who do not reason about the absence of green labels in order to avoid reaching the conclusion that their purchases are detrimental to the environment. In this paper, two behavioral models are proposed to explain why subjects exercise skepticism in a motivated way. This paper is one of the very first to bring the formation of motivated beliefs into strategic contexts.
In the second part of the project, two additional published articles study the role of memory in the motivated interpretation of evidence. One work is a theoretical piece that proposes to model selective memory as a disclosure game between an agent and his future self. In this paper, “Selective Memory of a Psychological Agent”, one notable achievement is the provision of a catalogue of belief-based utility functions. In “The Motivated Memory of Noise”, agents cannot forget disturbing signals about their own ability, so they forget the reliability of the signal-generating process instead. This paper complements the experimental literature on selective memory by shedding light on a new, complementary path through which individuals maintain a positive image of themselves.
The research was presented at various seminars and conferences (more than 40). The careers of all the young researchers involved in the project, two post-doctoral researchers and two PhD students, were greatly supported by the project.
The first part of the project resulted in two main contributions. First, the article “Competition, Cooperation and Social Perceptions” shows that social perceptions are indeed motivated. More precisely, it demonstrates experimentally that the same evidence about an individual can be interpreted differently when that individual is presented as a competitor or as a co-worker, all else being equal. In this way, this work is one of the first to bring the motivated interpretation of evidence into the social realm. Further work will examine how individuals strategically construct images of others in order to feel better or to motivate their own effort.
A second contribution lies in showing that the way individuals perceive themselves in relation to others affects how they strategically disclose information to algorithms. This is the object of the article “Strategic Information Disclosure to Classification Algorithms: an Experiment”. In this work, experimental subjects must manage their private data to prevent an algorithm from guessing who they are. Explaining to subjects how the algorithm works (as typically required by the transparency provisions of the General Data Protection Regulation) is not enough for them to play optimally; they must also understand how common their own characteristics are in the algorithm’s training data. Given that people have motivated and biased views of who they are, it may be particularly difficult to design policies that truly help them game algorithms.
In the second part of the project, the main achievement was obtained in “Motivated Skepticism”. The experimental literature on the formation of motivated beliefs usually studies the interpretation of information when this information is provided by a neutral experimenter. However, in many economic contexts, the information interpreted by individuals is itself transmitted strategically by another agent. For instance, firms transmit information about products to consumers in a strategic way. In the proposed experiment, individuals exercise skepticism as predicted by theory, that is, they interpret absent or vague information in the way least favorable to the sender, as long as these skeptical inferences do not lead to disturbing conclusions. Think, for instance, of consumers who do not reason about the absence of green labels in order to avoid reaching the conclusion that their purchases are detrimental to the environment. In this paper, two behavioral models are proposed to explain why subjects exercise skepticism in a motivated way. This paper is one of the very first to bring the formation of motivated beliefs into strategic contexts.
In the second part of the project, two additional published articles study the role of memory in the motivated interpretation of evidence. One work is a theoretical piece that proposes to model selective memory as a disclosure game between an agent and his future self. In this paper, “Selective Memory of a Psychological Agent”, one notable achievement is the provision of a catalogue of belief-based utility functions. In “The Motivated Memory of Noise”, agents cannot forget disturbing signals about their own ability, so they forget the reliability of the signal-generating process instead. This paper complements the experimental literature on selective memory by shedding light on a new, complementary path through which individuals maintain a positive image of themselves.
The research was presented at various seminars and conferences (more than 40). The careers of all the young researchers involved in the project, two post-doctoral researchers and two PhD students, were greatly supported by the project.
The project advanced our understanding of "the demand side of motivated beliefs" by showing that individuals form motivated beliefs in strategic settings and about others, as well as our understanding of "the supply side of motivated beliefs" by modeling and proposing new experiments on selective memory and motivated skepticism.