Research

Abstract:

Decision making in practice is often difficult, with many actions to choose from andmuch that is unknown. Experts play a particularly important role in such complex environments. We study the strategic provision of expert advice in a variation of the classic sender-receiver game in which the environment is complex, so knowledge of the sender’s preferred action may not reveal the receiver’s preferred action. We identify an equilibrium in which the action is exactly what the sender would choose if she held full decision making authority. This contrasts with the inefficient equilibria of the canonical model of Crawford and Sobel (1982) in which the decision making environment is simpler. Thus, strategic communication is not only more favorable to the expert when the environment is complex, it is also more effective. We explore the implications of this result on the size and structure of the choice set, the decision making mechanism, and how these vary in the complexity of the decision making problem.

Abstract:

In many real-world scenarios, experts must convey complex information using a limited number of messages. In this paper, we attempt to answer the question: how does an expert’s ability to persuade change with the availability of messages? We develop a geometric representation of the expert’s payoff when using a limited number of messages. The sender consistently performs worse with coarse communication and values additional signals. We identify bounds on this value. In a special class of games, the marginal value of a signal increases as the receiver becomes more difficult to persuade. Moreover, we show that an additional signal does not directly translate into more information in equilibrium, and the receiver might prefer coarse communication. This suggests that regulations on communication capacity have the potential to shift the balance of power from the expert to the decision-maker, ultimately improving welfare. Finally, we study the geometric properties of optimal information structures and show how they can be utilized to simplify the sender's problem into a finite algorithm.

Abstract:

Vast literatures have arisen showing how, in the face of uncertainty, a decision maker may benefit from expert advice or from strategic experimentation. Yet these literatures have largely treated these options as mutually exclusive. In practice, a decision maker has the choice of accepting expert advice, experimenting on his own, or doing both. This choice becomes particularly relevant in dynamic settings where the decision-maker can experiment during the initial periods and leverage the acquired information in subsequent ones. We develop a model that captures this possibility and show how experimentation and expertise can both emerge on the equilibrium path. While experimentation and expert advice are complements for the decision-maker, they act as substitutes for the expert. Notably, while efficient communication is feasible in a single-period interaction (Aybas and Callander, 2023), over a longer horizon, the decision-maker's ability to experiment makes communication inefficient and the quality of decision-making deteriorates.

Abstract:

We introduce a model in which homophily in social networks affects both the quality and diversity of the information to which people have access. Homophily provides higher-quality information about the actions that a group takes, since observing the payoffs of another person is more informative the more similar that person is to the decision maker. However, homophily can lead to observations about fewer actions if people similar to the decision maker choose a limited set of actions. This can lead to inefficiencies as well as inequalities across groups. We characterize conditions under which homophily hurts rather than helps social learning. Homophily lowers efficiency and increases inequality in sparse networks, but enhances efficiency and decreases inequality in dense enough networks. We also show that optimal (learning-maximizing) networks exhibit assortativity in payoff-determining characteristics, which results in incidental homophily on other innate characteristics, providing an explanation for some empirical patterns.

Abstract:

The transition to college is a challenging time during which many students suffer declines in well-being. Social connections play a key role in supporting mental health, but only tell part of the story of social life on campus. For instance, the personalities of one’s friends and neighbors on campus contribute to a “social microclimate.” Here, we quantify the collective impact of individual, social network, and community factors in the well-being of a first-year college cohort during (i) their first academic term and (ii) a stressor (the COVID-19 pandemic). Students who maintained supportive connections and belonged to emotionally stable and tight-knit microclimates reported greater well-being in their first academic term, and less anxiety when exposed to stress during the COVID-19 pandemic, highlighting the importance of both personal relationships and community factors in supporting mental health.

Abstract:

We study a model of trade with repeated interaction between a single buyer and many sellers. The buyer is initially uninformed about her valuation for the various goods and sellers are uninformed about the buyer’s demand. We model this interaction as a multi-armed bandit problem with strategic arms and seek to understand the welfare consequences of various models of buyer behavior. Similarly to Braverman et al. (2019), we show that a buyer using a no-regret (contextual) learning algorithm may be exploited by colluding sellers in an approximate Nash equilibrium for the sellers. We then show that a buyer with commitment power may extract almost all the gains from trade from the sellers in an approximate dominant strategy equilibrium for the sellers.