Nikhil Vellodi

We study a sequential experimentation model with endogenous feedback. Agents choose between a safe and risky action, the latter generating stochastic rewards. When making this choice, each agent is selfishly motivated (myopic). However, agents can disclose their experiences to a public record, and when doing so are pro-socially motivated (forward-looking). When prior uncertainty is large, disclosure is both polarized (only extreme signals are disclosed) and positively biased (no feedback is bad news). When prior uncertainty is small, a novel form of unraveling occurs and disclosure is complete. Subsidizing disclosure costs can perversely lead to less disclosure but more experimentation.

A present-biased decision maker (DM) faces a two-armed bandit problem whose risky arm generates random payoffs at exponentially distributed times. The DM learns about payoff arrivals through informative feedback. At the unique stationary Markov perfect equilibrium of the multi-self game, positive feedback supports greater equilibrium welfare than both negative and transparent feedback. Regardless of the form of feedback, the DM’s behavior exhibits indecision, deriving from their desire to procrastinate. We relate our findings to the theory of self-prospection -the process of imagining future goals and outcomes when seeking motivation in the present.

We study a sequential experimentation model with endogenous feedback. Agents choose between a safe and risky action, the latter generating stochastic rewards. When making this choice, each agent is selfishly motivated (myopic). However, agents can disclose their experiences to a public record, and when doing so are pro-socially motivated (forward-looking). Disclosure is both polarized (only extreme signals are disclosed) and positively biased (no feedback is bad news). The extent of disclosure is non-monotone in prior uncertainty. Subsidizing disclosure costs can paradoxically lead to less disclosure, but more experimentation.

We study a model of time preference in which both current consumption and the memory of past consumption enter “experienced utility”—or the felicity—of an individual. An individual derives overall utility from her own felicity and the anticipated felicities of future selves. These postulates permit an agent to anticipate future regret in current decisions, and generate a set of novel testable implications in line with empirical evidence. The model can be applied to disparate phenomena, including present bias, equilibrium savings behavior, anticipation of regret, and career concerns.

We study how regulating data usage impacts innovation in digital markets. Platforms commonly use proprietary data about third-party sellers to inform their own competing offerings, dampening incentives for innovation. We model this interaction and characterize how data usage restrictions reshape these incentives. An outright ban on data usage may boost or curtail innovation, depending upon the thickness of the right tail of demand for new products. More flexible rules controlling when and what data is made available can always improve the effectiveness of regulation. Our results contribute to an ongoing policy discussion regarding competition in digital markets.