WISE18: Abstracts

Mohammad Akbarpour (Stanford) “Just a Few Seeds More: Value of Network Information for Diffusion” (with Suraj Malladi and Amin Saberi)

Identifying the optimal set of individuals to first receive information (‘seeds’) in a social network is a widely-studied question in many settings, such as the diffusion of information, microfinance programs, and new technologies. Numerous studies have proposed various network-centrality based heuristics to choose seeds in a way that is likely to boost diffusion. Here we show that, for some frequently studied diffusion processes, randomly seeding s+x individuals can prompt a larger cascade than optimally targeting the best s individuals, for a small x. We prove our results for large classes of random networks, but also show that they hold in simulations over several real-world networks. This suggests that the returns to collecting and analyzing network information to identify the optimal seeds may not be economically significant. Given these findings, practitioners interested in communicating a message to a large number of people may wish to compare the cost of network-based targeting to that of slightly expanding initial outreach.

Itai Arieli (Technion) “When is the crowd wise?” (with Yakov Babichenko and Rann Smorodinsky)

Consider a setting where many individuals forecast the (unknown) state of nature based on signals they receive independently. We refer to the joint distribution over the states and signals as an "information structure." An information structure is identifiable (`the crowd is wise') if the collection of forecasts is sufficient to determine the state of nature, even without knowing the underlying information structure. We characterize the set of identifiable information structures and propose a scheme that uniquely identifies the state of nature for the finite case.

Yakov Babichenko (Technion) “Robust Forecast Aggregation” (with Itai Arieli and Rann Smorodinsky)

Bayesian experts who are exposed to different evidence often make contradictory probabilistic forecasts. An aggregator, ignorant of the underlying model, uses this to calculate her own forecast. We use the notions of scoring rules and regret to propose a natural way to evaluate an aggregation scheme. We focus on a binary state space and construct low regret aggregation schemes whenever there are only two experts which are either Blackwell-ordered or receive conditionally independent signals. In contrast, if there are many experts with conditionally i.i.d. signals, then no scheme performs (asymptotically) better than a (0:5; 0:5) forecast.

Aislinn Bohren (Carnegie Mellon University and University of Pennsylvania) “Misinterpreting Social Outcomes and Information Campaigns” (with Daniel N. Hauser)

Laura Doval (Caltech) “Sequential Information Design” (with Jeff Ely)

We study games of incomplete information as both the information structure and the extensive-form vary. An analyst may know the payoff-relevant data but not the players' private information, nor the extenstive-form that governs their play. Alternatively, a designer may be able to build a mechanism from these ingredients. We characterize all outcomes that can arise in an equilibrium of some extensive-form with some information structure.

Ignacio Esponda (UCSB) “Convergence and stability in misspecified Markov decision processes”

Mira Frick (Yale) “Misinterpreting Others and the Fragility of Social Learning” (with Ryota Iijima and Yuhta Ishii)

Drew Fudenberg (MIT) “Learning, Experimentation, and Equilibrium Refinements” (with Kevin He)

This talk will survey 3 papers with Kevin He that use the the theory of learning in games to provide foundations for equilibrium refinements. The core idea is that actions that are off of the equilibrium path are sometimes played by new agents who are uncertain of the distribution of strategies used in society, and that some experiments are more valuable and hence more common than others.

Ben Golub (Harvard) “Social Learning in a Dynamic Environment” (with Krishna Dasaratha and Nir Hak)

Agents learn about a state using private signals and the past actions of their neighbors. In contrast to most models of social learning in a network, the target being learned about is moving around. We ask: when can a group aggregate information quickly, keeping up with the changing environment? First, if private signal distributions are diverse enough across agents, then Bayesian learning achieves good information aggregation as long as individuals observe sufficiently many others. Second, without such diversity, Bayesian information aggregation can fall far short of good aggregation benchmarks, and can be Pareto-inefficient. Third, good aggregation requires anti-imitation; without it, agents' estimates are inefficiently confounded by "echoes." Our stationary equilibrium learning rules incorporate past information by taking linear combinations of other agents' past estimates (as in the simple DeGroot heuristic), and we characterize the coefficients in these linear combinations. The resulting tractability can facilitate structural estimation of equilibrium learning models and testing against behavioral alternatives, as well as the analysis of welfare and influence.

David Hirshleifer (UC Irvine) “Visibility Bias in the Transmission of Consumption Norms and Undersaving” (with Bing Han and Johan Walden)

We study how bias in the social transmission of information affects saving behavior. In the model, consumption is more salient than non-consumption. This visibility bias causes people to perceive that others are consuming heavily owing to favorable information about future wealth. The biased transmission of beliefs reduces saving and increases the equilibrium interest rate. Overconsumption increases with social network connectivity and with individual centrality within the network. Information asymmetry about the wealth of others dilutes the inference from high observed consumption that the future prospects are good. In consequence, in contrast with the Veblen wealth-signaling approach, information asymmetry about wealth reduces overconsumption. The visibility bias approach offers a novel explanation for the dramatic drop in the savings rate in the US and several other countries in the last thirty years based on shifts in observability in the social network. In contrast with other approaches, owing to the majority illusion in social networks, the visibility bias approach suggests that relatively simple disclosure-based policy interventions can ameliorate undersaving.

Matthew Jackson (Stanford) “Access to information in networks and distorted decisions”

Ilan Lobel (NYU) “Surge Pricing and Its Spatial Supply Response” (with Omar Besbes and Francisco Castro)

We consider the pricing problem faced by a platform matching price sensitive customers to flexible supply units within a geographic area. This can be interpreted as the problem faced in the short-term by a ride-hailing platform trying to match supply and demand within a city. We propose a framework in which a platform selects prices for the different locations, and drivers respond by choosing where to relocate based on prices, travel costs and driver congestion levels. Our contributions are along two dimensions. We first derive general results on the structure of optimal pricing policies. In particular, we derive structural properties of supply equilibria and the corresponding utilities that emerge and establish a form of spatial decomposition, which allows us to localize the analysis. In turn, uncovering an appropriate knapsack structure to the platform's problem, we establish a crisp local characterization of an optimal solution and the corresponding supply response. In a second set of results, we specialize the analysis to a family of models that isolates the impact of supply demand imbalances by introducing a demand shock. We derive in quasi-closed form the optimal solution across the city, highlighting the implications of the strategic nature of supply units. In particular, we show that the platform will use prices to create damaged regions where demand is shut-down or driver congestion is artificially high, incentivizing a suitable number of drivers to relocate towards the demand shock. Furthermore, the optimal solution, while better balancing supply and demand around the shock, also ends up inducing movement away from it.

Teddy Mekonnen (Caltech) “Information Acquisition and the Value of Transparency” (with René Leal Vizcaı́no)

We study the strategic effect of information in Bayesian games by comparing two different classes of endogenous information acquisition: one in which information acquisition is a covert activity (a player cannot observe how much information her opponents acquire) and another in which information acquisition is an overt activity. We provide a taxonomy of the strategic effect, which we call the Value of Transparency, and explore its connection to the demand and value of information in overt and covert games. We also analyze the strategic role of information acquisition as a barrier to entry in oligopolistic competition. (Talk is based on one of the applications from our paper, Bayesian Comparative Statics)

Manuel Mueller-Frank (IESE) “Social Learning Equilibria” (with Elchanan Mossel, Allan Sly and Omer Tamuz)

We consider social learning settings in which a group of agents face uncertainty regarding a state of the world, observe private signals, share the same utility function, and act in a general dynamic setting. We introduce Social Learning Equilibria, a static equilibrium concept that abstracts away from the details of the given extensive form, but nevertheless captures the corresponding asymptotic equilibrium behavior. We establish conditions for agreement, herding, and information aggregation in equilibrium, highlighting a connection between agreement and information aggregation.

Mallesh Pai (Rice) “Compromising Quality to Stay Relevant” (with Rahul Deb and Matt Mitchell)

We study a novel dynamic principal-agent framework which features adverse selection, moral hazard and no transfers. The model can be described as a bandit problem where the principal chooses between a safe and risky arm. The risky arm's type is known, and output is controlled, by a strategic agent. The principal prefers to pull the risky arm only if the arm is the high type whereas, irrespective of type, the agent wants to maximize the number of times the risky arm is pulled. Our main result shows that when the principal can commit, there are conditions under which the optimal dynamic mechanism induces efficient output from the risky arm. By contrast, in the absence of commitment, inefficient output must arise on path in all equilibria (subject to a mild refinement). We use our model to discuss reputation management by online content providers and by experts in organizations.

Luciano Pomatto (Caltech) “The cost of information acquisition” (with Philipp Strack and Omer Tamuz)

We provide an axiomatic theory of costly information acquisition. In the framework of Blackwell (1951), we pursue a nonparametric approach for determining the cost of an experiment. We consider three axioms of physical information acquisition: 1) The cost of an experiment depends only on its informational content; 2) Performing an experiment with probability 1/2 requires 1/2 the cost; and 3) the cost of observing two independent experiments is the sum of the individual costs. We provide a simple characterization of all continuous cost functions that satisfy the axioms: they are positive linear combinations of Kullback-Leibler divergences. If, as in the recent literature, information acquisition is modeled as the choice of a distribution over posteriors, our theory provides a cost of physical information that, while simple, is technically and conceptually different from the standard approach based on Shannon entropy, and allows to express the idea that some states of the world are intrinsically harder to distinguish than others.

Evan Sadler (Columbia) “False Information and Disagreement in Social Networks”

Lones Smith (Wisconsin) “Accept this paper” (with Andrea Wilson)

Rann Smorodinsky (Technion) “(temporary title) Keeping Up with the Kardashians is Rational” (with Itai Arieli, Gal Bahar and Moshe Tennenholtz)

Why do so many of us idol celebrities? Why do we care what Kim Kardashian wears and drinks, who Robert de-Niro votes for, what are Jane Fonda's opinions on environmental issues and whether Hugh Grant, denounces Trump's cut of funds to UN Palestinian refugee agency. In this paper we argue that celebrities have a valuable role in social learning even if they have no particular merit or prior unique knowledge. Thus, the celebrity phenomenon is a social structure from which we all benefit and so it has clear rational foundations.

Disclaimer: my co-authors are familiar neither with the title nor he abstract. The draft of this work has been circulating under a different title and a different abstract and can be found here.

Eduard Talamàs (UPenn) “No Holdup in Dynamic Markets” (with Matthew Elliott)

Different types of agents make non-contractible investments before bargaining over both who matches to whom and the terms of trade. In thin markets, the holdup problem—that is, underinvestment caused by agents receiving only a fraction of the returns from their investments—is ubiquitous. However, we show that holdup is not a problem in markets that attract traders over time—even when only a few traders are present in the market at any point in time. In particular, we characterize the type-symmetric Markov perfect equilibria of a non-cooperative investment and bargaining game with sequential entry, and we show that—in every such equilibrium—all the agents receive the full returns from their marginal investments in the limit as they become patient. Intuitively, the option to wait for future market participants creates competition—so even apparently-thin markets can be competitive. This provides non-cooperative foundations for the standard price-taking assumption in the literature investigating investment efficiency in competitive matching markets.

Omer Tamuz (Caltech) “Stochastic Dominance Under Independent Noise” (with Luciano Pomatto and Philipp Strack)

Junjie Zhou (NUS) “Coordination on Networks” (with Matt Leister and Yves Zenou)

We study a coordination game among agents on a network, who choose whether or not to take an action that yields value increasing in the actions of neighbors. In a standard global game setting, players receive noisy information of the technology’s common state-dependent value. At the noiseless limit, equilibrium strategies are threshold strategies: each agent adopts if the signal received is above a certain cutoff value. We characterize properties of the cutoffs as a function of the network structure. This characterization allows to partition players into coordination sets, i.e., sets of players where all members take a common cutoff strategy and are path connected. We also show that there is a single coordination set (all players use the same strategies, so they perfectly coordinate) if and only if the network is balanced, i.e., the average degree of each subnetwork is no larger than the average degree of the network. Comparative statics exercises as well as welfare properties are investigated. We show that, in order to maximize aggregate welfare or adoption, the planner needs to target coordination sets and not individuals.