Observational learning in an uncertain world
01 Dec 2010-pp 6645-6650
TL;DR: It is shown that information is correctly aggregated when preferences of different types are closely aligned, and even though learning is guaranteed to be incomplete ex ante, there are sample paths over which agents become certain about the underlying state of the world.
Abstract: We study a model of observational learning in social networks in the presence of uncertainty about agents' type distributions. Each individual receives a private noisy signal about a payoff-relevant state of the world, and can observe the actions of other agents who have made a decision before her. We assume that agents do not observe the signals and types of others in the society, and are also uncertain about the type distributions. We show that information is correctly aggregated when preferences of different types are closely aligned. On the other hand, if there is sufficient heterogeneity in preferences, uncertainty about type distributions leads to potential identification problems, preventing asymptotic learning. We also show that even though learning is guaranteed to be incomplete ex ante, there are sample paths over which agents become certain about the underlying state of the world.
Citations
More filters
Posted Content•
TL;DR: In this paper, the authors study diffusion social learning over weakly-connected graphs and show that the asymmetric flow of information hinders the learning abilities of certain agents regardless of their local observations.
Abstract: In this paper, we study diffusion social learning over weakly-connected graphs. We show that the asymmetric flow of information hinders the learning abilities of certain agents regardless of their local observations. Under some circumstances that we clarify in this work, a scenario of total influence (or "mind-control") arises where a set of influential agents ends up shaping the beliefs of non-influential agents. We derive useful closed-form expressions that characterize this influence, and which can be used to motivate design problems to control it. We provide simulation examples to illustrate the results.
9 citations
Dissertation•
10 Jul 2018
TL;DR: In this article, the authors present theory and empirical evidence for two distinct aspects of human behavior: social learning and motivated beliefs, and they develop a simple theory to revisit the classical social learning models by challenging the assumption of freely available information.
Abstract: This thesis contains four chapters presenting theory and empirical evidence for two distinct aspects of human behaviour: social learning and motivated beliefs. I develop a simple theory to revisit the classical social learning models by challenging the assumption of freely available information. My model suggests that when it is costly to acquire information, social learning (herding) is prevalent, and people do not have incentives to acquire private information (e.g. to form their own judgements). Classical information cascade models suggest that although herding is observed, information aggregation is still possible with communication channels (e.g. a survey); however, my model indicates that information aggregation is unattainable because people in the herd do not acquire private information. We then test my model in a laboratory and find that, as predicted, subjects can learn from others successfully. Also, individual heterogeneity exists in: there are herd animals biased against private information, lone wolves who are biased toward it and subjects who behave optimally. In aggregate, there is no overall bias for or against private information. We also document a new cognitive bias involved in processing social information. Individual characteristics, especially the cognitive ability, seems to be a very good indicator of subjects’ behaviour. Subjects with higher cognitive scores choose optimal information more frequently and follow information more frequently. Overconfidence can be driven by the consumption motive (e.g. savouring future payoff/self-image) and the instrumental motive (e.g. being optimistic about the outcome of effort for motivation). I develop a simple model incorporating these two motives and suggest that individuals hold a dynamic pattern of overconfidence. Then I conduct an online field experiment with students to test the theory. The experimental findings indicate that students are likely to adopt overconfident beliefs as a commitment device to deal with their self-control problem. However, I do not find evidence for the consumption motive of overconfidence.
2 citations
Cites background or methods from "Observational learning in an uncert..."
...This is a contradiction to Acemoglu et al. (2011) where asymptotic learning is possible with exogenous observation structure....
[...]
...Acemoglu et al. (2010) also model social learning with heterogeneous agents....
[...]
...In Acemoglu et al. (2010), information may fail to aggregate because agents cannot identify the history efficiently with the presence of type distribution uncertainty; while in classical information cascade models, information aggregation fails because the private signal is bounded (not informative…...
[...]
...Smith & Sørensen (2000); Banerjee & Fudenberg (2004); Acemoglu et al. (2010); Guarino et al. (2011); Guarino & Jehiel (2013). For experimental papers, see for example, Anderson & Holt (1997), Hung & Plott (2001) and Weizsäcker (2010)....
[...]
...The failure of information aggregation in Acemoglu et al. (2010) is fundamentally different from classical information cascade models by Bikhchandani et al. (1992) and Banerjee (1992)....
[...]
26 Jun 2011
TL;DR: The rule of weight adjustment is designed and it is testified that the updating rule with weight adjustment ensures learning on the whole social network.
Abstract: Almost all existing social learning models assume that there is only one type of agents in the society in order to avoid identification problem. In this work, we assume that there are various types of agents according to the communities they locate in. We design the rule of weight adjustment and testify that the updating rule with weight adjustment ensures learning on the whole social network. Furthermore, we show that how convergence speed is influenced by two updating-relevant parameters, and present instruction on how to attain the optimal social learning efficiency.
1 citations
TL;DR: A multi-armed bandit algorithm is proposed and used to train the weakly connected agents to converge to the most stable state over time, and the sublinearity of the regret bound for the proposed algorithm is compared to the sub linearity for the state-of-the-art algorithm for strongly connected networks.
Abstract: This article provides a study into the social network where influential personalities collaborate positively among themselves to learn an underlying truth over time, but may have misled their followers to believe a false information. Most existing work that study leader-follower relationships in a social network model the social network as a graph network, and apply non-Bayesian learning to train the weakly connected agents to learn the truth. Although this approach is popular, it has the limitation of assuming that the truth - otherwise called the true state - is time-invariant. This is not practical in social network, where streams of information are released and updated every second, making the true state arbitrarily time-varying. Thus, this article improves on existing work by introducing online reinforcement learning into the graph theoretic framework. Specifically, multi-armed bandit technique is applied. A multi-armed bandit algorithm is proposed and used to train the weakly connected agents to converge to the most stable state over time. The speed of convergence for these weakly connected agents trained with the proposed algorithm is slower by 66% on average, when compared to the speed of convergence for strongly connected agents trained with the state-of-the-art algorithm. This is because weakly connected agents are difficult to train. However, the speed of convergence of these weakly connected agents can be improved by approximately 50% on average, by fine-tuning the learning rate of the proposed algorithm. The sublinearity of the regret bound for the proposed algorithm is compared to the sublinearity of the regret bound for the state-of-the-art algorithm for strongly connected networks.
20 Oct 2020
TL;DR: In this article, a multi-armed bandit algorithm is proposed for weakly connected agents to predict the time-varying true state in a social network, where influential personalities collaborate positively among themselves to learn an underlying truth over time but may have misled their followers to believe a false information.
Abstract: This paper provides a study into the social network where influential personalities collaborate positively among themselves to learn an underlying truth over time but may have misled their followers to believe a false information. Most existing work models the social network as a graph network and applies non-Bayesian learning to understand the behavior of agents in the network. Although this approach is popular, it has the limitation of assuming that the truth - otherwise called the true state - is time-invariant. This is not practical in social network where streams of information are released and updated every second. Thus, this paper improves on existing work by introducing online reinforcement learning into the graph theoretic framework. Specifically, multi-armed bandit technique is applied. A multi-armed bandit algorithm is proposed for weakly connected agents to predict the time-varying true state. The result shows that the weakly connected agents can predict this time-varying true state, howbeit with a higher regret than the strongly connected agents.
References
More filters
TL;DR: In this article, the authors analyze a sequential decision model in which each decision maker looks at the decisions made by previous decision makers in taking her own decision, and they show that the decision rules that are chosen by optimizing individuals will be characterized by herd behavior.
Abstract: We analyze a sequential decision model in which each decision maker looks at the decisions made by previous decision makers in taking her own decision. This is rational for her because these other decision makers may have some information that is important for her. We then show that the decision rules that are chosen by optimizing individuals will be characterized by herd behavior; i.e., people will be doing what others are doing rather than using their information. We then show that the resulting equilibrium is inefficient.
5,956 citations
"Observational learning in an uncert..." refers background or methods in this paper
...The failure of information aggregation in environments with type uncertainties is different from the herding behavior observed by Banerjee [2] and Bikhchandani et al. [3]....
[...]
...The observational learning framework was independently introduced by Banerjee [2] and Bikhchandani, Hirshleifer, This research is supported in parts by the following grants: AFOSR grant FA9550-09-1-0420, ARO grant W911NF-09-1-0556, NSF grant SES0729361....
[...]
...The observational learning framework was independently introduced by Banerjee [2] and Bikhchandani, Hirshleifer,...
[...]
...With bounded private beliefs, once social belief qn passes a certain threshold, all agents discard their private signals; leading to the herding behavior observed by Banerjee [2] and Bikhchandani et al. [3]....
[...]
...With bounded private beliefs, once social belief qn passes a certain threshold, all agents discard their private signals; leading to the herding behavior observed by Banerjee [2] and Bikhchandani et al....
[...]
Posted Content•
TL;DR: It is argued that localized conformity of behavior and the fragility of mass behaviors can be explained by informational cascades.
Abstract: An informational cascade occurs when it is optimal for an individual, having observed the actions of those ahead of him, to follow the behavior of the preceding individual without regard to his own information. We argue that localized conformity of behavior and the fragility of mass behaviors can be explained by informational cascades.
5,412 citations
"Observational learning in an uncert..." refers background in this paper
...[3] and [5]), as ex post and ex ante asymptotic learnings coincide....
[...]
Book•
01 Jan 1990TL;DR: In this paper, a comprehensive introduction to probability theory covering laws of large numbers, central limit theorem, random walks, martingales, Markov chains, ergodic theorems, and Brownian motion is presented.
Abstract: This book is an introduction to probability theory covering laws of large numbers, central limit theorems, random walks, martingales, Markov chains, ergodic theorems, and Brownian motion. It is a comprehensive treatment concentrating on the results that are the most useful for applications. Its philosophy is that the best way to learn probability is to see it in action, so there are 200 examples and 450 problems.
5,168 citations
TL;DR: In this paper, the authors argue that localized conformity of behavior and the fragility of mass behaviors can be explained by informational cascades, where an individual, having observed the actions of those ahead of him, to follow the behavior of the preceding individual without regard to his own information.
Abstract: An informational cascade occurs when it is optimal for an individual, having observed the actions of those ahead of him, to follow the behavior of the preceding individual without regard to his own information. We argue that localized conformity of behavior and the fragility of mass behaviors can be explained by informational cascades.
4,731 citations