Bayesian persuasion

Last updated

In economics and game theory, Bayesian persuasion involves a situation where one participant (the sender) wants to persuade the other (the receiver) of a certain course of action. There is an unknown state of the world, and the sender must commit to a decision of what information to disclose to the receiver. Upon seeing said information, the receiver will revise their belief about the state of the world using Bayes' Rule and select an action. Bayesian persuasion was introduced by Kamenica and Gentzkow, [1] though its origins can be traced back to Aumann and Maschler (1995).

Contents

Bayesian persuasion is a special case of a principal–agent problem: the principal is the sender and the agent is the receiver. It can also be seen as a communication protocol, comparable to signaling games; [2] the sender must decide what signal to reveal to the receiver to maximize their expected utility. It can also be seen as a form of cheap talk. [3]

Example

Consider the following illustrative example. There is a medicine company (sender), and a medical regulator (receiver). The company produces a new medicine, and needs the approval of the regulator. There are two possible states of the world: the medicine can be either "good" or "bad". The company and the regulator do not know the true state. However, the company can run an experiment and report the results to the regulator. The question is what experiment the company should run in order to get the best outcome for themselves. The assumptions are:


For example, suppose the prior probability that the medicine is good is 1/3 and that the company has a choice of three actions:

  1. Conduct a thorough experiment that always detects whether the medicine is good or bad, and truthfully report the results to the regulator. In this case, the regulator will approve the medicine with probability 1/3, so the expected utility of the company is 1/3.
  2. Don't conduct any experiment; always say "the medicine is good". In this case, the signal does not give any information to the regulator. As the regulator believes that the medicine is good with probability 1/3, the expectation-maximizing action is to always reject it. Therefore, the expected utility of the company is 0.
  3. Conduct an experiment that, if the medicine is good, always reports "good", and if the medicine is bad, it reports "good" or "bad" with probability 1/2. Here, the regulator applies Bayes' rule: given a signal "good", the probability that the medicine is good is 1/2, so the regulator approves it. Given a signal "bad", the probability that the medicine is good is 0, so the regulator rejects it. All in all, the regulator approves the medicine in 2/3 of the cases, so the expected utility of the company is 2/3.

In this case, the third policy is optimal for the sender since this has the highest expected utility of the available options. Using the Bayes rule, the sender has persuaded the receiver to act in a favorable way to the sender.

Generalized model

The basic model has been generalized in a number of ways, including:

Practical application

The applicability of the model has been assessed in a number of real-world contexts:

Computational approach

Algorithmic techniques have been developed to compute the optimal signalling scheme in practice. This can be found in polynomial time with respect to the number of actions and pseudo-polynomial time with respect to the number of states of the world. [3] Algorithms with lower computational complexity are also possible under stronger assumptions.

The online case, where multiple signals are sent over time, can be solved efficiently as a regret minimization problem. [17]

Related Research Articles

<span class="mw-page-title-main">Statistical inference</span> Process of using data analysis

Statistical inference is the process of using data analysis to infer properties of an underlying distribution of probability. Inferential statistical analysis infers properties of a population, for example by testing hypotheses and deriving estimates. It is assumed that the observed data set is sampled from a larger population.

<span class="mw-page-title-main">Behavioral economics</span> Academic discipline

Behavioral economics is the study of the psychological and cognitive factors involved in the decisions of individuals or institutions, and how these decisions deviate from those implied by traditional economic theory.

From a legal point of view, a contract is an institutional arrangement for the way in which resources flow, which defines the various relationships between the parties to a transaction or limits the rights and obligations of the parties.

<span class="mw-page-title-main">Information asymmetry</span> Concept in contract theory and economics

In contract theory, mechanism design, and economics, an information asymmetry is a situation where one party has more or better information than the other.

Network congestion in data networking and queueing theory is the reduced quality of service that occurs when a network node or link is carrying more data than it can handle. Typical effects include queueing delay, packet loss or the blocking of new connections. A consequence of congestion is that an incremental increase in offered load leads either only to a small increase or even a decrease in network throughput.

<span class="mw-page-title-main">Loss function</span> Mathematical relation assigning a probability event to a cost

In mathematical optimization and decision theory, a loss function or cost function is a function that maps an event or values of one or more variables onto a real number intuitively representing some "cost" associated with the event. An optimization problem seeks to minimize a loss function. An objective function is either a loss function or its opposite, in which case it is to be maximized. The loss function could include terms from several levels of the hierarchy.

<span class="mw-page-title-main">Decision theory</span> Branch of applied probability theory

Decision theory or the theory of rational choice is a branch of probability, economics, and analytic philosophy that uses the tools of expected utility and probability to model how individuals should behave rationally under uncertainty. It differs from the cognitive and behavioral sciences in that it is prescriptive and concerned with identifying optimal decisions for a rational agent, rather than describing how people really do make decisions. Despite this, the field is extremely important to the study of real human behavior by social scientists, as it lays the foundations for the rational agent models used to mathematically model and analyze individuals in fields such as sociology, economics, criminology, cognitive science, and political science.

The expected utility hypothesis is a foundational assumption in mathematical economics concerning decision making under uncertainty. It postulates that rational agents maximize utility, meaning the subjective desirability of their actions. Rational choice theory, a cornerstone of microeconomics, builds this postulate to model aggregate social behaviour.

<span class="mw-page-title-main">Paul Milgrom</span> Economist and winner of the 2020 Nobel Prize in Economics

Paul Robert Milgrom is an American economist. He is the Shirley and Leonard Ely Professor of Humanities and Sciences at the Stanford University School of Humanities and Sciences, a position he has held since 1987. He is a professor in the Stanford School of Engineering as well and a Senior Fellow at the Stanford Institute for Economic Research. Milgrom is an expert in game theory, specifically auction theory and pricing strategies. He is the winner of the 2020 Nobel Memorial Prize in Economic Sciences, together with Robert B. Wilson, "for improvements to auction theory and inventions of new auction formats".

In game theory, cheap talk is communication between players that does not directly affect the payoffs of the game. Providing and receiving information is free. This is in contrast to signalling, in which sending certain messages may be costly for the sender depending on the state of the world.

<span class="mw-page-title-main">Signaling game</span> Game class in game theory

In game theory, a signaling game is a simple type of a dynamic Bayesian game.

In game theory, a Perfect Bayesian Equilibrium (PBE) is a solution with Bayesian probability to a turn-based game with incomplete information. More specifically, it is an equilibrium concept that uses Bayesian updating to describe player behavior in dynamic games with incomplete information. Perfect Bayesian equilibria are used to solve the outcome of games where players take turns but are unsure of the "type" of their opponent, which occurs when players don't know their opponent's preference between individual moves. A classic example of a dynamic game with types is a war game where the player is unsure whether their opponent is a risk-taking "hawk" type or a pacifistic "dove" type. Perfect Bayesian Equilibria are a refinement of Bayesian Nash equilibrium (BNE), which is a solution concept with Bayesian probability for non-turn-based games.

Bayesian experimental design provides a general probability-theoretical framework from which other theories on experimental design can be derived. It is based on Bayesian inference to interpret the observations/data acquired during the experiment. This allows accounting for both any prior knowledge on the parameters to be determined as well as uncertainties in observations.

In game theory, a Bayesian game is a strategic decision-making model which assumes players have incomplete information. Players may hold private information relevant to the game, meaning that the payoffs are not common knowledge. Bayesian games model the outcome of player interactions using aspects of Bayesian probability. They are notable because they allowed, for the first time in game theory, for the specification of the solutions to games with incomplete information.

In contract theory, signalling is the idea that one party credibly conveys some information about itself to another party.

The intuitive criterion is a technique for equilibrium refinement in signaling games. It aims to reduce possible outcome scenarios by restricting the possible sender types to types who could obtain higher utility levels by deviating to off-the-equilibrium messages, and to types for which the off-the-equilibrium message is not equilibrium dominated.

Mathematical models of social learning aim to model opinion dynamics in social networks. Consider a social network in which people (agents) hold a belief or opinion about the state of something in the world, such as the quality of a particular product, the effectiveness of a public policy, or the reliability of a news agency. In all these settings, people learn about the state of the world via observation or communication with others. Models of social learning try to formalize these interactions to describe how agents process the information received from their friends in the social network. Some of the main questions asked in the literature include:

  1. whether agents reach a consensus;
  2. whether social learning effectively aggregates scattered information, or put differently, whether the consensus belief matches the true state of the world or not;
  3. how effective media sources, politicians, and prominent agents can be in belief formation of the entire network. In other words, how much room is there for belief manipulation and misinformation?

Various experiments have been made to evaluate various procedures for fair division, the problem of dividing resources among several people. These include case studies, computerized simulations, and lab experiments.

In game theory, a Bayes correlated equilibrium is a solution concept for static games of incomplete information. It is both a generalization of the correlated equilibrium perfect information solution concept to bayesian games, and also a broader solution concept than the usual Bayesian Nash equilibrium thereof. Additionally, it can be seen as a generalized multi-player solution of the Bayesian persuasion information design problem.

References

  1. ↑ Kamenica, Emir; Gentzkow, Matthew (2011-10-01). "Bayesian Persuasion". American Economic Review. 101 (6): 2590–2615. doi:10.1257/aer.101.6.2590. ISSN   0002-8282.
  2. ↑ Kamenica, Emir (2019-05-13). "Bayesian Persuasion and Information Design". Annual Review of Economics. 11: 249–272. doi:10.1146/annurev-economics-080218-025739.
  3. 1 2 Dughmi, Shaddin; Xu, Haifeng (June 2016). "Algorithmic Bayesian persuasion". Proceedings of the forty-eighth annual ACM symposium on Theory of Computing. pp. 412–425. arXiv: 1503.05988 . doi:10.1145/2897518.2897583. ISBN   978-1-4503-4132-5.
  4. ↑ Hedlund, Jonas (2017-01-01). "Bayesian persuasion by a privately informed sender". Journal of Economic Theory. 167: 229–268. doi:10.1016/j.jet.2016.11.003.
  5. ↑ Kolotilin, Anton (2018-05-29). "Optimal information disclosure: A linear programming approach". Theoretical Economics. 13 (2): 607–635. doi:10.3982/TE1805. hdl: 10419/197158 .
  6. ↑ Rayo, Luis; Segal, Ilya (2010-10-01). "Optimal Information Disclosure". Journal of Political Economy. 118 (5): 949–987. doi:10.1086/657922.
  7. ↑ Camara, Modibo K.; Hartline, Jason D.; Johnsen, Aleck (2020-11-01). "Mechanisms for a No-Regret Agent: Beyond the Common Prior". 2020 IEEE 61st Annual Symposium on Foundations of Computer Science (FOCS). IEEE. pp. 259–270. arXiv: 2009.05518 . doi:10.1109/focs46700.2020.00033. ISBN   978-1-7281-9621-3.
  8. ↑ Gentzkow, Matthew; Kamenica, Emir (2016-10-18). "Competition in Persuasion". The Review of Economic Studies. 84: 300–322. doi:10.1093/restud/rdw052.
  9. ↑ Gentzkow, Matthew; Shapiro, Jesse M. (2008). "Competition and Trust in the Market for News". Journal of Economic Perspectives. 22 (2): 133–154. doi:10.1257/jep.22.2.133.
  10. ↑ Li, Fei; Norman, Peter (2021). "Sequential Persuasion". Theoretical Economics. 16 (2): 639–675. doi: 10.3982/TE3474 .
  11. ↑ Bergemann, Dirk; Morris, Stephen (2019-03-01). "Information Design: A Unified Perspective". Journal of Economic Literature. 57: 44–95. doi:10.1257/jel.20181489.
  12. ↑ Ely, Jeffrey C. (January 2017). "Beeps". American Economic Review. 107 (1): 31–53. doi:10.1257/aer.20150218.
  13. ↑ Goldstein, Itay; Leitner, Yaron (September 2018). "Stress tests and information disclosure". Journal of Economic Theory. 177: 34–69. doi:10.1016/j.jet.2018.05.013.
  14. ↑ Boleslavsky, Raphael; Cotton, Christopher (May 2015). "Grading Standards and Education Quality". American Economic Journal: Microeconomics. 7 (2): 248–279. doi:10.1257/mic.20130080.
  15. ↑ Habibi, Amir (January 2020). "Motivation and information design". Journal of Economic Behavior & Organization. 169: 1–18. doi:10.1016/j.jebo.2019.10.015.
  16. ↑ Ely, Jeffrey; Frankel, Alexander; Kamenica, Emir (February 2015). "Suspense and Surprise". Journal of Political Economy. 123: 215–260. doi:10.1086/677350.
  17. ↑ Bernasconi, Martino; Castiglioni, Matteo (2023). "Optimal Rates and Efficient Algorithms for Online Bayesian Persuasion". Proceedings of Machine Learning Research. 202: 2164–2183. arXiv: 2303.01296 .
This page is based on this Wikipedia article
Text is available under the CC BY-SA 4.0 license; additional terms may apply.
Images, videos and audio are available under their respective licenses.