Designing an Automatic Agent for Repeated Language–based Persuasion Games

Designing an Automatic Agent for Repeated Language–based Persuasion Games Maya Raifer, Guy Rotman, Reut Apel, Moshe Tennenholtz, Roi Reichart Technion—Israel Institute of Technology, Israel {mayatarno, grotman, reutapel}@campus.technion.ac.il {roiri, moshet}@technion.ac.il Abstract Persuasion games are fundamental in eco- nomics and AI research and serve as the basis for important applications. However, work on this setup assumes communication with styl- ized messages that do not consist of rich human language. In this paper we consider a repeated sender (expert) – receiver (decision maker) game, where the sender is fully informed about the state of the world and aims to persuade the receiver to accept a deal by sending one of several possible natural language reviews. We design an automatic expert that plays this re- peated game, aiming to achieve the maximal payoff. Our expert is implemented within the Monte Carlo Tree Search (MCTS) algorithm, with deep learning models that exploit behav- ioral and linguistic signals in order to predict the next action of the decision maker, and the future payoff of the expert given the state of the game and a candidate review. We demonstrate the superiority of our expert over strong base- lines and its adaptability to different decision makers and potential proposed deals.1 1 Introduction Natural Language Processing (NLP) has made substantial progress in recent years, excelling on text understanding applications such as machine trans- lation (Bahdanau etal., 2015;Johnson etal., 2017), information extraction (Stanovsky et al., 2018), and question answering (Andreas et al., 2016; Kwiatkowski et al., 2019). However, these appli- cations do not assume that language is used for interaction between strategic participants whose objectives overlap only partially. In contrast, in the fields of economics and artificial intelligence (AI), such setups have been widely explored. For example, the settings of person- alized advertising and targeted recommendation systems (Shapiro and Varian, 1998; Emek et al., 1Our code and data are available at: https://github .com/mayaraifer/automaticagent. 2014; Bahar et al., 2016) suggest personalized services for their customers, and solutions are formed as strategic sender–receiver interactions (Arieli and Babichenko, 2019). However, this work assumes stylized messaging that does not involve real-world natural language. In this paper we address the setting of sender– receiver interaction, but, in contrast to previous research, we assume natural language interaction between the players in an iterative non zero-sum persuasion game. In our setting the two partici- pants are strategic players with their own private utilities. Crucially, the sender has more informa- tion about the world than the receiver does. Taking the NLP perspective, we are particularly interested in the persuasion game setting, where the sender’s objective is to persuade the receiver, using natural language messages, to select an action from a set of alternatives. The receiver, in turn, has different payoffs for the different actions. The receiver’s payoff depends on properties of the setup that are unavailable to her, and she has a higher level of uncertainty about the setup than the sender has. Our focus is on repeated non-cooperative se- tups, where the utilities of the players do …

Content truncated to 500 words