Modeling Non-Cooperative Dialogue: Theoretical and Empirical Insights

Modeling Non-Cooperative Dialogue: Theoretical and Empirical Insights Anthony Sicilia Intelligent Systems Program, University of Pittsburgh, Pittsburgh, USA anthonysicilia@pitt.edu Tristan Maidment Intelligent Systems Program, University of Pittsburgh, Pittsburgh, USA tristan.maidment@pitt.edu Pat Healy Department of Informatics and Networked Systems, University of Pittsburgh, Pittsburgh, USA pat.healy@pitt.edu Malihe Alikhani Department of Computer Science and Intelligent Systems Program, University of Pittsburgh, Pittsburgh, USA malihe@pitt.edu Abstract Investigating cooperativity of interlocutors is central in studying pragmatics of dialogue. Models of conversation that only assume co- operative agents fail to explain the dynamics of strategic conversations. Thus, we investigate the ability of agents to identify non-cooperative interlocutors while completing a concurrent visual-dialogue task. Within this novel setting, we study the optimality of communication strategies for achieving this multi-task ob- jective. We use the tools of learning theory to develop a theoretical model for identify- ing non-cooperative interlocutors and apply this theory to analyze different communica- tion strategies. We also introduce a corpus of non-cooperative conversations about im- ages in the GuessWhat?! dataset proposed by De Vries et al. (2017). We use reinforcement learning to implement multiple communica- tion strategies in this context and find that empirical results validate our theory. 1 Introduction A robust dialogue agent cannot always assume a cooperative conversational counterpart when de- ployed in the wild. Even in goal-oriented settings, where the intent of an interlocutor may seem to be granted, bad actors and disinterested parties are free to interact with our dialogue systems. These non-cooperative interlocutors add harmful noise to data, which can elicit unexpected behaviors from our dialogue systems. Thus, the need to study non- cooperation increases daily as we build and deploy conversational systems which interact with peo- ple from different demographics, political views, and intents, continuously learning from the col- lected data. Examples include Amazon Alexa, task-oriented systems that help patients recover- ing from injuries or can teach a person a new language, and systems that help predict deceptive behaviors in courtrooms. To effectively commu- nicate in the presence of unwanted behaviors like bullying (Cercas Curry and Rieser, 2018), sys- tems need to understand users’ strategic behaviors (Asher and Lascarides, 2013) and be able to iden- tify non-cooperative actions. Designing agents that learn to identify non-cooperative interlocu- tors is challenging since it requires processing the context of the dialogue in addition to mod- eling the choices that interlocutors make under uncertainty—choices which typically affect their ability to complete tasks unrelated to identifying non-cooperation as well. In light of this, we ask: What communication strategies are ef- fective for identifying non-cooperative interlocutors, while also achieving the goals of a distinct dialogue task? To answer this question, we appeal to a sim- ple non-cooperative version of the visual dialogue game Guess What?! (De Vries et al., 2017). See Figure 1 for an example. The game consists of a multi-round dialogue between two players: a question-player and an answer-player. Both have access to the same image whereas only the answer- player has access to an image-secret; that is, a 1084 Transactions of the Association …

Content truncated to 500 words