Keeping Up Appearances: Computational Modeling of Face Acts in Persuasion Oriented Discussions

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 7473–7485, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 7473 Keeping Up Appearances: Computational Modeling of Face Acts in Persuasion Oriented Discussions Ritam Dutt Rishabh Joshi Language Technologies Institute Carnegie Mellon University {rdutt, rjoshi2, cprose}@cs.cmu.edu Carolyn Penstein Ros´e Abstract The notion of face refers to the public self- image of an individual that emerges both from the individual’s own actions as well as from the interaction with others. Modeling face and understanding its state changes throughout a conversation is critical to the study of mainte- nance of basic human needs in and through in- teraction. Grounded in the politeness theory of Brown and Levinson (1978), we propose a generalized framework for modeling face acts in persuasion conversations, resulting in a reli- able coding manual, an annotated corpus, and computational models. The framework reveals insights about differences in face act utiliza- tion between asymmetric roles in persuasion conversations. Using computational models, we are able to successfully identify face acts as well as predict a key conversational outcome (e.g. donation success). Finally, we model a la- tent representation of the conversational state to analyze the impact of predicted face acts on the probability of a positive conversational out- come and observe several correlations that cor- roborate previous findings. 1 Introduction Politeness principles, displayed in practice in day- to-day language usage, play a central role in shap- ing human interaction. Formulations of politeness principles are related to basic human needs that are jointly met in and through interaction (Grice et al., 1975; Brown et al., 1987; Leech, 2016). Natural language offers various ways to enact politeness. One of the most influential politeness theories from linguistics is proposed in (Brown and Levinson, 1978), in which a detailed exposition is offered of the individual actions whose cumulative effect re- sults in saving face and losing face, along with a consideration of cost. Using this framework, it is possible to analyze how interlocutors make deci- sions about where and how these devices should be used based on an intricate cost-benefit analysis (Brown et al., 1987). We refer to these component actions here as face acts. The idea of face acts appears quite attractive from a computational standpoint for their potential role in understanding what is “meant” from what is “said” (Grice et al., 1975; Brown et al., 1987; Leech, 2016). Consequently, politeness has been widely researched in various domains of language tech- nologies (Walker et al., 1997; Gupta et al., 2007; Wang et al., 2012; Abdul-Mageed and Diab, 2012; Danescu-Niculescu-Mizil et al., 2013) in addition to foundational work in pragmatics and sociolin- guistics (Brown et al., 1987; Grice et al., 1975; Leech, 2016). However, much prior work modeling politeness reduces the problem to a rating task or binary prediction task, separating polite and impo- lite behavior. Consequently, what the models end up learning is mainly overt markers of politeness or rudeness, rather than the underlying indirect strate- gies for achieving politeness or rudeness …

Content truncated to 500 words