Will I Sound Like Me? Improving Persona Consistency in Dialogues through Pragmatic Self-Consciousness

We explore the task of improving persona consistency of dialogue agents. Recent models tackling consistency often train with additional Natural Language Inference (NLI) labels or attach trained extra modules to the generative agent for maintaining consistency. However, such additional labels and training can be demanding. Also, we find even the bestperforming persona-based agents are insensitive to contradictory words. Inspired by social cognition and pragmatics, we endow existing dialogue agents with public self-consciousness on the fly through an imaginary listener. Our approach, based on the Rational Speech Acts framework (Frank and Goodman, 2012), can enforce dialogue agents to refrain from uttering contradiction. We further extend the framework by learning the distractor selection, which has been usually done manually or randomly. Results on Dialogue NLI (Welleck et al., 2019) and PersonaChat (Zhang et al., 2018) dataset show that our approach reduces contradiction and improves consistency of existing dialogue models. Moreover, we show that it can be generalized to improve contextconsistency beyond persona in dialogues.

Introduction. In the study of dialogue agents, consistency has been a long-standing issue. To resolve this, much research has been conducted to endow dialogue agents with personas. Li et al. (2016) propose to encode persona in embeddings and Zhang et al. (2018) introduce a persona-conditioned dialogue dataset. On top of these works, many efforts have been made to improve consistency. In spite of such recent significant progress, there is much room for improving persona-based dialogue agents. We observe that even the best performing persona-based generative models (See et al., 2019; Wolf et al., 2019b; Roller et al., 2020) are highly insensitive to contradictory words, and thus fail to deliver consistent persona to the interlocutor (Figure 1). Also, extra modules other than the generative model is often required for improving consistency. Recent works on consistency in persona-based dialogue actively adopt the NLIbased approach (Welleck et al., 2019; Song et al., 2019; Li et al., 2020; Song et al., 2020), which have the following prerequisites.

Discussion / Conclusion. This work investigated how modeling public selfconsciousness can help dialogue agents improve persona-consistency. We showed existing dialogue agents are highly insensitive to contradiction, and introduced an orthogonally applicable method using the RSA framework (Frank and Goodman, 2012) to alleviate the issue. We also designed a learning method for distractor selection, named Distractor Memory and proposed a better update for the listener’s world prior. Furthermore, we demonstrated how our approach can be generalized to improve dialogue context-consistency. Our self-conscious agents improved the base agents on the Dialogue NLI (Welleck et al., 2019) and PersonaChat (Zhang et al., 2018) dataset, without consistency labels and NLI models. An important future direction will be generating the distractors and learning the rationality coefficients.