"Hiding in Plain Sight": Designing Synthetic Dialog Generation for Uncovering Socially Situated Norms

Naturally situated conversations capture the underlying social norms appropriate for the topic of conversation, the relationship between interlocutors and their communicative intent. This paper proposes a framework for controlled generation of dialogues, spanning a wide range of interlocutors attributes (such as age group, profession and personality types), relationship types, conversation topics and conversational trajectories. We use this framework to generate NormHint, a collection of dialogues consistent with these rich settings and analyzed for norm violation leading to conflicts, and potential steps for avoiding these conflicts by adhering to social norms and preferring respectful utterances maintaining the communicative intents of the original utterance. We present the results of human validation and automated analysis of NormHint and show it captures a wide range of conversational topics and scored highly by humans for the naturalness of the conversations based on the prompted context.

"We Demand Justice!": Towards Grounding Political Text in Social Context

Social media discourse from US politicians frequently consists of 'seemingly similar language used by opposing sides of the political spectrum'. But often, it translates to starkly contrasting real-world actions. For instance, "We need to keep our students safe from mass shootings" may signal either "arming teachers to stop the shooter" or "banning guns to reduce mass shootings" depending on who says it and their political stance on the issue. In this paper, we define and characterize the context that is required to fully understand such ambiguous statements in a computational setting and ground them in real-world entities, actions, and attitudes. To that end, we propose two challenging datasets that require an understanding of the real-world context of the text to be solved effectively. We benchmark these datasets against baselines built upon large pre-trained models such as BERT, RoBERTa, GPT-3, etc. Additionally, we develop and benchmark more structured baselines building upon existing 'Discourse Contextualization Framework' and 'Political Actor Representation' models. We perform analysis of the datasets and baseline predictions to obtain further insights into the pragmatic language understanding challenges posed by the proposed social grounding tasks.