Social Bias Frames: Reasoning about Social and Power Implications of Language

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 5477–5490 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 5477 SOCIAL BIAS FRAMES: Reasoning about Social and Power Implications of Language Maarten Sap† Saadia Gabriel†‡ Lianhui Qin†‡ Dan Jurafsky⋄ Noah A. Smith†‡ Yejin Choi†‡ †Paul G. Allen School of Computer Science & Engineering, University of Washington ‡Allen Institute for Artificial Intelligence ⋄Linguistics & Computer Science Departments, Stanford University Abstract Warning: this paper contains content that may be offensive or upsetting. Language has the power to reinforce stereo- types and project social biases onto others. At the core of the challenge is that it is rarely what is stated explicitly, but rather the im- plied meanings, that frame people’s judgments about others. For example, given a statement that “we shouldn’t lower our standards to hire more women,” most listeners will infer the implicature intended by the speaker — that “women (candidates) are less qualified.” Most semantic formalisms, to date, do not capture such pragmatic implications in which people express social biases and power differentials in language. We introduce SOCIAL BIAS FRAMES, a new conceptual formalism that aims to model the pragmatic frames in which people project so- cial biases and stereotypes onto others. In ad- dition, we introduce the Social Bias Inference Corpus to support large-scale modelling and evaluation with 150k structured annotations of social media posts, covering over 34k implica- tions about a thousand demographic groups. We then establish baseline approaches that learn to recover SOCIAL BIAS FRAMES from unstructured text. We find that while state- of-the-art neural models are effective at high- level categorization of whether a given state- ment projects unwanted social bias (80% F1), they are not effective at spelling out more de- tailed explanations in terms of SOCIAL BIAS FRAMES. Our study motivates future work that combines structured pragmatic inference with commonsense reasoning on social impli- cations. 1 Introduction Language has enormous power to project social biases and reinforce stereotypes on people (Fiske, What do you call a movie with an all- Muslim cast? A box office bomb. We shouldn’t lower our standards just to hire more women. likely yes no Women Implies women are less qualified offensive? lewd? intentional? yes yes no offensive? lewd? intentional? yes group referenced? which group implied statement yes group referenced? which group Muslims Implies Muslims are all terrorists implied statement Figure 1: Understanding and explaining why an ar- guably innocuous statement is potentially unjust re- quires reasoning about conversational implicatures and commonsense implications with respect to the under- lying intent, offensiveness, and power differentials be- tween different social groups. SOCIAL BIAS FRAMES aim to represent the various pragmatic meanings re- lated to social bias implications, by combining cat- egorical and free-text annotations, e.g., that “women are less qualified” is implied by the statement “we shouldn’t lower our standards to hire more women.” 1993). The way such biases are projected is rarely in what is stated explicitly, but in all the implied layers of meanings that …

Content truncated to 500 words