David I. Beaver

Venkata S Govindarajan, Katherine Atwell, Barea Sinno et al.

Current studies of bias in NLP rely mainly on identifying (unwanted or negative) bias towards a specific demographic group. While this has led to progress recognizing and mitigating negative bias, and having a clear notion of the targeted group is necessary, it is not always practical. In this work we extrapolate to a broader notion of bias, rooted in social science and psychology literature. We move towards predicting interpersonal group relationship (IGR) - modeling the relationship between the speaker and the target in an utterance - using fine-grained interpersonal emotions as an anchor. We build and release a dataset of English tweets by US Congress members annotated for interpersonal emotion -- the first of its kind, and 'found supervision' for IGR labels; our analyses show that subtle emotional signals are indicative of different biases. While humans can perform better than chance at identifying IGR given an utterance, we show that neural models perform much better; furthermore, a shared encoding between IGR and interpersonal perceived emotion enabled performance gains in both tasks. Data and code for this paper are available at https://github.com/venkatasg/interpersonal-bias

Anshun Asher Zheng, Kanishka Misra, David I. Beaver et al.

We introduce HERO'S JOURNEY, a benchmark for rule induction in goal-directed episodic tasks, where agents must infer hidden rules from demonstrations and act on them through multi-step execution. HERO'S JOURNEY covers eight tasks across attribute and procedural induction families, each with four structural rule forms, controllable lexical grounding, and identifiability conditions. Evaluating state-of-the-art LLMs, we find that models show evidence of rule induction, but the ability is limited and uneven across tasks. Meanwhile, process execution adds an execution bottleneck for models, whereas surface semantics has minimal effect. Induction-specific steering methods improve performance on attribute tasks but show no reliable gains on procedural tasks, suggesting the gap in procedural induction remains an open challenge.

Venkata S Govindarajan, Kyle Mahowald, David I. Beaver et al.

While existing work on studying bias in NLP focues on negative or pejorative language use, Govindarajan et al. (2023) offer a revised framing of bias in terms of intergroup social context, and its effects on language behavior. In this paper, we investigate if two pragmatic features (specificity and affect) systematically vary in different intergroup contexts -- thus connecting this new framing of bias to language output. Preliminary analysis finds modest correlations between specificity and affect of tweets with supervised intergroup relationship (IGR) labels. Counterfactual probing further reveals that while neural models finetuned for predicting IGR labels reliably use affect in classification, the model's usage of specificity is inconclusive. Code and data can be found at: https://github.com/venkatasg/intergroup-probing

Anshun Asher Zheng, Junyi Jessy Li, David I. Beaver

Language is often used strategically, particularly in high-stakes, adversarial settings, yet most work on pragmatics and LLMs centers on cooperativity. This leaves a gap in the systematic understanding of strategic communication in adversarial settings. To address this, we introduce SDA (Strategic Discourse Assessment), a framework grounded in Gricean and game-theoretic pragmatics to assess strategic use of language. It adapts the ME Game jury function to make it empirically estimable for analyzing dialogue. Our approach incorporates two key adaptations: a commitment-based taxonomy of discourse moves, which provides a finer-grained account of strategic effects, and the use of estimable proxies grounded in Gricean maxims to operationalize abstract constructs such as credibility. Together, these adaptations build on discourse theory by treating discourse as the strategic management of commitments, enabling systematic evaluation of how conversational moves advance or undermine discourse goals. We further derive three interpretable metrics-Benefit at Turn (BAT), Penalty at Turn (PAT), and Normalized Relative Benefit at Turn (NRBAT)-to quantify the perceived strategic effects of discourse moves. We also present CPD (the Crooked Path Dataset), an annotated dataset of real courtroom cross-examinations, to demonstrate the framework's effectiveness. Using these tools, we evaluate a range of LLMs and show that LLMs generally exhibit limited pragmatic understanding of strategic language. While model size shows an increase in performance on our metrics, reasoning ability does not help and largely hurts, introducing overcomplication and internal confusion.