We study fine-grained stereotypes of LLMs activated by linguistic features representing demographic groups, developing a probe-based framework to extract 235 fine-grained traits (e.g., friendly, uneducated, honest) from model representations of prompts. Using these probes, we conduct two studies analyzing the underlying structure of traits and specific stereotypes of demographic groups. We find that fine-grained trait information is represented in LLMs similarly to coarse stereotype content dimensions (e.g., warmth and competence), and that the latent factors of trait probes (e.g., morality, competence, sociability) largely align with those in the psychology literature (Fiske et al., 2002) despite traits being modeled independently. Finally, we extract LLMs’ associations between traits and demographics signaled through language to surface granular stereotypes encoded in LLMs. We, for example, find that the evaluated models encode associations between African American Language and traits clumsy, uneducated, and criminal. These associations may reinforce harmful stereotypes, with implications for both general dialogue responses and in domains like hiring. Overall, we discuss the need to consider fine-grained traits in stereotype evaluations in order to distinguish stereotypes of out-groups and between models.

Psychological insights have long shaped pivotal NLP breakthroughs, from attention mechanisms to reinforcement learning and social modeling. As Large Language Models (LLMs) develop, there is a rising consensus that psychology is essential for capturing human-like cognition, behavior, and interaction.This paper reviews how psychological theories can inform and enhance stages of LLM development. Our review integrates insights from six subfields of psychology, including cognitive, developmental, behavioral, social, personality psychology, and psycholinguistics. With stage-wise analysis, we highlight current trends and gaps in how psychological theories are applied. By examining both cross-domain connections and points of tension, we aim to bridge disciplinary divides and promote more thoughtful integration of psychology into NLP research.

There are many settings where it is useful to predict and explain the success or failure of a dialogue. Circumplex theory from psychology models the social orientations (e.g., Warm-Agreeable, Arrogant-Calculating) of conversation participants and can be used to predict and explain the outcome of social interactions. Our work is novel in its systematic application of social orientation tags to modeling conversation outcomes. In this paper, we introduce a new data set of dialogue utterances machine-labeled with social orientation tags. We show that social orientation tags improve task performance, especially in low-resource settings, on both English and Chinese language benchmarks. We also demonstrate how social orientation tags help explain the outcomes of social interactions when used in neural models. Based on these results showing the utility of social orientation tags for dialogue outcome prediction tasks, we release our data sets, code, and models that are fine-tuned to predict social orientation tags on dialogue utterances.