Meat reduction benefits human and planetary health, but social norms keep meat central in shared meals.To date, the development of _communication strategies that promote meat reduction while minimizing social costs_ has required the costly involvement of human participants at each stage of the process.We present work in progress on simulating multi-round dialogues on meat reduction between _Generative Agents_ based on large language models (LLMs). We measure our main outcome using established psychological questionnaires based on the _Theory of Planned Behavior_ and additionally investigate _Social Costs_.We find evidence that our preliminary simulations produce outcomes that are (i) consistent with theoretical expectations; and (ii) valid when compared to data from previous studies with human participants. Generative agent-based models are a promising tool for identifying novel communication strategies on meat reduction—tailored to highly specific participant groups—to then be tested in subsequent studies with human participants.

Lexicons are indispensable tools for textual analysis through labelled term associations.Despite their utility, lexicons are static and require manual effort to curate and maintain.Regular updates are essential to stay relevant amid semantic shifts and neologisms during language evolution.In this work, we explore the potential of supervised learning for expanding lexicons using dictionaries.We study the effect of using dictionaries with varying properties such as noise, size, labels, structure and curation method.Definitions are used as input features to a transformer model (BERT) that assigns categories to terms.We analyse the expansions using varying English dictionaries and lexicons for estimated accuracy, coverage and labelling consistency and apply the expanded versions to a downstream task.Our analyses show dictionary based expansion is a robust approach.We release our expanded lexicons, code, and pretrained models.

The Picture Story Exercise (PSE) is a projective measure in personality psychology where individuals create narratives based on ambiguous images. Traditionally, the coding of these narratives has been labor-intensive. We introduce the Automated Motive Coder (AMC), which employs recent advances in natural language processing and machine learning to automate the coding of PSE narratives. Trained on an extensive dataset, the AMC demonstrates accuracy comparable to expert coders for both original and translated texts. The model offers support for multiple languages that were absent in prior methods while improving in accuracy and speed. To illustrate its effectiveness, we tested and successfully replicated the established psychological effect of gender difference in the affiliation motive. The AMC can be utilized through established machine learning tools, offering a pragmatic and reliable method for coding across several languages. This tool provides an option to reduce the workload involved in PSE coding, promoting efficiency and consistency in motive assessment.

Inferring interpersonal processes such as perspective-taking and empathy from text is a valuable yet not widely developed tool to study inter-individual differences in social interaction. Going beyond lexical categorization, more pragmatically ori-ented classification procedures can profit from insights from traditional qualitative tools in social-cognitive research, espe-cially narrative coding. In this short paper, I briefly discuss the concept of perspective-taking and its points of contact with language. I then illustrate a specific narrative coding sys-tem (Feffer’s decentring scales) and make suggestions as to how this system offers targets for automatized classification procedures, offering impulses for the integration of natural language processing and research into social interaction.

Traditional psychological models of belief revision focus on face-to-face interactions, but with the rise of social media, more effective models are needed to capture belief revision at scale, in this rich text-based online discourse. Here, we use a hybrid approach, utilizing large language models (LLMs) to develop a model that predicts successful persuasion using features derived from psychological experiments.

Our approach leverages LLM generated ratings of features previously examined in the literature to build a random forest classification model that predicts whether a message will result in belief change. Of the eight features tested, epistemic emotion and willingness to share to share were the top-ranking predictors of belief change in the model. Our findings provide insights into the characteristics of persuasive messages and demonstrate how LLMs can enhance models of successful persuasion based on psychological theory. Given these insights, this work has broader applications in fields such as online influence detection and misinformation mitigation, as well as measuring the effectiveness of online narratives.

Humans develop cooperation heuristics in social decision-making, either intuitively or deliberatively. Large language models (LLMs), which exhibit human-like heuristics across cognitive domains, may acquire prosocial tendencies through instruction tuning, latently encoded in their representations to foster cooperative behavior in social reasoning games. However, most studies of this kind either focus on cooperative language generation or explicitly instruct LLMs to cooperate, deviating from the inherent cooperation heuristics of humans. Our negotiation role-play simulations using BATNA (Best Alternative to a Negotiated Agreement) with a GPT-based LLM reveal that LLMs may struggle with cooperation in the absence of explicit instructions, showing a 50–90% lower success rate than in instructed scenarios and a 40–80% lower success rate than human performance reported in past studies. Implicitly inducing cooperation through personality traits had inconsistent effects, with agreeableness showing only a marginal influence and other traits exhibiting no systematic impact. These findings suggest that personality-based cooperation cues are subtle and that explicit instructions may still be essential for multi-agent LLMs to approximate human-like negotiation.

Pragmatic inference remains a persistent challenge even for state-of-the-art large language models (LLMs). While prior efforts have focused on explicit tuning or training, this paper suggests that their limitations may not stem from a lack of inherent capability but from the absence of a multi-agent setup—since pragmatics, by nature, emerges in interaction. This study therefore proposes an adversarial multi-agent LLM framework, modeled after courtroom dynamics, in which a Semantic and a Pragmatic Agent debate scalar implicatures—an essential test of pragmatic competence—and a neutral Judge Agent adjudicates to reveal whether LLMs can derive pragmatic inferences or still default to literal semantics. Experimental results show a clear difference between the single-agent and multi-agent conditions. In the former, the Judge Agent received the same prompt but without access to agent reasoning, and defaulted to a semantic interpretation as observed in prior findings. In the latter, however, the same model successfully derived the implicature—closely aligning with human scalar implicature patterns. This contrast suggests that effective pragmatic reasoning in LLMs arises not from additional tuning, but from contextualized interaction—specifically the kind enabled by a multi-agent framework, which allows latent pragmatic abilities to surface through exposure to competing interpretations. Their pragmatic inference, however, is not indiscriminate. It is modulated by discourse structure, computing scalar implicatures when the implicature-bearing term is foregrounded but defaulting to semantic interpretations when less salient. This study also contributes to the ongoing debate on whether LLMs genuinely engage in pragmatic reasoning or merely simulate it statistically, raising broader discussions about their validity as cognitive models of human linguistic behavior.

According to Self-Determination Theory, human well-being depends on the ability of each person’s environment to support their basic psychological needs (BPNs) for autonomy, competence, and relatedness. The rise of AI and its permeation into human communication increasingly make AI part of that environment. So far, limited work in NLP has leveraged Self-Determination Theory and the concept of BPNs. In this work, we argue that Self-Determination Theory poses a promising framework to extract the antecedents of human well-being from text by considering its commonalities with previously employed emotion theories and by reviewing the surrounding literature. In addition, we argue for AI-mediated communication as a target domain, given the centrality of relationships for BPN satisfaction and the possibility of shaping the field through targeted LLM development. We then outline ten challenges on the path to need-aware, AI-augmented communication.

While automatic fact verification has grown into an established research field in the last years, work on the follow-up step, i.e., conveying fact checking results to users, is still lacking. Presumably, this is because it requires a multidisciplinary approach, as NLP methods to generate such result briefs with a meaningful impact may depend on the consideration of complex psychological processes. In this extended abstract, we discuss the setting and the tasks, questions and considerations connected to it.

As part of this ongoing work, we prompted an LLM with three waves of texts from a longitudinal panel dataset to generate a text for wave 4. We compared generated to ground truth texts using cosine similarity on embeddings and tested whether text similarity was associated with psychological variables. We found limited evidence for an association but do find differences in the topics used in generated versus ground-truth texts. An explanation for differences in text similarities remains the subject of ongoing investigation.