Lorenzo Cima
Publications
Contextualized Counterspeech Can Be More Persuasive Than Generic Counterspeech
AI-generated counterspeech offers a scalable and effective strategy to mitigate online toxicity by promoting more constructive dialogue. Yet, existing approaches adopt a generic, one-size-fits-all paradigm, overlooking the conversational context and characteristics of the targeted users. Here, we propose and evaluate multiple strategies for generating contextualized counterspeech that is adapted to the moderation setting and personalized to the moderated user. In detail, we explore a range of configurations that integrate different forms of contextual information and fine-tuning techniques. We conduct a comprehensive evaluation combining quantitative indicators with a pre-registered, mixed-design crowdsourcing experiment. To ensure robustness, we implement algorithmic measures of counterspeech quality based on ROUGE, BLEU, and BERTScore, observing overall consistent results across metrics. Furthermore, we analyze which characteristics of both the generated counterspeech and the moderated toxic message most strongly influence perceived persuasiveness, yielding insights into how contextualized interventions can be made more effective. Our findings show that personalization can be effective, but not uniformly so. Lightweight strategies combining conversational context and user history improve perceived adequacy and persuasiveness, whereas several other contextualization strategies degrade human-perceived counterspeech quality. Taken together, these results provide actionable directions for developing more personalized, effective, and responsible counterspeech systems, ultimately advancing human-AI collaboration in online content moderation.
A Geometric Analysis of Small-sized Language Model Hallucinations
Hallucinations -- fluent but factually incorrect responses -- pose a major challenge to the reliability of language models, especially in multi-step or agentic settings. This work investigates hallucinations in small-sized LLMs through a geometric perspective, starting from the hypothesis that when models generate multiple responses to the same prompt, genuine ones exhibit tighter clustering in the embedding space, we prove this hypothesis and, leveraging this geometrical insight, we also show that it is possible to achieve a consistent level of separability. This latter result is used to introduce a label-efficient propagation method that classifies large collections of responses from just 30-50 annotations, achieving F1 scores above 90%. Our findings, framing hallucinations from a geometric perspective in the embedding space, complement traditional knowledge-centric and single-response evaluation paradigms, paving the way for further research.
Assessing the Reliability of Persona-Conditioned LLMs as Synthetic Survey Respondents
Using persona-conditioned LLMs as synthetic survey respondents has become a common practice in computational social science and agent-based simulations. Yet, it remains unclear whether multi-attribute persona prompting improves LLM reliability or instead introduces distortions. Here we contribute to this assessment by leveraging a large dataset of U.S. microdata from the World Values Survey. Concretely, we evaluate two open-weight chat models and a random-guesser baseline across more than 70K respondent-item instances. We find that persona prompting does not yield a clear aggregate improvement in survey alignment and, in many cases, significantly degrades performance. Persona effects are highly heterogeneous as most items exhibit minimal change, while a small subset of questions and underrepresented subgroups experience disproportionate distortions. Our findings highlight a key adverse impact of current persona-based simulation practices: demographic conditioning can redistribute error in ways that undermine subgroup fidelity and risk misleading downstream analyses.