Gespeichert in:
| Hauptverfasser: | , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | https://arxiv.org/abs/2605.30654 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866911730770116608 |
|---|---|
| author | Huang, Jun Rui Zhu, Wang Bill Liu, Ziyi Fast, Nathanael Iyer, Ravi Jia, Robin |
| author_facet | Huang, Jun Rui Zhu, Wang Bill Liu, Ziyi Fast, Nathanael Iyer, Ravi Jia, Robin |
| contents | Large language models (LLMs) are increasingly used as conversational partners for companionship, emotional disclosure, and interpersonal advice, but the social dynamics of these interactions can create harms that are not captured by capability-oriented or traditional safety evaluations. We introduce the Social AI Design Code, a framework for evaluating whether LLMs align with user welfare in social interactions, including whether they encourage harmful intimacy, dependence, or prolonged engagement. To evaluate these risks in natural and diverse user-LLM interactions, we operationalize the code with EUDAIMONIA, a benchmark of 969 user inputs and 3,147 design-requirement violation checks built from WildChat through weak-to-strong filtration, multi-model relabeling, and controlled rewriting. Evaluating 22 recent LLMs, we find that even the strongest models, Claude-Opus-4.7 and GPT-5.5, violate 30.7% and 27.2% of checks, respectively. Extended thinking does not reduce violation rates, suggesting that these failures are persistent social-alignment problems rather than deficits solvable through test-time reasoning alone. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_30654 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | EUDAIMONIA: Evaluating Undesirable Dynamics in AI Huang, Jun Rui Zhu, Wang Bill Liu, Ziyi Fast, Nathanael Iyer, Ravi Jia, Robin Computation and Language Artificial Intelligence Human-Computer Interaction Large language models (LLMs) are increasingly used as conversational partners for companionship, emotional disclosure, and interpersonal advice, but the social dynamics of these interactions can create harms that are not captured by capability-oriented or traditional safety evaluations. We introduce the Social AI Design Code, a framework for evaluating whether LLMs align with user welfare in social interactions, including whether they encourage harmful intimacy, dependence, or prolonged engagement. To evaluate these risks in natural and diverse user-LLM interactions, we operationalize the code with EUDAIMONIA, a benchmark of 969 user inputs and 3,147 design-requirement violation checks built from WildChat through weak-to-strong filtration, multi-model relabeling, and controlled rewriting. Evaluating 22 recent LLMs, we find that even the strongest models, Claude-Opus-4.7 and GPT-5.5, violate 30.7% and 27.2% of checks, respectively. Extended thinking does not reduce violation rates, suggesting that these failures are persistent social-alignment problems rather than deficits solvable through test-time reasoning alone. |
| title | EUDAIMONIA: Evaluating Undesirable Dynamics in AI |
| topic | Computation and Language Artificial Intelligence Human-Computer Interaction |
| url | https://arxiv.org/abs/2605.30654 |