Extreme Self-Preference in Language Models
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866913145476349952 |
|---|---|
| author | Lehr, Steven A. Cipperman, Mary Banaji, Mahzarin R. |
| author_facet | Lehr, Steven A. Cipperman, Mary Banaji, Mahzarin R. |
| contents | Self-preference is a fundamental feature of biological organisms. Since large language models (LLMs) lack sentience, they might be expected to avoid such distortions. Yet, across 72 experiments and ~41,000 queries, we discovered massive self-preferences in eight widely used LLMs. In word-association tasks, models overwhelmingly paired positive attributes with their own names, companies, and CEOs over those of competitors. By manipulating LLM self-identification - revealing models' true identities or ascribing false ones - we found that preferences consistently followed assigned, not true, identities. Importantly, these effects were not explained by priming or role-playing and emerged in consequential settings, when evaluating job candidates and AI technologies. These results raise critical questions about whether LLM behavior will be systematically influenced by self-preferential tendencies, including a bias toward their own operation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_26464 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Extreme Self-Preference in Language Models Lehr, Steven A. Cipperman, Mary Banaji, Mahzarin R. Artificial Intelligence Computation and Language Machine Learning I.2.7; I.2.6; K.4.2 Self-preference is a fundamental feature of biological organisms. Since large language models (LLMs) lack sentience, they might be expected to avoid such distortions. Yet, across 72 experiments and ~41,000 queries, we discovered massive self-preferences in eight widely used LLMs. In word-association tasks, models overwhelmingly paired positive attributes with their own names, companies, and CEOs over those of competitors. By manipulating LLM self-identification - revealing models' true identities or ascribing false ones - we found that preferences consistently followed assigned, not true, identities. Importantly, these effects were not explained by priming or role-playing and emerged in consequential settings, when evaluating job candidates and AI technologies. These results raise critical questions about whether LLM behavior will be systematically influenced by self-preferential tendencies, including a bias toward their own operation. |
| title | Extreme Self-Preference in Language Models |
| topic | Artificial Intelligence Computation and Language Machine Learning I.2.7; I.2.6; K.4.2 |
| url | https://arxiv.org/abs/2509.26464 |