Cultural Conditioning or Placebo? On the Effectiveness of Socio-Demographic Prompting
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866909227758387200 |
|---|---|
| author | Mukherjee, Sagnik Adilazuarda, Muhammad Farid Sitaram, Sunayana Bali, Kalika Aji, Alham Fikri Choudhury, Monojit |
| author_facet | Mukherjee, Sagnik Adilazuarda, Muhammad Farid Sitaram, Sunayana Bali, Kalika Aji, Alham Fikri Choudhury, Monojit |
| contents | Socio-demographic prompting is a commonly employed approach to study cultural biases in LLMs as well as for aligning models to certain cultures. In this paper, we systematically probe four LLMs (Llama 3, Mistral v0.2, GPT-3.5 Turbo and GPT-4) with prompts that are conditioned on culturally sensitive and non-sensitive cues, on datasets that are supposed to be culturally sensitive (EtiCor and CALI) or neutral (MMLU and ETHICS). We observe that all models except GPT-4 show significant variations in their responses on both kinds of datasets for both kinds of prompts, casting doubt on the robustness of the culturally-conditioned prompting as a method for eliciting cultural bias in models or as an alignment strategy. The work also calls rethinking the control experiment design to tease apart the cultural conditioning of responses from "placebo effect", i.e., random perturbations of model responses due to arbitrary tokens in the prompt. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_11661 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Cultural Conditioning or Placebo? On the Effectiveness of Socio-Demographic Prompting Mukherjee, Sagnik Adilazuarda, Muhammad Farid Sitaram, Sunayana Bali, Kalika Aji, Alham Fikri Choudhury, Monojit Computation and Language Socio-demographic prompting is a commonly employed approach to study cultural biases in LLMs as well as for aligning models to certain cultures. In this paper, we systematically probe four LLMs (Llama 3, Mistral v0.2, GPT-3.5 Turbo and GPT-4) with prompts that are conditioned on culturally sensitive and non-sensitive cues, on datasets that are supposed to be culturally sensitive (EtiCor and CALI) or neutral (MMLU and ETHICS). We observe that all models except GPT-4 show significant variations in their responses on both kinds of datasets for both kinds of prompts, casting doubt on the robustness of the culturally-conditioned prompting as a method for eliciting cultural bias in models or as an alignment strategy. The work also calls rethinking the control experiment design to tease apart the cultural conditioning of responses from "placebo effect", i.e., random perturbations of model responses due to arbitrary tokens in the prompt. |
| title | Cultural Conditioning or Placebo? On the Effectiveness of Socio-Demographic Prompting |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2406.11661 |