OpenAI published a blog post on 19 February 2019 calling for social scientists to join AI safety research efforts [source]. The post acknowledged that technical AI safety work alone is insufficient and that understanding human behaviour, societal impacts, and deployment contexts requires expertise from social science disciplines.

The company stated that AI systems interact with complex social systems and that predicting their effects requires knowledge of psychology, economics, political science, and sociology. OpenAI identified several research areas where social scientists could contribute, including studying how people interact with AI systems, understanding distributional effects of AI deployment, and examining how AI might affect social institutions.

The post noted that most AI safety research at the time focused on technical problems such as robustness and alignment, but that deployment decisions involve trade-offs that require empirical social science research. OpenAI indicated it was hiring social scientists to work alongside machine learning researchers.

This represents an early acknowledgment by a major AI provider that technical capabilities alone do not address the full scope of AI safety challenges. The post predates widespread public concern about large language model outputs, misinformation risks, and societal-scale effects that emerged after GPT-3's release in 2020.

The call for interdisciplinary collaboration contrasts with subsequent provider communications that have emphasised technical safety measures such as reinforcement learning from human feedback and constitutional AI. Whether social science expertise has been integrated into core safety processes at scale remains unclear from public documentation.

The post provides historical context for understanding how provider safety frameworks evolved and what gaps were identified before the current generation of models entered widespread deployment.