Send message to Author

Short abstract

Through interviews with expert data labelers and a socio-technical analysis of RLHF, this paper shows how expertise stabilizes uncertainty and legitimizes ground truths in non-convergent domains like the humanities. It also reveals the opacity of labor platforms when assigning expert credentials.

Long abstract

This paper examines the recent practice of hiring “experts“ – defined by model developers and labor platforms as Master’s and PhD degree holders in relevant domains – to enhance model performance in the Reinforcement Learning with Human Feedback(RLHF) phase of LLM development. It combines a socio-technical analysis of the RLHF process with in-depth interviews with “expert” workers across different fields on platforms like Surge AI to understand how model developers conceive of expertise, and the assumptions underlying the technical infrastructure that attempts to “encode” expertise.

The logics of RLHF rely on expert reviewers converging on a set of responses to prompts. As literature in the sociology of expertise notes, such a consensus is often challenging. Annotation firms manage this by using strict rubrics, detailed guidelines, and layers of review. What emerges is a form of expertise constructed by a tension between the variability of expert analysis and attempts to quantitatively stabilize uncertainty. This epistemic weakness reaches its limit when “expert” logic is applied to creative fields, reducing creativity to credentials and rubrics, foreclosing radical uncertainty.

The contingent nature of this ground truth is compounded by the unstable category of the expert. Despite advertising highly credentialed “experts”, data annotation firms resort to opaque assessments to establish expertise, creating a gamified environment that prioritizes credential acquisition. This instability is exacerbated by the inconsistency, isolation, and surveillance characteristic of gig work. This paper highlights the epistemic limits of a ground truth based on “expert” consensus, especially as it relies on distributed gig labor.