Premium Labels: When Experts Enter the AI Data Annotation Economy

• In Event: Labor and Technology: Platforms, AI, and Inequalities

Abstract

Research on platform labor has long treated data annotation as paradigmatic “ghost work”: low-paid, invisible, and globally distributed labor that undergirds modern machine learning systems. Yet recent shifts in artificial intelligence (AI) research, particularly the limits of scaling large pretrained models, are reshaping this terrain. As attention turns toward post-training, fine-tuning, and evaluation work, a growing set of platforms and in-house industry programs are recruiting academics and professionals to perform what we call expert annotation. This labor commands unusually high wages, often exceeding $75 per hour, and relies on academic credentials and disciplinary authority as key gatekeeping mechanisms. Drawing on a comparative study of AI annotation platforms and interviews with academics and professionals engaged in expert annotation, this paper introduces the concept of premium labels: annotations produced by converting professional capital—degrees, disciplinary knowledge, and prestige—into market value within AI’s data supply chain. We analyze public-facing marketing materials, internal platform affordances, and interview accounts to examine how expertise is recruited, valued, and experienced in this emerging segment of data labor. We identify three core dynamics. First, expertization as valuation: companies frame expert knowledge as a quality assurance mechanism, transforming disciplinary training into a premium commodity. Second, boundary construction and justificatory framing: participants describe their work as consulting, service, moonlighting, or exploitation, drawing moral and professional boundaries between “data work” and recognized professional labor. Third, structured ambivalence: while some experts view annotation as a troubling outsourcing that risks undermining their fields, others see it as an opportunity to influence emerging technologies from within; across positions, high pay, prestige, and insider access exert a powerful pull. Theoretically, the paper extends platform labor scholarship by moving beyond ghost work to theorize a stratified annotation economy in which precarious clickwork coexists with elite, well-compensated expert labor.

Authors

• Skyler Wang, McGill University

• Leo Holton, McGill University