Academic Roots Enable Safety Evaluations
Jemma White, COO of Prolific, on why humans ensure AI safety
Prolific’s academic roots matter because they make quality legible in a market where many vendors sell labor volume. The company began as research infrastructure for Oxford academics, then carried that same workflow into AI, matching specific participants to studies, tracking response quality over time, and giving customers transparent control over screening and pay. That history helps explain why Prolific can sell safety evals and nuanced research, not just generic annotation work.
-
Academic customers are still a meaningful business line, not just origin story branding. They run AI studies and behavioral research on the same platform, which keeps participant supply active across many use cases and gives Prolific years of performance history on who responds well to what kind of task.
-
The competitive contrast is concrete. Scale grew out of large scale labeling, Handshake turned university credentials into expert contractor supply, while Prolific is built around deeply profiled participants and fast self serve studies. That makes it better suited to red teaming, cultural nuance, trust and safety, and other jobs where how a person thinks matters as much as what degree they hold.
-
Credibility also lowers friction with frontier labs. When a model team needs an external second opinion on whether an output is harmful, biased, or culturally off, a participant pool shaped by academic research norms feels more defensible than a black box managed service or a low cost crowdwork marketplace.
The market is moving toward higher stakes human evaluation, not away from it. As AI work shifts from raw labeling to safety, behavior, and localization, vendors with durable participant data and research grade processes should capture more of the value, and Prolific’s academic heritage becomes more important as regulation and customer scrutiny rise.