Preprint Traces AI's Turing Test Back to Pygmalion Displacement

A preprint published on SocArXiv by Lelia Erscoi, Annelies Kleinherenbrink, and Olivia Guest argues that AI development has long followed a pattern they call "Pygmalion displacement": the progressive replacement of women by inanimate technological objects — from automata to algorithms — in ways that remain largely unacknowledged within the field.

The authors trace this pattern to AI's intellectual foundations. The paper argues that Pygmalion displacement "prefigures heavily, but in an unacknowledged way, in the original Turing test, the imitation game" — the thought experiment central to how the field conceptualizes machine intelligence. The mythological connection is not purely metaphorical: according to a published review of the paper, Joseph Weizenbaum explicitly invoked Pygmalion when introducing ELIZA, his 1964 chatbot. The review also notes that voice systems like Siri and Alexa are characterized as gendered feminine through disembodied voices performing "emotional labor."

To make the analysis actionable, the paper introduces a 10-question diagnostic framework for identifying Pygmalion displacement in AI products. The questions ask whether a system carries feminised external characteristics, whether it displaces women from their work, whether it frames AI as competitive with women, and whether it serves psychological ego-service functions, among others. According to the authors, if multiple questions register affirmatively, "Pygmalion displacement is occurring, causing documented harm."

The honest caveat is that this is a philosophical and historical analysis rather than an empirical study. What available summaries do not provide is data from applying the diagnostic to specific commercial systems, or evidence that the 10 criteria have been validated against measurable outcomes. The preprint dates to 2023, and it is not clear from available sources whether it has since been peer-reviewed.

For AI practitioners, the most concrete output is the diagnostic itself. If the framework gains traction with ethics teams, regulators, or standards bodies, it could become a structured tool for auditing products at the design stage rather than after deployment.