Comment More uncanny, worse agents, allocated less (Score 2) 210
The observed disparity in monetary rewards between the male- and female-presenting AI assistants can be explained by several methodological factors rather than purely gender bias. Most notably, participants rated the female-presenting assistant as significantly less human-like than the male counterpart, a factor explicitly linked to lower reward allocations throughout the study. Additionally, the researchers used specific off-the-shelf voice and visual models, meaning technical variances in audio quality or visual realism between those specific assets could have driven user preference. Furthermore, initializing the underlying language model with gendered names to establish character roles may have subtly altered the tone or quality of its responses, inadvertently affecting the user experience. Finally, because the experimental task asked participants to fund the AI's future "development" rather than pay a direct wage for completed work, the financial allocations likely measured perceived organizational utility rather than a true evaluation of the assistant's labor.