Article · AI translation
IT Home reported on October 26 that researchers at the University of Limerick in Ireland said in a research paper published earlier this month that, in an experiment involving 189 knowledge workers, people preferred AI agents with humanlike appearances, but also showed gender bias.
Participants worked with 3 AI agent colleagues based on the same underlying model in a virtual office. The 3 virtual colleagues appeared as a text chat window, a small virtual robot, and a humanlike figure, respectively; the humanlike figure had 2 versions, a female one named "Johanna" and a male one named "Johan". After working with the agents, participants rated each AI assistant and allocated resources for further development based on its performance.
The results showed that humanlike figures received higher levels of trust, credibility, and "pay" than the more robotic agents. There was no significant difference between "Johanna" and "Johan" in trust or credibility, but the feminized "Johanna" still received statistically significantly lower "pay" than her male "colleague," even though most participants interviewed did not think virtual "gender" would affect their feedback.
Other noteworthy findings included that the text chat window and small virtual robot received comparable ratings for trust and credibility, "Johann" received a significantly lower rating for human likeness than "Johan," and participants preferred the appearance of "Johann" as an AI assistant.
Reference
https://dl.acm.org/doi/10.1145/3829807.3829910
Advertising disclosure: External links in this article (including but not limited to hyperlinks, QR codes, and access codes) are intended to provide more information and save readers time in selecting sources. The results are for reference only. All IT Home articles containing external links include this disclosure.