Researchers at the Massachusetts Institute of Technology (MIT), Vassar College, and the Max Planck Institute for Human Development in Germany surveyed several hundred people to determine whether they view artificial intelligence (AI) like an agent or a tool.
The study focused on a painting produced using a generative adversarial network that sold at auction in 2018.
Study participants were asked to rate how much each party that took part in the creation of the painting should be given credit, and how much they agreed with various statements about the fictional algorithm that implied agency.
The researchers found a significant correlation between the extent to which participants agreed with statements about the algorithm's agency, and how much credit they gave to different parties.
Said MIT's Ziv Epstein, "By tweaking the language, to up-play the agent-ness of the AI ... we can manipulate the allocation of money and responsibility."
From "People's Notions About AI Are Terrible, an MIT Study Asks Whether They Can Be Helped"
ZDNet (09/18/20) Tiernan Ray
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA