HCL+PrioR mtg. Reed Orchinik. Thursday 4/23 12:30pm-2:00pm ET

[Antonio Alonso Arechar]

Good morning everyone,

Reed Orchinik will present this Thursday (4/23 12:30pm-2:00pm ET; see blurb below).

Lab meeting will be hybrid, with a zoom link (https://mit.zoom.us/j/91655448125) and in person in the conference room on the 15th floor of E94 (1579; 245 First Street).

Best regards,

Antonio

---

Title: The indirect harms of AI-generated videos

Abstract: Generative AI is changing the content that people are exposed to online. Across 4 pilot studies, we explore the ways in which the presence of generative AI changes perceptions of both real and AI-generated videos. In 3 of the pilots, we systematically vary the proportion and believability of AI-generated content and find that two harms may emerge depending on the composition of the video feed. When AI slop (obviously AI-generated content) is present, people are more likely to find other videos to be realistic (a contrast effect). When AI-generated content is more ambiguous, people may become skeptical of all videos (an assimilation effect). In a fourth pilot, we test the impact of warning labels on discernment between real and AI-generated videos and find potential differences in effectiveness by the plausibility of the video that is being labeled. NOTE: This work is at a very early stage so input on experimental design and help making sense of the pilot results would be very much appreciated.