MIT Media Lab Study Finds AI-Edited Images Generate False Memories and False Confidence
In an experiment led by Pattie Maes from MIT's Media Lab, presented at the CHI 2025 conference, it was found that participants who viewed AI-modified images and videos were more likely to misremember the original photographs compared to a control group, and they had greater confidence in their incorrect recollections. The study involved 200 participants who first saw 24 baseline images, then were split into four groups: one control group revisited the originals, while others viewed AI-altered stills, five-second AI-generated videos from the originals, and videos derived from edited images. Those exposed to the edited videos exhibited 2.05 times more false memories, with confidence levels 1.19 times higher than the control group. Additionally, researchers Gideon Nave and Steven Shaw from the University of Pennsylvania are examining the cognitive impacts of AI.
Key facts
- MIT Media Lab professor Pattie Maes and colleagues ran the experiment on AI-edited photos and memory
- The study was published at the CHI 2025 conference
- 200 participants viewed 24 baseline photographs of scenes such as a wedding, a city street, and a race
- Participants were divided into four groups of 50, including a control group that re-viewed the originals
- AI edits added, removed, or changed objects, people, and surrounding environment elements
- The group watching AI-generated videos based on AI-edited photos produced 2.05 times as many false memories as the control
- That group's confidence in incorrect answers was 1.19 times the control level
- Gideon Nave of the Wharton School warned of a world where "thinking becomes optional"
Entities
Artists
- Pattie Maes
- Gideon Nave
- Steven Shaw
Institutions
- MIT Media Lab
- Massachusetts Institute of Technology
- The Washington Post
- CHI 2025
- University of Pennsylvania
- Wharton School
- PetaPixel
Locations
- Cambridge
- United States
- Philadelphia
- Pennsylvania