Raw capture — raw/seeing-is-not-believing.md

url: https://www.media.mit.edu/publications/seeing-is-not-believing/ title: "Seeing Is Not Believing: Realistic AI Videos Disrupt Confidence in Authentic Videos and Perceived Reality" authors: "Eitan Wolf, Yasith Samaradivakara, Om Gokhale, Sally Ahmed, Yuhan Wang, Pat Pataranutaporn, and Pattie Maes" date_fetched: 2026-07-25 date_published: 2026-03-13 venue: "Extended Abstracts of the 2026 CHI Conference on Human Factors in Computing Systems (CHI EA '26), ACM" doi: "10.1145/3772363.3798358" affiliations: "MIT Media Lab, Opera of the Future" license: "Creative Commons Attribution 4.0 International"

Seeing Is Not Believing: Realistic AI Videos Disrupt Confidence in Authentic Videos and Perceived Reality

Authors

Eitan Wolf, Yasith Samaradivakara, Om Gokhale, Sally Ahmed, Yuhan Wang, Pat Pataranutaporn, and Pattie Maes

Venue

Published in Extended Abstracts of the 2026 CHI Conference on Human Factors in Computing Systems (CHI EA '26), ACM. DOI: 10.1145/3772363.3798358

Abstract

Visual media has traditionally served as a reference for what feels real, but generative video models now produce synthetic footage resembling real-world scenes. This content increasingly appears in short-form social feeds where AI-generated and human-generated videos are consumed in rapid succession. Prior work has focused on detecting synthetic content, while less is understood about how such exposure shapes perception of subsequent human-generated media.

Methodology

A two-phase study with N = 100 participants was conducted. The study compared: - An AI-exposure group — participants first exposed to AI-generated videos - A control group — participants viewing human-generated videos

After this initial phase, both groups transitioned to human-generated content for the second phase.

Key design feature: AI-generated videos were accompanied by clear disclosure that they were synthetic.

Key Findings

Despite disclosure that the AI videos were synthetic, the AI-exposure group reported:

  1. Increased doubt about the authenticity of subsequent videos
  2. Reduced judgment confidence regarding whether videos were real or AI-generated
  3. Greater perceptual disruption
  4. Lower social connectedness

Conclusions

Perceptual and psychological effects occur even when synthetic content is disclosed. The authors call for design strategies prioritizing "perceptual safety" — addressing experiential impacts beyond deception.

Affiliated Groups & Project

Related Content on the Page