Communications of the ACM
People Are Bad at Spotting Fake LinkedIn Profiles Generated by AI
Study participants accepted friend requests from 90% of the deepfake profiles that were consistent, and between 79% and 85% of those with obvious errors.
Credit: Chris Batson/Alamy
Scientists at the University of Illinois at Urbana-Champaign (UIUC) and Santa Clara University found fake profiles produced by artificial intelligence (AI) on the business social network LinkedIn can easily deceive people.
Nearly 300 participants of the study examined three profiles each, two of which contained either a deepfake profile picture or AI-generated text.
Participants accepted friend requests from 90% of the deepfake profiles that were consistent, and between 79% and 85% of those with obvious errors.
UIUC's Jaron Mink said they apparently did not notice age differences, and were less suspicious of grammatical errors than of image glitches.
When informed that deepfake profiles have previously been used to trick people, acceptance levels for inconsistent profiles fell by up to 43%.
UIUC's Gang Wang observed that although trained users were better at spotting fake profiles, "The overall stats show they're still not super-good at them."
From New Scientist
View Full Article
Abstracts Copyright © 2022 SmithBucklin, Washington, DC, USA
No entries found