People trust AI faces more than real ones, says new study - ThePrint
Summary
A new study led by psychology PhD student Alexis McGuire from Lancaster University, published in the *Journal of Vision*, reveals that people tend to trust AI-generated faces more than real human faces. This research, the first to examine the trustworthiness of AI faces, found that participants rated faces created using diffusion models (DM) as most trustworthy (4.70 average rating), followed by GAN faces (4.36), and then real faces (4.03). In an initial test, participants could only distinguish fake from real faces with 58.4% accuracy. The findings highlight a significant risk, especially as AI-powered scams are projected to cause \$40 billion in losses by 2027 in the US alone, according to Deloitte. Researchers emphasize the ease with which AI models can create convincing fake faces, leading to potential misuses like misinformation, identity fraud, and catfishing.
Key takeaway
For public safety officials and policy makers addressing online fraud, this study underscores the urgent need for public education campaigns. Your constituents are more susceptible to AI-generated deepfakes than previously understood, trusting them more than real faces. You should prioritize developing and deploying clear public advisories on identifying AI-generated content and the risks of identity fraud, misinformation, and catfishing. Consider adopting AI-powered solutions, like Japan's AIko, to proactively combat these evolving threats.
Key insights
People inherently trust AI-generated faces more than real ones, posing risks for misinformation and fraud.
Principles
- AI-generated faces, especially DM, are perceived as more trustworthy.
- Public awareness of AI deepfake creation is crucial.
- Generative AI democratizes creation of deceptive content.
Method
Participants rated 96 faces (real, GAN, DM) on a 1-7 trustworthiness scale after an initial fake/real identification task. Average accuracy was 58.4%.
In practice
- Inform public about AI deepfake creation ease.
- Develop strategies to mitigate AI-driven harms.
- Implement AI tools for scam detection (e.g., AIko).
Topics
- AI Deepfakes
- Generative AI Trust
- Online Fraud
- Diffusion Models
- Public Awareness
- Digital Identity
Best for: Investor, CTO, VP of Engineering/Data, AI Ethicist, Policy Maker, General Interest
Related on AIssential
See Counsel's argued verdicts on the open AI decisions leaders are weighing →
Editorial summary, takeaway, and curation by AIssential. Original article published by artifical intelligence via Google News.