Who gets to be “free” with identity-altering AI?
How people judge the ethics of deepfakes, voice cloning, and AI avatars, and why those judgments, and the funding that follows, split along group lines.
Collaborative research with Dr. Ariel Mosley · presented at SPSSI
Why it matters now
Identity-altering AI is already mainstream. The same tool that lets one person explore a new creative identity can let another commodify someone else's. As companies ship synthetic-media and avatar features, the open question is not only what the technology can do, but how users will judge it, and whether those judgments turn into trust, backlash, or lost revenue.
This work moves that question from opinion to evidence: three controlled experiments testing who is granted the “freedom” to use these tools, how ethical that use is perceived to be, and what people do with their money as a result.
The technologies tested
- Voice cloningsynthesized speech and vocal style
- Generated facesAI-made likenesses and imagery
- Video deepfakesswapped identities in motion
- Digital avatarsembodied identities in virtual space
Explore the three experiments
Each study answers a sharper version of the same question. Switch between them to see the design and what it showed.
Do people judge identity-altering AI differently depending on their own group, and why?
How it was designed
A 2 × 2 experiment crossing the perceiver's race with whether the AI represented an in-group or out-group identity, using four scenarios drawn from real news headlines (voice, imagery, deepfake, avatar).
What it showed
White participants granted more “freedom” to use out-group representations and rated that use as more ethical than Black participants did. Black participants pulled the most funding from the companies behind it; White participants did not differentiate. The gap was statistically explained by how much freedom people felt the dominant group was owed.
What it means for AI perception
Ethics judgments are not universal
The same AI feature reads as harmless self-expression to one audience and as exploitation to another. Whose identity is represented, and who is doing the representing, changes the verdict.
Moral judgments move money
How ethical people found a use directly predicted whether they would fund or withhold money from the company behind it. Perception translates into measurable financial behavior.
“Ownership” of identity is asymmetric
Borrowing a marginalized group's likeness was treated differently from the reverse, and the penalties fell unevenly across groups.
Restriction messaging can backfire
Framing the technology as under threat of regulation increased perceived ethicality and support among dominant-group users. Heavy-handed “we must restrict this” comms can entrench the very behavior it targets.
Where this applies
Tap a domain to see how the findings translate into practice.
Freedom shapes moral judgment, and whose freedom gets centered determines what people count as ethical. For anyone building or governing identity-altering AI, that is a design and policy problem, not just a philosophical one.
Collaborative research with Dr. Ariel Mosley. Reframes the manuscript “When AI Becomes ‘Other’: The Role of Racial Representation and the Uncanny Valley in Human–AI Relationships” (under review, Journal of Experimental Psychology) and an SPSSI conference talk for an applied audience.