Want to solve deepfakes? Ask citizens what to do

As deepfake technology becomes more sophisticated and accessible, traditional approaches to combating AI-generated misinformation are hitting a wall. Instead of relying solely on tech companies and government regulators, a growing movement suggests an unconventional solution: ask ordinary citizens what to do.

Why technical solutions aren't enough

The current arms race between deepfake creators and detection systems resembles a digital game of whack-a-mole. Meta's latest detection algorithms achieve 85% accuracy at best, while new deepfake tools emerge monthly with improved capabilities to fool existing safeguards.

But the deeper issue isn't just technical—it's philosophical. Who decides what constitutes harmful content? How do we balance free expression with protection from deception? These questions require social consensus, not just better algorithms.

Stanford University researchers studying AI governance argue that "deepfake regulation touches on fundamental values that can't be encoded into software." The challenge lies in navigating competing priorities: protecting democracy from manipulation while preserving legitimate uses of synthetic media in entertainment, education, and art.

The citizen jury experiment

Taiwan pioneered an innovative approach: citizen assemblies dedicated to AI governance. In their most recent experiment, 150 randomly selected citizens spent three days deliberating on deepfake policy after hearing from experts across technology, law, and civil society.

The results surprised many observers. Rather than demanding aggressive content removal or sophisticated detection systems, participants prioritized mandatory labeling of synthetic content and enhanced media literacy education. They recognized that perfect detection might be impossible, but informed consumption could be achievable.

The European Union has scaled this approach across 12 countries, engaging 2,400 citizens in discussions about AI regulation. Early findings suggest that ordinary people often demonstrate more nuanced thinking about trade-offs than assumed by policymakers who tend toward binary solutions.

Beyond traditional governance

This participatory approach challenges conventional wisdom about complex policy-making. Critics argue that citizens lack the technical expertise to make informed decisions about AI governance. Supporters counter that lived experience with technology often provides insights that experts miss.

MIT's Center for Collective Intelligence found that diverse citizen groups consistently outperformed expert panels when addressing multi-faceted problems with significant social implications. The key advantage appears to be citizens' ability to consider broader impacts beyond narrow technical metrics.

Several tech companies are now experimenting with citizen input mechanisms. Anthropic recently convened focus groups to help shape their AI safety policies, while OpenAI has explored public consultation processes for sensitive AI applications.

The implementation challenge

Translating citizen input into actionable policy remains complex. Democratic deliberation takes time—a luxury when dealing with rapidly evolving threats. There's also the question of representation: ensuring participant diversity across age, education, geography, and digital literacy levels.

University of Oxford governance researchers suggest hybrid models combining citizen assemblies with expert advisory panels and real-time public feedback mechanisms. This approach could provide both democratic legitimacy and technical competence.

The stakes extend beyond deepfakes. How societies handle AI governance today will establish precedents for addressing future technological challenges, from autonomous weapons to artificial general intelligence.