Study Finds AI Use by 13% of UK Voters Yielded Knowledge Gains Comparable to Web Search

A recent study conducted in the week leading up to the 2024 UK general election revealed a "cautiously optimistic result" regarding artificial intelligence and disinformation. The research indicated that 13% of all voters utilized AI tools for political topics. This finding comes amidst widespread concerns about the potential for AI to spread misinformation during electoral periods.

According to a social media post by Wharton Professor Ethan Mollick, a randomized trial associated with the study found that using AI led to gains in "true knowledge" similar to those achieved through traditional web searches. Mollick stated in his tweet, "> A randomized trial found this may be good: using AI led to similar gains in true knowledge as doing web search, regardless of model & prompts." This suggests that, in certain contexts, AI can serve as an effective tool for information acquisition without necessarily increasing vulnerability to disinformation.

The 2024 election cycle, dubbed a "super election year" globally, saw significant apprehension regarding the proliferation of AI-generated content. Experts and organizations, including the World Economic Forum, had highlighted AI-powered misinformation as a top short-term risk. Instances of AI-generated robocalls, such as those impersonating President Joe Biden in the US, and deepfake video ads, like those targeting UK Prime Minister Rishi Sunak, underscored these fears.

However, other analyses have offered a more nuanced perspective. Research from the Alan Turing Institute's Centre for Emerging Technology and Security (CETaS) found no evidence that AI-enabled misinformation meaningfully impacted the results of recent UK or European elections. Their studies identified a limited number of viral AI disinformation cases, suggesting that such content primarily reinforced existing political beliefs rather than swaying undecided voters.

Ethan Mollick, a prominent voice on the practical implications of AI and co-director of the Generative AI Labs at Wharton, frequently emphasizes the need to understand how AI tools integrate into daily life. His work, including his book "Co-Intelligence," explores how humans can effectively collaborate with AI. The study's findings contribute to a growing body of research that seeks to balance alarmist predictions with evidence-based assessments of AI's actual impact on information consumption and political discourse.