Commentary: AI polling needs a pause, prioritize human listening

Artificial intelligence is emerging as a powerful tool in understanding public sentiment, with researchers demonstrating AI's ability to simulate survey responses with up to 85 percent accuracy based on in-depth interviews. This development, highlighted by Stanford University researchers, suggests AI could become a valuable addition to the policymaking and research toolkit, particularly for testing new policy ideas and gauging potential public reactions. However, the application of AI in polling is not without controversy, as seen when a US media company referenced AI-generated "findings" on public trust in doctors that were later found to be computer simulations based on "silicon sampling", the use of large language models to mimic survey responses from digital personas.

Proponents argue that AI offers significant benefits. In an era of information overload, AI can rapidly process vast amounts of data, detect subtle patterns, and flag emerging issues for closer attention. It can also provide quick pulse-checks on potential public reactions to policies. Furthermore, AI-driven polling may eventually become more cost-effective and time-saving compared to traditional survey methods. This efficiency could allow policymakers and researchers to respond more thoughtfully and proactively to public concerns.

“

AI is good at producing neat answers, but public opinion is often less tidy than that.

However, academics Shane Pereira and Elvin Xing from Singapore's Institute of Policy Studies Social Lab caution against over-reliance on AI. They emphasize that while AI excels at producing neat answers, public opinion is inherently complex and often less tidy. People's views are deeply intertwined with their context, emotions, and everyday experiences. An individual might support a policy in principle but harbor concerns about its implementation, or their initial supportive stance might shift upon deeper reflection. This nuanced reality underscores the continued importance of direct human interaction.

Pereira and Xing advocate for AI to serve as a supportive tool rather than a replacement for traditional listening methods. They point to Singapore's REACH unit, which uses a combination of in-person and online surveys, dialogues, and focus groups to gather public views. This multi-channel approach acknowledges that understanding public opinion requires more than just data processing; it necessitates genuine listening and engagement with people to capture the full spectrum of their thoughts and feelings, especially on complex issues tied to daily life.

“

This is why speaking and listening to people matters, especially when the issues are complex or closely tied to everyday life.