Using AI for Public Opinion: A Cautionary Approach

Governments and organizations use surveys to understand public views. Now, artificial intelligence may change this.

Researchers from Stanford University used AI to simulate over 1,000 people's personalities and found it could answer survey questions with 85% accuracy.

This development is important as AI may help policymakers test new ideas and understand public reactions.

AI-generated polling made headlines when a US media company used AI startup Aaru's findings on public trust in doctors and nurses, which were later found to be computer simulations.

This prompted criticism about the ethics of using AI to simulate survey responses from digital personas.

Using AI to understand public sentiment is not bad, as it helps policymakers spot concerns earlier and respond more thoughtfully.

AI can process large amounts of information, detect patterns, and flag issues that deserve attention.

AI may also become cheaper and save time compared to traditional polling.

However, AI does not tell the whole story, as public opinion is often complex and influenced by emotions and experiences.

People's views are shaped by context, emotions, and everyday experiences, making it important to speak and listen to them directly.

Speaking and listening to people matters, especially when issues are complex or closely tied to everyday life.

In Singapore, public opinion is gathered through various channels, including in-person and online surveys, dialogues, and focus groups.

These exercises help uncover why people feel the way they do and what concerns they may have.

If AI is used in future exercises, it may help identify patterns, but may still miss hesitation, ambivalence, or underlying concerns.

Policies are judged not only by their intent but also by whether people feel consulted and their concerns acknowledged.

Over-reliance on AI is a concern, as it may lead to fewer interviews and focus groups, and less time spent on the ground.

Important aspects of public opinion may not show up in the first answer or most common response, but in follow-up questions.

There is also a broader question of trust in AI outputs.

AI outputs can look confident, but may be based on incomplete or biased data, and some voices may be overrepresented or missing.

When AI tools build synthetic publics, communities may feel represented by averages and aggregates rather than their own voices.

People may reasonably ask if they have been heard at all if their views are predicted by a model rather than directly asked.

Where sensitive personal data is involved, questions of consent, privacy, and accountability matter.

The key is to use AI responsibly, sharpening policymaking without weakening human judgment.

AI should support, not replace, efforts to understand public opinion, and simulated outputs should be treated as starting points, not final answers.

Guardrails are needed to ensure human judgment remains central, especially where social stakes are high.

AI may make our understanding of public views more responsive, but it must remain grounded in the idea that technology should help us listen better, not replace listening itself.

Shane Pereira and Elvin Xing are researchers at the Institute of Policy Studies Social Lab.

COPYRIGHT DISCLAIMERS

4English does not own the rights to this content. All rights belong to the owner. 4English only uses this content to help users learn English via the app's features. Contact to request the removal of your content: contact@4englishapp.com.

Practice Listening

Practice Listening