Arnold Kling argued that public trust in AI could help move society beyond the post-truth era, though fears of political bias may limit whether people accept the systems.
In a Nov. 1 essay on Substack, Kling framed the possibility against problems he sees in public debate: audiences can choose media that match their politics, trust in institutions has weakened and social platforms contribute to polarization. He pointed to Walter Cronkite and prominent academics as examples of people who once held broad public confidence, then asked whether AI might earn similar standing among the public.
Kling also cited Tyler Cowen’s proposal for the National Science Foundation, prize committees, universities, tenure committees and journals to use AI to review their own work. Cowen wrote that the world was not rushing to conduct such audits, and Kling noted Cowen’s example of AI criticism of research by two prominent left-leaning economists.
His own example came from using Claude to research 20th-century American history. Kling said the chatbot appeared to favor progressive reformers. He also said the model gave greater emphasis than he would to historical mistreatment of Black people, women and workers. He considered strong attention to Black people’s mistreatment justified and said reformers were mostly on the right side through 1964. For women and workers, he credited the market system with providing the most beneficial changes.
Kling described competing political concerns about AI. He wrote that people on the left could lose influence if models expose their side’s biases, while conservatives may see the systems as reinforcing left-leaning ideas because of the training texts and employees who refine them. He said trusting AI on 20th-century American history would not make a reader a left-wing radical. Kling said he would favor public trust if models tested claims tied to left-leaning views more carefully.
