Cheap Talk with the Bayesian Truth Serum

The data collected through surveys form the groundwork for economic and business research. If biases exist in survey responses, they could seriously harm and mislead our decision-making. To mitigate survey response bias, we suggest an alternative way of combining two existing strategies, cheap talk and the Bayesian Truth Serum (BTS). In our three proof-of-concept experiments, we found that our alternative approach, named the C-BTS, helps elicit more truthful survey responses even in situations where neither the BTS nor cheap talk alone works well enough, especially in the context of economic valuation of goods. We presume that the success of the C-BTS in our experiments can be attributed
to its structure being similar to real choices: The cheap talk component induces people to think about their decision-making in the context of their real life, and the BTS component ensures that their choices have consequences that directly affect their utility. By applying the C-BTS, we have also found that 12 popular AI-powered services in daily life provide an annual consumer value of approximately $189.1 billion in the United States, which corresponds to about 0.74% of the total U.S. GDP in 2022.