Recent evidence and trends have been undermining some predictions that robots are about to steal all our jobs. Researchers have evolved from believing that automation must lead to substantial unemployment. Many now argue that automation may actually increase employment. This can happen because Artificial Intellignece (AI) raises firm productivity and also because, in some situations, AI acts as a complement to human expertise. Rather than having nothing to do, it looks like we may instead learn to work alongside our new robotic friends. This grant supports Nikhil Agarwal and Tobias Salz at MIT who are investigating the collaborative nature of interactions between people and AI in knowledge-intensive environments. Their goal is to understand better how human decision-makers combine their own contextual information or intuition with machine generated predictions.   Grant funds will allow Agarwal and Salz to develop theoretical models of human decision-making with and without AI assistance, then test these models by running experiments on how human experts actually make use of AI tools in practice. The team will initially test their models through observing how radiologists interpret patients’ chest X-rays, varying the availability and timing of AI predictions and the presence or absence of contextual data such as the patients’ clinical histories. This will allow the team to explore, to take one example, the weight that radiologists give to AI predictions under different circumstances. The findings of this project, however, will have implications that go far beyond the practice of radiology, including potential further applications concerning the use of AI in financial transactions, corporate operations, and risk assessments.

This grant provides ongoing support for Stefano DellaVigna at the University of California, Berkeley, and Eva Vivalt at the University of Toronto, who are scaling up their Social Science Prediction Platform (SSPP), an online platform for collecting and cataloguing forecasts about the results of social science experiments. Documenting such forecasts is an increasingly used and useful way to help evaluate the importance of social scientific studies. Among other reasons, it creates a baseline from which to measure the novelty or unexpectedness of a social scientific result or finding. It can also serve as a useful measure of scientific consensus around important or contested issues in a field. Grant funds will allow DellaVigna and Vivalt to include more research projects in the platform, to include more than 5,000 new forecasts, and to include new applications for predictions such as measuring the effectiveness of policy interventions. Funds will enable the team to run conferences and workshops; to produce training materials and outreach activities; to recruit a large and diverse sample of forecasters; and to develop new methodologies and platform capabilities.

This grant supports Sanjay Patnaik, director at the Brookings Center on Regulation and Markets (CRM), which promotes independent, unbiased, and non-partisan research on regulatory economics. Grant funds will allow CRM to conduct research on topics including artificial intelligence and emerging technologies, financial market regulation and fintech, and consumer protection and antitrust in the digital age. Through its research, CRM seeks to explore how novel technologies can best be regulated without stifling innovation; how the regulatory process could be adapted to quickly respond to changing market conditions; and how regulation can deal with new approaches to data privacy. In addition to producing high-quality and policy-relevant research on regulatory economics, CRM will disseminate this work through academic papers, policy briefs, and events that connect researchers with practitioners.