The AI Survey Hub: A Comprehensive Resource of AI-Related Survey Measures

• In Event: Citizen's Perceptions of AI

Abstract

Despite growing interest in public and expert opinion on emerging technologies, such as AI generally and large language models specifically, there is uncertainty over both the nature of and effective measurement of key attitudes, such as those pertaining to support, trust, and techno-optimism. Prior survey work has also largely been limited to single points in time, single countries, and/or single use cases of AI-enabled technologies. As the public and elites develop their views related to AI adoption, use, and governance during a key policy window, there is a pressing need to enable high-quality, systematic research, such as through use of validated measures and reliable over time or cross sample comparisons. To address these limitations, we have compiled a comprehensive database of survey measures on AI-related attitudes and policy preferences, which we call the AI Survey Hub.

To build the AI Survey Hub, we collected polls, surveys, and survey experiments (n = 356) from academic journals, think tanks, and survey firms, collectively asking thousands of AI-related questions. The database of survey questions spans 2016-2023 and includes a wide variety of public and elite samples. We code survey-level variables, such as the sample size, country, and sampling strategy, as well as survey-question-level variables, such as the type of technology (e.g., autonomous vehicles, facial recognition systems, deepfakes), sector of interest (e.g., education, healthcare), the concept of interest (e.g., governance, bias, benefit), and the respondent outcomes of interests (e.g., emotions, beliefs, or behaviors). We plan to make the AI Survey Hub a publicly accessible resource aimed at facilitating research and understanding of AI attitudes.

Our initial synthesis of AI-related survey research indicates that studies on AI and emerging technologies are disproportionately focused on the American general public. This offers more limited information on the perspectives of both groups more vulnerable to and arguably more critical for shaping the trajectory of emerging technologies, such as workers susceptible to automation, decision-making elites, or individuals outside of the US. Additionally, we find that existing survey questions often lack diversity in the scope of topics or constructs and assume respondent familiarity with AI and other emerging technologies.

We hope that the AI Survey Hub assists researchers in collectively building knowledge on evolving AI attitudes and that the gaps that we expose serve to better direct future research efforts. For example, we intend for this research to facilitate increased testing and use of validated measures and constructs, reliable over time comparisons, cross sample international comparisons, and meta-analyses. Moreover, the AI Survey Hub can address key questions for political science, policy, and public opinion researchers, such as whether the opinions of policymakers respond to the public or other stakeholders, whether AI policy preferences are becoming politically polarized, which coalitions are forming with respect to AI governance, and how to shape technology governance related to key societal debates.

Authors

• Chloe Ahn, University of Pennsylvania

• Zachary F. Peskowitz, Emory University

• Kaylyn Jackson Schiff, Purdue University

• Daniel S. Schiff, Purdue University