Research

The AI Governance Regulatory Archive (AGORA) database is a collaboration between GRAIL and the Center for Security and Emerging Technology (CSET) at Georgetown, with the purpose of compiling AI-related legislation into a single archive.

This project is building a database that includes survey questions about AI attitudes, opinions, and policy preferences across polls and surveys run by academic institutions, think tanks, and news organizations. This will assist in understanding how AI attitudes vary over time, across countries, and across subgroups of the population

The FAIRWORK project investigates how artificial intelligence is reshaping labor markets, job structures, and workforce policy.

This database includes deepfake images and videos about political actors, institutions, or events. We are compiling information about the political deepfakes such as whether the images/videos are presented as real/fake, whether there is external verification, who the targets are, what potential harms are depicted, any real-world evidence of harms, how they are spread on social media, and what framing/narratives they evoke.

This project will examine how police executives impact accountability practices and outcomes within law enforcement agencies. The research team will create comprehensive datasets on police chiefs, sheriffs, civilian review boards, and accountability mechanisms like body-worn cameras and AI systems for identifying misconduct.

This paper proposes a conceptual framework to clearly define and evaluate the effectiveness of AI ethics initiatives, outlining specific ethical aims, the activities designed to achieve these aims, and how they are interconnected.

This paper explores the demand for AI ethics and governance skills, examining job postings from 2018 to 2023 to identify trends and key skills needed in this field.

This study analyzes 112 documents addressing the ethical implications of AI from 25 countries (2016–2019), and finds significant differences across public, private, and NGO sources. Unlike private sector documents, those from NGOs and public institutions cover a broader range of ethical topics, engage more with law and regulation, and involve more participatory development processes, reflecting differing values and priorities.

By drawing on the advocacy coalition framework that emphasizes the centrality of beliefs amongst coalitions (or groups) of actors, this study investigates the deep core beliefs, policy core beliefs, and secondary beliefs of a subset of actors in AI policy (the citizens and civil society and advocacy groups).

This study investigates how national AI ethics guidelines in East Asia reflect distinct governance logics, institutional priorities, and normative frameworks through a comparative analysis of government-issued AI ethics guidelines in China, Japan, Singapore, and South Korea.

As AI becomes more integrated into society and the economy, its ethical implications are getting more attention. This article looks at current efforts in responsible AI, especially in businesses, and identifies six key factors causing a gap between principles and actions: conflicting incentives, the complexity of AI’s impacts, different fields not working together well, how responsibilities are divided in organizations, managing organizational knowledge about AI, and too many tools without clear choices.

This study analyzes whether public discourse on AI influences U.S. congressional messaging by examining five million AI-related tweets, as well as congressional statements from the 115th and 116th Congresses. Using text analysis and time series models, it finds that public discussion—especially on AI’s economic dimensions—can shape subsequent policymaker attention.

This article explores the factors influencing AI policy agenda-setting, focusing on different approaches to innovation policy and their implications for technology governance.

This paper examines potential drivers of AI policy politicization through a mixed-methods study focused on early AI policymaking across the U.S. states. We administer a survey to state legislators, resulting in a broadly representative sample of 129 policymakers from 44 states. Additionally, we perform four case studies featuring legislative debates over AI regulation in Idaho, Colorado, and Illinois.

This paper analyzes AI ethics documents and examines three key issues: challenges associated with the homogeneity of their creators, a typology of motivations behind their creation, and their varied impacts on the global AI governance landscape.

This chapter challenges the dominant conceptions of AI ethics education and encourages scholars, practitioners, and policymakers to consider the true versus intended outcomes of students at the heart of the sociological change expected by the AI ethics field.

This paper presents a systematic literature review and qualitative analysis of the early years of AI ethics education as a formalized field to analyze whether its future trajectory is aligned with educational best practices. Our review highlights core challenges in AI ethics education and the content, assessment, and pedagogy used in real interventions over recent years.

This study examines how automation risk affects worker well-being across 402 occupations using General Social Survey data from 2002–2018. It finds that while higher automation risk is associated with lower stress, it also correlates with worse health and little to no improvement in job satisfaction—highlighting the mixed effects of technological complementarity on workers.

The global discourse on artificial intelligence (AI) has surged in the past decade, prompting over 30 countries to establish national AI policy strategies by 2021. Despite extensive discussions on AI's ethical dimensions, a thematic analysis of 24 of these strategies reveals a significant gap in addressing AI in education (AIED) and its ethical implications. While policymakers prioritize AI for workforce preparation and expert training, they largely overlook the transformative potential and ethical considerations of AI in education.

This study investigates how digital transparency affects citizen participation in online crowdsourcing, focusing on Sacramento’s 311 platform. Using PSM-DID analysis, it finds that increased digital transparency boosts short-term participation, with effects varying across socioeconomic groups.

This study investigates how Computing Ethics courses should be taught in higher education programs in the U.S. by surveying 318 educators, focusing on their perceptions of instructional approaches and curricular integration.

This paper presents the results of a five-year, mixed-methods longitudinal study tracking a cohort of undergraduate students at the Georgia Institute of Technology in order to illustrate the key influences and inhibitors shaping social responsibility development among STEM undergraduates.

This study investigates the development of personal and professional social responsibility attitudes among undergraduate engineering students. It uses a mixed-methods approach to explore influences, including family, collegiate experiences, and the professional environment, shaping students’ microethical and macroethical perspectives.

AI in education, or AIEd, is a novel tool that holds the potential to revolutionize learning methods. It is speculated that AIEd could enhance the learning experience and promote fairness in education. However, concerns have been raised about its potential to introduce mechanical and detached aspects to learning. Unlike AI applications in other domains such as automotive or medical fields, the impact of AI on educational institutions has been a topic that has received relatively less attention. This paper evaluates the status of AIEd, with a focus on intelligent tutoring systems and anthropomorphized educational agents.

Ethical challenges in AI and social media have raised concerns about the adequacy of ethics education in undergraduate computer science programs. Despite calls for reforms, there's limited evidence on how well these programs foster professional ethics compared to other fields. This study compares the professional ethics attitudes of computing students to those in other disciplines and examines variations within computing based on demographics.

This study assesses 1,465 university students across Germany, the UK, and the US to measure AI literacy and related variables. We compare AI literacy, AI self-efficacy, interest in AI, attitudes towards AI, AI use, and prior learning experiences between countries. We provide an AI literacy test validated for cross-national research.

This study examines whether citizen engagement—through comments and follows on service requests—influences government responsiveness. Using data from 100 cities, it finds that collaborative input significantly increases the likelihood and speed of issue resolution, highlighting the impact of digital citizen feedback on public service delivery.

This project evaluates the effectiveness of AI ethics commitments on building public trust and reducing demand for state intervention.

Practitioners in public administration increasingly employ AI tools. This study delves into contextual factors influencing public reactions to AI in policing, exploring bureaucratic proximity, algorithmic targets, and agency capacity. Results show public favor for AI in local law enforcement, but preferences vary based on political affiliation, race, and the purpose of AI application.

In a comprehensive study focused on AI policy and involving over 7300 U.S. state legislative offices, researchers uncover intriguing insights into the dynamics of policymaker engagement. Contrary to expectations in highly technical domains like AI policy, narratives were found to be equally as effective as expert information in capturing the attention of state legislators.

This paper introduces and validates two measures (AI Dread and AI Controllability Concern) to explain public perceptions of AI risks. Using data from Canada and Japan, the authors find that trust in scientists, conspiracy thinking, and job impact concerns are significant predictors of AI attitudes across both contexts.

This study investigates how people think about and govern AI using surveys and statistical modeling. The authors find that people's views on AI's benefits influenced their attitudes towards its use, but not how it should be governed.

This paper introduces IEEE 7010, a new standard for assessing the well-being implications of artificial intelligence (AI).

This study uses a survey experiment to assess how the use of automated decision systems (ADS) in government affects public perceptions, focusing on the values of fairness, transparency, and human responsiveness. Findings show that perceived failures in fairness and transparency significantly reduce citizen support for government, regardless of policy area or political ideology.

This study investigates how politicians benefit from falsely labeling scandalous news as fake or deepfaked—known as the “liar’s dividend.” Across five survey experiments with over 15,000 Americans, it finds that such misinformation claims can boost politician support, especially for text-based scandals, outperforming other scandal response strategies like silence or apology.

This chapter examines the global proliferation of AI ethics documents, exploring why they are being produced and what they reveal about the motivations, practices, and policies surrounding AI.

This study aims to assess factors surrounding the adoption and legitimization of AI by international organizations, including by examining prominent case studies.