DEVELOPING AN EFFECTIVE ARTIFICIAL INTELLIGENCE GOVERNANCE FRAMEWORK
Restricted (Penn State Only)
- Author:
- Robles, Pedro
- Graduate Program:
- Public Administration
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- June 28, 2023
- Committee Members:
- Goktug Morcol, Program Head/Chair
Michele Tantardini, Major Field Member
Rhoda Joseph, Outside Unit & Field Member
Eric Best, Major Field Member
Dan Mallinson, Chair & Dissertation Advisor - Keywords:
- Information Technology Governance
Artificial Intelligence
Policy Diffusion
Conceptual Framework
Autonomous Vehicle - Abstract:
- Public organizations are increasingly adopting Artificial Intelligence (AI) without proper public policies and governance frameworks, giving rise to significant ethical and legal concerns. For example, the utilization of AI for facial recognition in airport security systems has resulted in instances where innocent individuals are mistaken for criminals. Moreover, AI and machine learning algorithms are being employed to expedite the hiring process, but these algorithms often exhibit biases by excluding certain genders or discriminating based on race, ethnicity, or skin color. Public organizations have implemented AI to provide services like chatbots, virtual information assistants, and public safety information. However, there is a growing concern among experts, public organization leaders, and the general public regarding the risks and potential harm posed by AI algorithms, which can be biased and prejudiced. Existing AI governance frameworks in the literature have limitations in addressing the political environment of public organizations. This study aims to bridge this research gap by developing an appropriate AI governance framework that fulfills the requirements of public organizations. Conducting exploratory research guided by academic theories is vital to achieving this objective. The outcomes of this research can contribute to addressing public policy concerns and alleviating public apprehensions, thereby enabling the adoption of AI for public service delivery. The proposed research will take a pragmatic approach and present an opportunity to further expand the epistemology of academic research. The initial challenge lies in the wide array of AI technologies that are labeled differently, such as machine learning, artificial neural networks, genetic algorithms, and natural language processing. Despite the numerous advancements and contributions of AI to public services, challenges have arisen in policymaking, defining the appropriate technology for public policy, and establishing aligned AI governance methods. Policymakers face specific issues that hinder their ability to develop suitable policies, grapple with fundamental concepts and definitions, and create laws that are adaptable to diverse AI environments. Chapter 2 delves into the complexity of algorithms, making them challenging to comprehend, particularly since many applications are new to the public sector. Governments across the globe are struggling with how to regulate rapidly emergent AI technology without stifling innovation. Scholars of AI governance have been likewise working in a fragmented manner to develop a theoretical framework to guide their work and inform government practice. These frameworks can serve as catalysts for researching the most effective AI governance strategies and as models for regulators. To foster a cohesive research program on AI governance, this systematic review identifies and examines each framework developed by researchers. This systematic review identifies 10 different frameworks that have emerged. While each is imbued with context, shared principles for AI governance can be identified. This review synthesizes these frameworks and offers a Unified AI Governance Framework that contains a set of core and peripheral principles which can guide a less fragmented research program. While Chapter 2 establishes an appropriate governance framework that promotes the credibility of public organizations and reinforces public trust in the government to harness the benefits and address the challenges of AI, Chapter 3 expands on the importance of public trust in AI algorithms. Advancement in information technology continues to evolve especially in the field of artificial intelligence (AI). Research studies have been conducted to evaluate the perceptions of Americans on the development and utilization of AI technology and if it is appropriate to use AI in public administrative duties. The research revealed that society is fragmented regarding the acceptance of AI, and whether AI decisions could have long term effects on the labor industry, legal system, and national security. The 2018 AI Public Opinion Survey revealed significant concerns among the American public regarding AI, yet also a recognition of its promise. The goal of this article is to further develop a governance framework for AI that considers the place of the public in AI policy. First, it presents an AI governance framework that pulls together disparate strands of other proposed frameworks. Then, highlighting the importance of public trust and support, it evaluates public opinion on AI technology that specifically pertains to governance. The article concludes with a discussion of future directions for AI governance research. Chapter 3 argues that public trust should be at the center of discussions on AI governance. It does so by highlighting public concerns regarding AI governance drawn from a national sample survey. Chapter 4 examines current public policy surrounding autonomous vehicle (AV) policies across state governments in the United States, with a particular focus on public policy diffusion. Notably, AVs have the potential to significantly enhance various sectors of the U.S. transportation industry, including trucking, public transportation, and other public services (i.e., food delivery and mail). A mixed methods study is conducted to better understand AV policy development and implementation. The quantitative research collected AV policy data from the National Council of State Legislature (NCSL) database, providing a solid foundation for a comprehensive mixed methods analysis. Additionally, qualitative data was gathered through interviews to obtain deeper insights. These interviews proved valuable in capturing information about the learning mechanisms that operate across various levels of government and domains, highlighting the crucial role of fostering policy learning between state governments and stakeholders. The analysis of interview transcripts was then compared with the quantitative data results, enabling a thorough exploration of AV policy development and implementation. The combined mixed research using qualitative and quantitative methods offers a robust understanding of how policymakers collaborate within and across states on AV policy, shedding light on policy diffusion mechanisms and contributing to the effective implementation of policies. In conclusion, this chapter presents comprehensive mixed research results that can assist academic researchers and practitioners address complex policy problems associated with autonomous vehicle policy, enabling them to adapt to changing circumstances and achieve desired policy outcomes.