“Promoting AI’s Safe Usage for Elections” – Fall 2024

The utilization of artificial intelligence (AI) in electoral processes is rapidly evolving, presenting both opportunities and challenges. Our contributed book, “PROMISE – PROMoting aI’s Safe usage for Elections”, explores the various dimensions of AI’s role in safeguarding and enhancing electoral integrity. It is being published by Spring Nature. Below you will find the table of contents and the abstracts.

Table of contents

Foreword by Michael Huhns, University of South Carolina

Part 1: Introduction: Elections – Status Today from Multiple Perspectives

Chapter 1a: Elections: The Information Technology (IT) and Artificial Intelligence (AI) Perspective
Biplav Srivastava, University of South Carolina

Chapter 1b: Securing the Artificial Intelligence (AI)-enabled Election Software Stack
Ben Lazarine, Sagar Samtani, and Ramesh Venkataraman

Chapter 1c: AI and Elections: The Narrative Approach
Andrea Hickerson, University of Mississippi

Chapter 1d: Perspectives of electronic election systems
Tarmo Koppel and Tanel Tammet, Tallinn University of Technology

Chapter 1e: AI Red Teaming for Elections
Anita Nikolich, School of Information Sciences at the University of Illinois-Urbana Champaign

Part 2: Challenges in Elections

Chapter 2a: Nothing but the Truth: Voting in the Time of AI
Joan Zaleski, Ph.D., League of Women Voters of South Carolina

Chapter 2b: Funding and Procurement for Government AI Research and Innovation
Ott Velsberg, Chief Data Officer of Estonia

Part 3: Resources and AI Techniques for Elections

Chapter 3a: A Dataset and Visualization of Generalizable Election-Related Questions Compiled from Leading Global Democracies for Building AI-Enabled Tools
Kausik Lakkaraju, Bharath Muppasani, Sara Elizabeth Jones, Biplav Srivastava, AI Institute, University of South Carolina

Chapter 3b: Creating Frameworks and Datasets at the Intersection of AI Safety and Elections
Alphaeus Dmonte (1), Yuxia Wang (2), Preslav Nakov (2), Haonan Li (2), Xudong Han (2), Marcos Zampieri (1), Kevin Lybarger (1), Massimiliano Albanese (1), Timothy Baldwin (2); 1-George Mason University, 2-Mohamed Bin Zayed University of Artificial Intelligence

Chapter 3c: Verification AI in the Newsroom: A Cross-Cultural Study of Journalists’ Use of Deepfake Detection Tools
Saniat Javid Sohrawardi, Y. Kelly Wu, Matthew Wright, Rochester Institute of Technology

Part 4: Perspectives on AI / data-driven technology usage

Chapter 4a: The Brazilian Elections: a Brief History, Recent Events, and Suggestions for Moving Forward
Eduardo Lopes Cominetti, Escola Politecnica da Universidade de São Paulo

Chapter 4b: The Use of Technology in Indian Elections with a Special Emphasis on the Use of Artificial Intelligence
Arvind Gupta, Digital India Foundation and Aakash Guglani, Digital India Foundation

Chapter 4c: Towards Better Elections: A Discussion about the United Kingdom and Africa
Aurelia Ayisi, Deepak P, Marquita Smith, Biplav Srivastava, Anita Nikolich, Andrea Hickerson, Tarmo Koppel, Kausik Lakkaraju

Prof. Aurelia Ayisi, University of Ghana and Prof. Marquita Smith, University of Mississippi

Part 5: Best Practices and Use Cases

Chapter 5a: The Power of AI to Strengthen Civic Engagement
Marquita Smith, University of Mississippi

Chapter 5b: Technology for Redistricting: Assessing and Generating District Maps
Matthew J. Saltzman, Clemson University, Anna Marie Morra, Epic Systems Corporation and Blake Splitter, Tennessee Wesleyan University

Chapter 5c: Voting Methods to Prevent Social Disappointment in Elections
Mohammad Ali Javidian, Appalachian State University, Pooyan Jamshidi, University of South Carolina, Rasoul Ramezanian, Ferdowsi University of Mashhad, Marco Valtorta, University of South Carolina

Part 6: AI Code of Ethics for Conduct of Trustworthy Elections

Chapter 6: A Unified Code of Ethics and Conduct for AI and Trustworthy Elections
Anita Nikolich, School of Information Sciences at the University of Illinois-Urbana Champaign

---

FOREWORD

Foreword by Michael Huhns, University of South Carolina

ABSTRACTS

Part 1: Introduction: Elections – Status Today from Multiple Perspectives

Chapter 1a: Elections: The Information Technology (IT) and Artificial Intelligence (AI) Perspective

Biplav Srivastava, University of South Carolina

Abstract: In this chapter, we will give a brief introduction to Artificial Intelligence (AI), situate it in the wider context of Information Technology (IT) and discuss how it could affect the election process. With more and more election related processes becoming IT and AI enabled, a natural question to ask is how advanced their AI is and if they pose risks to stakeholder trust. We will describe a framework to assess AI capability of a system that is simple to apply and communicate with technical and non-technical stakeholders. Finally, we will relate the concepts of AI in the context of elections and discuss its implications. This chapter will, thus, help the reader understand the balance between the benefits of AI, its maturity, and how its risks can be mitigated for elections and elsewhere.

Chapter 1b: Securing the Artificial Intelligence (AI)-enabled Election Software Stack

Ben Lazarine, Indiana University
Sagar Samtani, Indiana University
Ramesh Venkataraman, Indiana University

Abstract: In recent years, political organizations’ operations have been increasingly impacted by Artificial Intelligence (AI), as they leverage it to perform data-driven campaigning (e.g., custom-generated content for constituents). The adoption of AI in election campaigns has been facilitated by open-source software (OSS). AI-enabled election software stacks incorporate machine learning OSS (MLOSS) tools to facilitate critical analysis for data-driven campaigning. While OSS has made AI-enabled election software stacks accessible to more government organizations, it has also introduced a new set of security issues. In this chapter, we review campaign software OS AI assets, their vulnerabilities, and how associated risks can be mitigated. We present a generalized framework for securing AI-enabled election software stacks and illustrate the analysis it enables through five example OS AI campaign tools. The framework consists of identifying critical software stack assets, selecting appropriate vulnerability assessment tools, and understanding the risks that identified vulnerabilities pose. Lastly, we demonstrate the value of our framework through a case study, examining two organizations (the Hungarian National Government and the National Hispanic Voter Educational Foundation) that have used OS AI tools. The case study illustrates critical vulnerabilities identified in the technologies both organizations adopted, including code injection, and highlights their potential impact.

Chapter 1c: AI and Elections: The Narrative Approach

Andrea Hickerson, University of Mississippi

Abstract: Although the concept of “narrative intelligence,” or the ability to comprehend the impact and influence of narratives public behavior and opinion is relatively new, there is a rich research tradition related to narratives that can inform how AI enables and challenges what we can and ought to know about narrative circulation and reception. Specifically, AI can be useful to help understand what claims circulate in and across narratives, as well as in identifying who circulates what narratives. Less common and more difficult, AI can be used in planning and modeling narrative circulation and evaluating audience impact and reception. Regardless of application, AI brings new tools to the study of narratives, but understanding the role of narratives in elections nevertheless requires a robust, holistic, and interdisciplinary approach. This chapter brings together research from journalism, communication, international relations and computing to describe and set a research agenda for how AI impacts election narratives.

Chapter 1d: Perspectives of electronic election systems

Tarmo Koppel, Tallinn University of Technology
Tanel Tammet, Tallinn University of Technology

Abstract: As elections are fundamental to democratic governance, the integration of e-voting and internet voting systems represents a significant evolution in the electoral landscape. These technologies offer increased convenience and accessibility, allowing citizens to vote remotely and at any time, thus potentially increasing voter turnout, particularly among specific groups such as abstainers and occasional voters. However, the implementation of e-voting systems is not without challenges, encompassing technical, political, social, and legal dimensions. This chapter emphasizes the importance of a comprehensive approach to e-voting, recognizing that it extends beyond a purely technical issue. To address the challenges inherent in electronic voting, the chapter proposes several risk mitigation strategies, including the involvement of a broad range of specialists, system duplication, and the use of open-source software. In-house development of e-voting software, coupled with independent auditing and a minimalist design approach, is recommended to maintain control over security practices and ensure system transparency. The chapter argues that these measures are crucial in building a secure, trustworthy, and reliable electronic voting system that upholds the democratic principles of fairness, transparency, and public confidence. The chapter asserts that while digital voting systems offer superior security, efficiency, and convenience compared to traditional methods, they require a robust governance framework and meticulous risk management to ensure their successful integration into modern electoral processes. The adoption of these strategies will not only minimize the risks associated with e-voting but also reinforce the integrity and legitimacy of democratic elections in the digital age.

Chapter 1e: AI Red Teaming for Elections

Anita Nikolich, School of Information Sciences, University of Illinois-Urbana Champaign

Abstract: The growing integration of artificial intelligence (AI) in the ecosystem of election technologies necessitates the incorporation of AI considerations into traditional security assessments. AI introduces new complexities and risks that must be addressed using both traditional and emerging AI security methods. AI Red Teaming is a new field that focuses on proactively identifying and mitigating potential AI-related vulnerabilities by simulating adversarial attacks and stress-testing AI systems to ensure their robustness and reliability.

Part 2: Challenges in Elections

Chapter 2a: Nothing but the Truth: Voting in the Time of AI

Joan Zaleski, Ph.D., League of Women Voters of South Carolina

Abstract: This chapter addresses strategies the League of Women Voters, a non-partisan political organization, recommends for voters to be able to sift through the election information, to make informed decisions, and to vote with confidence. Recognizing election mis/disinformation can help voters restore their trust in the election process.

Chapter 2b: Funding and Procurement for Government AI Research and Innovation

Ott Velsberg, Chief Data Officer of Estonia

Abstract: This chapter explores the role of funding and procurement in supporting AI implementation in public sector services worldwide. As governments aim to enhance efficiency, consistency, and resource allocation, ethical considerations and robust oversight become crucial. Examining AI use cases—such as facial verification tools and prediction systems—reveals significant risks, including racial biases, false arrests, and privacy breaches. To counter these, the chapter discusses different approaches and strategic funding models, emphasizing the need for AI-specific procurement guidelines and adaptable funding to encourage innovation while managing risks. Estonia’s e-government practices serve as a case study for effective AI adoption through collaborative efforts with academia and the private sector. The chapter outlines strategic recommendations for procurement, flexible funding, and community-driven innovation through open data. It also underscores the benefits of public-private partnerships and real-world testing environments. This chapter offers insights for both governments and organizations that seek to implement trustworthy AI solutions that align with societal expectations and ethical standards.

Part 3: Resources and AI Techniques for Elections

Chapter 3a: A Dataset and Visualization of Generalizable Election-Related Questions Compiled from Leading Global Democracies for Building AI-Enabled Tools

Kausik Lakkaraju, AI Institute, University of South Carolina
Bharath Muppasani, AI Institute, University of South Carolina
Sara Elizabeth Jones, AI Institute, University of South Carolina
Biplav Srivastava, AI Institute, University of South Carolina

Abstract: Elections represent the unified voice of citizens in a country, empowering voters to shape their government and future. Understanding the concerns of stakeholders like voters, candidates, journalists, and election administrators is crucial for improving electoral participation and engagement. In our study, we collected 227 election-related questions from various online sources, reflecting diverse geopolitical contexts across 8 different democratic countries. We identified 116 questions that were asked frequently as per online sources across multiple regions and refined these into 85 parameterizable queries. We also built a user interface (UI), available at https://ai4society.github.io/election-dataset/, which categorizes these questions by country and allows individuals to submit election-related questions specific to their country. Our analysis reveals a notable disparity between the questions we collected and those typically found in official FAQs, highlighting gaps that could guide future research to improve electoral processes. This dataset can help developers and researchers build Artificial Intelligence (AI)-enabled tools to reduce information gaps in the election ecosystem globally.

Chapter 3b: Creating Frameworks and Datasets at the Intersection of AI Safety and Elections

Alphaeus Dmonte, George Mason University,
Yuxia Wang, Mohamed Bin Zayed University of Artificial Intelligence,
Preslav Nakov, Mohamed Bin Zayed University of Artificial Intelligence, Haonan Li, Mohamed Bin Zayed University of Artificial Intelligence,
Xudong Han, Mohamed Bin Zayed University of Artificial Intelligence,
Marcos Zampieri, George Mason University,
Kevin Lybarger, George Mason University,
Massimiliano Albanese, George Mason University,
Timothy Baldwin, Mohamed Bin Zayed University of Artificial Intelligence

Abstract: This chapter explores the intersection of Artificial Intelligence (AI) and elections, focusing on the critical challenge of ensuring safety in the age of Large Language Models (LLMs) including AI-generated misinformation and its impact on electoral integrity. This chapter is divided in two complementary parts. The first part describes Do-Not-Answer, a framework featuring a three-level hierarchical taxonomy of LLM risks including hallucination, bias, toxic language, and misinformation. Do-Not-Answer has been used to create datasets in multiple languages that serve to evaluate LLMs with respect to mitigation strategies, content filtering, and model alignment. The second part discusses the ElectAI taxonomy and dataset. ElectAI has been created to aid claim understanding with respect to election processes, equipment, and claims of fraud in both AI- and human-generated social media posts. The two parts combined present the reader with a comprehensive overview of both general and election-related AI safety issues along with strategies to address them.

Chapter 3c: Verification AI in the Newsroom: A Cross-Cultural Study of Journalists’ Use of Deepfake Detection Tools

Saniat Javid Sohrawardi, Rochester Institute of Technology,
Y. Kelly Wu, Rochester Institute of Technology,
Matthew Wright, Rochester Institute of Technology

Abstract: This chapter examines how journalists in the United States and Bangladesh perceive and utilize deepfake detection tools in their news verification workflows. Through a semi-structured, scenario-based role-play study, we investigate the factors influencing journalists’ adoption of these tools, their placement within the verification process, and the potential biases associated with their use. Our findings show that while journalists recognize the potential of deepfake detection tools, their usage is not yet standardized or universally embraced. The tools are often employed midway through the workflow, only after initial assessments based on traditional methods prove inconclusive. Factors influencing tool usage include uncertainty about content authenticity, the perceived importance of the news story, and explicit speculations of deepfakes. The study also revealed instances of automation bias and confirmation bias among journalists, emphasizing the need for a balanced approach that combines technological tools with traditional journalistic methods and critical thinking skills. We conclude by discussing the implications of our findings for the development and deployment of deepfake detection tools that are effective, reliable, and ethically sound.

Part 4: Perspectives on AI / data-driven technology usage

Chapter 4a: The Brazilian Elections: A Brief History, Recent Events, and Suggestions for Moving Forward

Eduardo Lopes Cominetti, Escola Politecnica da Universidade de São Paulo

Abstract: In this chapter, we present a brief history of the Brazilian Elections, the creation of the Brazilian voting machine, the Urna Eletronica, and the recent events and attacks on its integrity. We finish by shortly mentioning the current proposal to our Electoral Authority, the Tribunal Superior Eleitoral (TSE): the creation of an End-to-End verifiable voting system that is incremental to the currently used system.

Chapter 4b: The Use of Technology in Indian Elections with a Special Emphasis on the Use of Artificial Intelligence

Arvind Gupta, Digital India Foundation,
Aakash Guglani, Digital India Foundation

Abstract: The integration of technology in election processes has transformed democratic engagement in India. This chapter examines the evolution and impact of technological advancements on Indian elections, focusing on innovations like Electronic Voting Machines (EVMs), social media platforms, and Artificial Intelligence (AI). Initially, technology facilitated accurate voter registration and streamlined voting through EVMs. With the rise of social media, election campaigns shifted towards digital engagement, enhancing voter reach and interaction. In recent years, AI has emerged as a pivotal tool, offering benefits such as personalized voter outreach and sentiment analysis, while also presenting challenges like the spread of misinformation and deepfakes. This chapter highlights both the opportunities and risks associated with these technologies, providing insights into their role in shaping transparent, efficient, and inclusive election processes.

Chapter 4c: Towards Better Elections: A Discussion about the United Kingdom and Africa

Prof. Aurelia Ayisi, University of Ghana,
Prof. Deepak P, University of Belfast
Prof. Marquita Smith, University of Mississippi
Biplav Srivastava, University of South Carolina
Anita Nikolich, University of Illinois-Urbana Champaign
Andrea Hickerson, University of Mississippi
Tarmo Koppel, Tallinn University of Technology
Kausik Lakkaraju, University of South Carolina

Abstract: This article contains the observations of Dr. Deepak P., Prof. Aurelia Ayisi, and Prof. Marquita Smith. Dr. Deepak P. is a Senior Lecturer in Computer Science at the School of EEECS, specializing in AI ethics, social choice, and NLP. Prof. Aurelia Ayisi is a lecturer and researcher in the Department of Communication Studies at the University of Ghana, focusing on digital communication, media, and information literacy, with an emphasis on access, competencies, and global disparities. The interview was conducted in August 2024 to discuss the challenges of elections in Africa (especially Ghana) and the UK.

Part 5: Best Practices and Use Cases

Chapter 5a: The Power of AI to Strengthen Civic Engagement

Marquita Smith, University of Mississippi

Abstract: Many debate the benefits of using Artificial Intelligence (AI) to improve civic engagement and encourage more inclusive and participatory governance. This chapter explores how AI-driven platforms can be used to boost voter participation. In addition to enhancing voter engagement, AI can help counter misinformation by facilitating better communication between citizens and governments. This helps to support and sustain democratic societies. A review of existing literature highlighted the potential for AI to empower marginalized communities, bridge information gaps, and lower barriers to civic participation. However, ethical considerations such as data privacy and information bias are also addressed to ensure that AI’s role in civic engagement promotes accurate information for citizens.

Chapter 5b: Technology for Redistricting: Assessing and Generating District Maps

Matthew J. Saltzman, Clemson University,
Anna Marie Morra, Epic Systems Corporation,
Blake Splitter, Tennessee Wesleyan University

Abstract: In the United States, most political jurisdictions are based on geographic districts. After each decennial census, districts must be redrawn to account for changes in population and other demographics. “Gerrymandering” is the process of drawing maps with the intent to give an electoral advantage to a political party or candidate or to disadvantage an ethnic or social population. We describe mathematical and computational tools to detect gerrymandering, measure district characteristics, and generate maps with particular characteristics (including anti-gerrymandering measures such as fairness and competitiveness).

Chapter 5c: Voting Methods to Prevent Social Disappointment in Elections

Mohammad Ali Javidian, Appalachian State University,
Pooyan Jamshidi, University of South Carolina,
Rasoul Ramezanian, Ferdowsi University of Mashhad,
Marco Valtorta, University of South Carolina

Abstract: Mechanism design is concerned with settings where a policymaker (or social planner) faces the problem of aggregating the announced preferences of multiple agents into a collective (or social), system-wide decision. One of the most important ways for aggregating preference that has been used in multi-agent systems is election. In an election, the aim is to select the candidate who reflects the common will of society. Despite the importance of this subject, in some situations the result of the election does not respect the purpose of those who execute it and the election leads to dissatisfaction of a large amount of people and in some cases causes polarization in societies. To analyze these situations, we introduce a new notion called social disappointment and show which voting rules can prevent it in elections. In addition, we propose a new protocol to prevent social disappointment in elections. A version of the impossibility theorem is proved regarding social disappointment in elections, showing that there is no voting rule for four or more candidates that simultaneously satisfies avoiding social disappointment and Condorcet winner criteria. We give conditions under which our new protocol always selects the Condorcet winner under the assumption of single peakness. We empirically compare our protocol with eight well-known other voting protocols and observe that our protocol is capable of preventing social disappointment and is more robust against manipulations.

Part 6: AI Code of Ethics for Conduct of Trustworthy Elections

Chapter 6: A Unified Code of Ethics and Conduct for AI and Trustworthy Elections

Anita Nikolich, School of Information Sciences, University of Illinois-Urbana Champaign

Abstract: There is a critical need for a unified code of ethics and conduct for the development and application of Artificial Intelligence (AI) in electoral processes. As AI increasingly influences various aspects of elections, existing professional guidelines and ethical frameworks fall short in addressing the unique challenges at this intersection. Through a review and comparison of current codes from professional societies and election organizations, this study identifies the gaps in addressing AI-specific concerns within the electoral context and highlights the necessity for interdisciplinary collaboration among experts in cybersecurity, journalism, machine learning, technology, and election administration to safeguard democratic processes in the digital age. While individual fields operate under their own ethical guidelines, the interconnected nature of modern election security, driven by AI, demands a more cohesive approach.

Technology in Elections: Code of Ethics – The 7 Easy Reckoner

Promoting Computing: (ACM – https://www.acm.org/code-of-ethics)

• Contribute to society and to human well-being, acknowledging that all people are stakeholders in computing
• Maintain high standards of professional competence, conduct, and ethical practice

Promoting Communication: (https://www.spj.org/ethicscode.asp)

• Seek truth and report it
• Be accountable and transparent

Promoting Model Citizenship Responsibility

• Minimize harm
• Respect everyone’s view and give them space to express them
• Honor people and their free will to vote

For more details and rationale, see Chapter 6 – A Unified Code of Ethics and Conduct for AI and Trustworthy Elections.