The rise of artificial intelligence (AI) has fueled concerns about potential threats of AI on elections, particularly after instances like an AI-generated robocall imitating President Joe Biden.
However, a recent gathering in New York organized by investigative journalist Julia Angwin and former White House science advisor Alondra Nelson brought scholars, election officials, and journalists together to explore AI’s potential impact on the 2024 election.
The event, part of their new initiative, AI Democracy Projects, aimed to assess the capabilities of AI language models in responding to election-related queries.
Participants were divided into groups to test four AI language models, or chatbots, on prompts related to elections. The focus was on assessing the quality and differences in responses, with prompts ranging from voter registration to finding polling locations. While not providing a comprehensive evaluation, the initiative represented a starting point for understanding AI’s influence on election processes.
AI on elections Scale of Threats
Concerns about the potential scale of AI threats to elections were acknowledged during the event. Quinn Raymond, co-founder of Voteshield, emphasized that the threat lies in the ease of entry and increased verisimilitude, enabling a small number of motivated individuals to cause significant damage. The consensus was that while the threat is not entirely new, AI lowers the barrier to entry, making election disruption more accessible.
The Brookings Institute’s recent explainer on AI’s possible impact on elections aligns with the event’s conclusions. The consensus suggests that AI-driven election disruption relies on traditional tactics but benefits from lower barriers to entry and increased verisimilitude.
Despite the potential for harm, participants noted that AI language models, at least in their current state, demonstrated limitations and often provided unhelpful or inaccurate responses.
AI’s Limitations and Room for Collaboration
During the tests, AI language models were found lacking, sometimes delivering answers of questionable accuracy. Despite concerns about the evolving technology, participants expressed optimism about collaboration between election officials and AI experts.
The need for accurate election information was emphasized, highlighting that AI’s current focus on “guessing” does not align with the precision required for election-related queries.
The interaction between election officials and AI experts showcased a model for effective collaboration, instilling confidence in our ability to address the shortcomings of AI models and counteract the growing threat of disinformation. The event served as a positive step, demonstrating the potential for interdisciplinary collaboration in navigating the challenges posed by AI in election processes.
Facing the Inevitability of AI Challenges
Acknowledging that incidents like AI-generated robocalls are inevitable, participants stressed the importance of proactive responses and updates to technology policies to address real-world conditions.
Some states are already taking measures such as requiring explicit labels on images created using AI. The event’s collaborative approach signals a promising way forward, urging legislators to consider similar interdisciplinary strategies in their efforts to address AI’s impact on elections.
Amid concerns about the perceived inaccessibility of AI, participants encouraged active engagement with the technology. While recognizing the fears associated with AI, there was a consensus that understanding its pros and cons requires direct interaction. Encouraging individuals to experiment with AI tools like ChatGPT and Microsoft’s Image Creator was seen as essential for consumers of information to comprehend the tangible impacts of AI on various aspects of our lives.
In conclusion, the collaborative approach taken during the event reflects a constructive way to navigate the challenges posed by AI in elections. The focus on testing AI language models, acknowledging their limitations, and fostering collaboration provides a foundation for addressing the evolving landscape of AI’s influence on democratic processes.