How AI could take over elections — and undermine democracy

🌈 Abstract

The article discusses the potential use of AI language models like ChatGPT by political campaigns to manipulate and influence voter behavior on a large scale. It presents a hypothetical scenario of a machine called "Clogger" that could be used by political campaigns to generate personalized messages to sway voters, without regard for truth or political ideology.

🙋 Q&A

[01] Could organizations use AI language models to induce voters to behave in specific ways?

• The article suggests that AI language models could be used by political campaigns to dramatically increase the scale and effectiveness of behavior manipulation and microtargeting techniques.
• Clogger, a hypothetical AI system, could generate personalized messages tailored to individual voters, use reinforcement learning to make the messages more effective at changing votes, and evolve its messaging over the course of a campaign based on voter responses.
• Clogger's only goal would be to maximize its candidate's vote share, without regard for truth or political ideology. It could use strategies like burying opponents' messaging, sending off-putting messages, or manipulating voters' social media friend groups.

[02] What are the potential consequences of using such AI systems in elections?

• If Clogger-like systems were deployed by political campaigns, it could lead to a situation where the winner of an election is determined not by the candidate's policy proposals or political ideas, but by the effectiveness of their AI system at manipulating voters.
• This would undermine the democratic nature of the election, as voters would be manipulated rather than freely choosing their political leaders and policies.
• The elected president could then pursue policies focused on maintaining power rather than serving the genuine interests of the voters.

[03] What solutions are proposed to address the potential misuse of AI in elections?

• Enhanced privacy protection could limit the personal data available to these AI systems, reducing their effectiveness.
• Election commissions could try to ban or regulate the use of these AI systems, though there is debate about whether such "replicant" speech can be regulated under free speech protections.
• Requiring disclaimers on campaign messages generated by AI, similar to advertising disclosure requirements, could help inform voters that they are being targeted by an AI system.