The reliability of AI chatbots in providing election voting advice is under scrutiny, with a recent study revealing their inaccuracies and inconsistencies. This is a critical issue, especially as AI usage among the Hungarian population continues to grow. The research, conducted by the civil liberties group Liberties, highlights the potential dangers of relying on these systems for political guidance. The study focused on the Hungarian parliamentary elections, where AI chatbots were tested for their ability to recommend the right party for voters. The findings were alarming, to say the least.
One of the most concerning aspects was the overwhelming bias in party visibility. In 90% of cases, ChatGPT failed to recommend the opposition Tisza party, despite it being a significant force in the election. This bias could have potentially influenced the outcome, especially in a more competitive race. The researchers also noted that the chatbots often recommended smaller parties or parties not even running in the election, which is a clear sign of their unreliability. The study's author, Eva Simon, emphasizes the need for transparency and accountability in AI systems providing political advice.
The issue lies in the training data gaps and language-processing limitations of these AI models. Since Tisza emerged as a prominent party only after 2024, the static training models struggled to recognize it. This highlights the importance of regular updates and improvements in AI systems to ensure they remain accurate and relevant. Furthermore, the chatbots' responses often appeared confident and authoritative, making users more likely to trust them. However, the underlying methods are opaque, and the results are unstable, which raises serious concerns about their reliability.
The study also points out the regulatory gap in the EU's AI Act and Digital Services Act. These acts cover systemic risks to electoral processes, but AI chatbots fall between the two, creating a loophole. The researchers argue that safeguards should be developed to ensure the transparency, accuracy, and consistency of AI systems providing political advice. They also suggest that AI providers should not offer personalized voting recommendations unless they can guarantee these essential qualities.
In conclusion, the study's findings are a stark reminder of the potential pitfalls of relying on AI chatbots for political guidance. As AI usage continues to grow, it is crucial to address these issues to ensure the integrity of democratic processes. The study's authors emphasize the need for transparency, accountability, and regular updates in AI systems to provide reliable and accurate information to voters.
