Research reveals AI's ability to influence political views - ForkLog: cryptocurrencies, AI, singularity, future

# Research Reveals AI’s Ability to Influence Political Views

Widely used AI systems can alter voters’ preferences by 15% under controlled conditions. This is stated in a study by Cornell University and the British Institute for Artificial Intelligence Security.

“There is significant public concern regarding the potential use of technology for political persuasion and its possible impact on elections and democracy. We are informing about these concerns by using pre-registered experiments to assess the ability of large language models to influence voter attitudes,” — experts noted.

Nearly 6,000 people from the USA, Canada, and Poland participated in the study. They evaluated a political candidate, interacted with a chatbot, and then re-evaluated.

In the American part of the study, 2,300 people participated before the 2024 presidential election. AI had an amplifying effect when participants’ opinions aligned.

If the chatbot supported a candidate that the participant did not support, “more noticeable shifts” occurred — people changed their attitudes toward the political figure more strongly.

Similar results were recorded in Canada and Poland.

The study showed that politically oriented messages had a greater persuasive effect than personality-based ones.

The accuracy of statements depended on the conversation. Chatbots supporting right-leaning candidates made more inaccurate claims compared to those supporting left-leaning candidates.

“These findings contain an uncomfortable assumption: AI political influence can exploit biases in what the models know, leading to uneven spread of inaccuracies even with explicit instructions to tell the truth,” — scientists noted.

AI Bias

In a separate study published in Science, experts attempted to understand why the persuasion effect occurs at all. They tested 19 language models on 76,977 adults in the UK on over 700 political issues.

“There is a widespread fear that conversational artificial intelligence will soon be able to exert unprecedented influence on human beliefs,” — the authors wrote.

They found that prompt engineering methods influence the convincing ability more than the size of the model. Prompts that encourage LLMs to introduce new information increased persuasiveness but reduced accuracy.

AI Is Not Neutral

Analytical centers noted that young conservatives are more willing to delegate AI the authority to make important government decisions compared to liberals.

Donald Kendall, director of the Glenn Haskins Center for New Issues at the Heartland Institute, mentioned that voters often misjudge the neutrality of large language models.

“One thing I try to convey to people is to dispel the illusion that artificial intelligence is impartial. It is clearly biased, in some cases passively,” — he added.

The expert stated that large Silicon Valley corporations developing AI “do not shy away from influencing what content is spread on their platforms.”

Recall that a simulated Federal Reserve meeting using AI agents showed that political pressure divided board members during the discussion of interest rates.