A recent study conducted by computer and information science researchers from the UK and Brazil suggests that ChatGPT, a prominent chatbot powered by large language models (LLMs), may have a political bias, particularly leaning towards the left of the political spectrum. The st udy, Authored by Fabio Motoki, Valdemar Pinho Neto, and Victor Rodrigues and published on August 17 in the journal Public Choice, claims to provide "strong evidence" of ChatGPT's political bias.
The researchers raise concerns that the presence of factual errors and biases in LLM-generated texts like ChatGPT could mislead readers and exacerbate existing issues of political bias in traditional media. They contend that this bias could potentially have negative political and electoral repercussions similar to the bias present in conventional and social media.
The study employed an empirical approach, using questionnaires to interact with ChatGPT. The researchers tested the chatbot's responses using the Political Compass Test, which gauges individuals' political inclinations. The study also examined ChatGPT's responses in simulations of average Democrats and Republicans. The findings indicated that ChatGPT's algorithm exhibited a bias towards the responses of the US Democratic Party.
The researchers suggest that ChatGPT's political bias is not confined to American political contexts; it appears biased towards the US Democratic Party, Brazil's Lula, and the UK Labor Party. The study's authors acknowledge that identifying the precise source of the bias is challenging. They attempted to glean insight by trying to prompt ChatGPT into a developer mode, but the chatbot reiterated its impartiality.
The researchers posit two possible sources of bias—training data and algorithm design. They conclude that both factors likely contribute to ChatGPT's output bias and emphasize that disentangling these components for further study is essential.
While concerns about political bias are emerging, AI tools like ChatGPT face a range of other challenges and risks, including privacy concerns, ethical considerations, and potential misuse. The broader adoption of AI technologies has led to discussions about their regulation and the need for transparent and unbiased AI development.
