A study conducted by computer and information science researchers from the UK and Brazil has raised concerns about the objectivity of ChatGPT.
The researchers claim to have discovered substantial political bias in ChatGPT’s responses, leaning towards the left side of the political spectrum.
Published in the journal Public Choice this week, the study – conducted by Fabio Motoki, Valdemar Pinho, and Victor Rodrigues – argues that the presence of political bias in AI-generated content could perpetuate existing biases found in traditional media.
The research highlights the potential impact of such bias on various stakeholders, including policymakers, media outlets, political groups, and educational institutions.
Utilising an empirical approach, the researchers employed a series of questionnaires to gauge ChatGPT’s political orientation. The chatbot was asked to answer political compass questions, capturing its stance on various political issues.
Furthermore, the study examined scenarios where ChatGPT impersonated both an average Democrat and a Republican, revealing the algorithm’s inherent bias towards Democratic-leaning responses.
The study’s findings indicate that ChatGPT’s bias extends beyond the US and is also noticeable in its responses regarding Brazilian and British political contexts. Notably, the research even suggests that this bias is not merely a mechanical result but a deliberate tendency in the algorithm’s output.
Determining the exact source of ChatGPT’s political bias remains a challenge. The researchers investigated both the training data and the algorithm itself, concluding that both factors likely contribute to the bias. They highlighted the need for future research to delve into disentangling these components for a clearer understanding of the bias’s origins.
OpenAI, the organisation behind ChatGPT, has not yet responded to the study’s findings. This study joins a growing list of concerns surrounding AI technology, including issues related to privacy, education, and identity verification in various sectors.
As the influence of AI-driven tools like ChatGPT continues to expand, experts and stakeholders are grappling with the implications of biased AI-generated content.
This latest study serves as a reminder that vigilance and critical evaluation are necessary to ensure that AI technologies are developed and deployed in a fair and balanced manner, devoid of undue political influence.
(Photo by Priscilla Du Preez on Unsplash)
See also: Study highlights impact of demographics on AI training
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with Digital Transformation Week.
Explore other upcoming enterprise technology events and webinars powered by TechForge here.
