Study Reveals Age Bias in Popular AI Chatbots Despite Efforts to Reduce Gender Discrimination
Key Takeaways
- ▸Five major chatbots show significant age bias despite reducing gender stereotypes in their responses
- ▸AI systems demonstrate a double standard: politically correct on gender while readily assigning age-based stereotypes
- ▸Age bias appears to stem from societal attitudes rather than technical limitations—developers have internalized anti-sexism but not anti-ageism
Summary
A new international study led by researchers at Universitat Oberta de Catalunya has found that five of the most popular AI chatbots—ChatGPT, Gemini, Copilot, Jasper, and Perplexity—exhibit significant age-related biases despite successfully reducing gender stereotypes. The research, published in Big Data & Society, used qualitative interviews with the chatbots to assess how they reproduce social stereotypes, treating AI models as conversational participants and analyzing their responses to questions about fictional characters' digital habits and age/gender assignments.
The study reveals a notable double standard: while the chatbots demonstrate "political correctness" by avoiding sexist assumptions and stereotypical gender roles, they readily assign age profiles and abilities based on users' digital behaviors. For instance, heavy Instagram or TikTok users are consistently categorized as younger, while Facebook political debate followers are assigned older age categories. Researcher Mireia Fernández-Ardèvol suggests this disparity reflects broader societal attitudes, noting that AI developers have internalized the wrongness of sexism but not ageism, resulting in systems that absorb and perpetuate age discrimination during training and deployment.
- Chatbots associate social media platforms (Instagram, TikTok) with youth and traditional platforms (Facebook) with older age groups
Editorial Opinion
This study highlights a critical blind spot in AI safety and bias mitigation efforts. While the AI industry has made commendable progress in reducing gender bias—a highly visible issue—the persistence of age discrimination reveals that bias work remains incomplete and often reactive to public pressure rather than comprehensive. The findings suggest that truly fair AI requires active effort to identify and address all forms of stereotyping, not just those that have achieved cultural consensus about their wrongfulness.
