Examining Age Bias in ChatGPT Responses to Cancer Decision-Making Questions
Author Information
Author(s): McDarby Meghan, Mroz Emily, Hahne Jessica, Carpenter Brian, Parker Patricia
Primary Institution: Memorial Sloan Kettering Cancer Center
Hypothesis
Does ChatGPT exhibit age bias in its responses to decision-making questions about advanced cancer?
Conclusion
ChatGPT responses may show age bias, with younger patients receiving more mentions of clinical trials compared to older patients.
Supporting Evidence
- ChatGPT responses more frequently mentioned consideration of clinical trials for younger patients.
- There was a non-significant trend toward suggesting second opinions more for younger patients.
- No age differences were found in mentions of hospice care or palliative care.
Takeaway
The study looked at how ChatGPT answers questions about cancer based on a person's age, finding that it tends to suggest more options for younger people.
Methodology
Responses from ChatGPT were analyzed based on variations in hypothetical patient characteristics, and three coders categorized the responses.
Potential Biases
Potential bias in AI responses based on user demographics was not fully explored.
Limitations
The study only examined a limited set of patient characteristics and may not represent all demographics.
Statistical Information
P-Value
0.01
Statistical Significance
p=0.01
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website