The Bias Lens

Medical research shapes how we treat patients, but too often it’s built on biased data. The Bias Lens scores gender bias in medical studies, helping professionals assess its quality and applicability.

Comment

Inspiration

We were inspired by the growing recognition that medical research often fails to represent women and minorities. Some case studies we looked at included hip replacements that are 29% more likely to fail for women compared to men, and lack of autism research on girls resulting in 80% of cases not being diagnosed.

What it does

The Bias Lens is a web app that scans medical papers and outputs a gender bias score. It gives clinicians and researchers a holistic view of how gender balanced a study really is, so they can quickly decide whether it’s reliable enough to inform patient care. It provides justification on its score, giving context from the research article including its sample representation, inclusion in analysis, study outcomes, and methodological fairness.

How we built it

Frontend: React/Next.js, TypeScript, Tailwind CSS APIs: PubMed E-Utilities API for papers, Gemini API for language analysis with grounded context Backend/Auth: Firebase and Auth0 Other tools: GitHub, Google AI Studio, prompt engineering

Challenges we ran into

  • Many journals block automated access with robots.txt, limiting the dataset.
  • Balancing ethical use of LLMs while keeping results accurate.
  • Designing clear scoring metrics for bias that professionals can trust.
  • Ensuring the app outputs results in a way that is easy to interpret.

What's next for The Bias Lens

  • Partner with more journals and medical organizations to expand data access.
  • Broaden beyond gender bias to other categories, such as race and age.
  • Add user accounts and a database so professionals can track their analyses.

Built With

Share this project:

Updates