[Paper Review] Fairness via AI: Bias Reduction in Medical Information

Shiri Dori-Hacohen, Roberto E. Montenegro+5 more|arXiv (Cornell University)|Sep 6, 2021

Ethics in Clinical Research15 references4 citations

TL;DR

This paper introduces a novel 'Fairness via AI' framework to proactively detect and mitigate biased health information—termed 'bisinformation'—using AI in recommender systems and online content platforms. By focusing on root causes of societal inequities in medical information rather than merely debiasing AI models, the approach aims to reduce disparities in health outcomes through interdisciplinary collaboration and targeted computational methods.

ABSTRACT

Most Fairness in AI research focuses on exposing biases in AI systems. A broader lens on fairness reveals that AI can serve a greater aspiration: rooting out societal inequities from their source. Specifically, we focus on inequities in health information, and aim to reduce bias in that domain using AI. The AI algorithms under the hood of search engines and social media, many of which are based on recommender systems, have an outsized impact on the quality of medical and health information online. Therefore, embedding bias detection and reduction into these recommender systems serving up medical and health content online could have an outsized positive impact on patient outcomes and wellbeing. In this position paper, we offer the following contributions: (1) we propose a novel framework of Fairness via AI, inspired by insights from medical education, sociology and antiracism; (2) we define a new term, bisinformation, which is related to, but distinct from, misinformation, and encourage researchers to study it; (3) we propose using AI to study, detect and mitigate biased, harmful, and/or false health information that disproportionately hurts minority groups in society; and (4) we suggest several pillars and pose several open problems in order to seed inquiry in this new space. While part (3) of this work specifically focuses on the health domain, the fundamental computer science advances and contributions stemming from research efforts in bias reduction and Fairness via AI have broad implications in all areas of society.

Motivation & Objective

• To address systemic biases in medical information dissemination that disproportionately harm minority populations.
• To propose a shift from 'Fairness in AI' to 'Fairness via AI'—using AI not just to correct bias in models, but to actively root out societal inequities at their source.
• To introduce and define 'bisinformation' as a distinct, harmful category of biased, yet often factually true, medical content that perpetuates health disparities.
• To advocate for interdisciplinary collaboration between computer science, medicine, sociology, and public health to study and mitigate bisinformation at scale.
• To identify key research gaps and open problems in detecting, measuring, and countering bisinformation in online health ecosystems.

Proposed method

• Propose a new framework—Fairness via AI—where AI is used to study, detect, and remediate structural inequities in health information systems.
• Introduce 'bisinformation' as a distinct category from misinformation: information that is factually accurate but embedded with harmful biases (e.g., racial or gendered assumptions) without structural context.
• Leverage NLP and machine learning techniques to scale the detection of biased language and imagery in medical curricula and online health content.
• Apply population-sensitive modeling to understand differential dissemination of bisinformation across demographic groups, informed by prior work on online controversy distribution.
• Integrate insights from medical education, antiracism theory, and sociology to ground AI interventions in real-world structural inequities.
• Use recommender system analysis to identify how biased content is amplified in search and social media platforms, particularly during public health crises like the COVID-19 infodemic.

Experimental results

Research questions

• RQ1 What societal problems related to health information are most amenable to intervention through Fairness via AI?
• RQ2 Which existing or novel AI techniques are most effective for detecting and mitigating bisinformation in medical content?
• RQ3 How can interdisciplinary collaboration be systematically fostered to enhance the ethical and societal impact of fairness research in AI?
• RQ4 How and where does bisinformation spread online, and is its dissemination distributed unequally across different population groups?
• RQ5 Which types of bisinformation and misinformation are most harmful and should be prioritized for fact-checking and countermessaging?

Key findings

• The term 'bisinformation' captures a critical but under-researched form of biased health information that is factually correct but socially harmful due to lack of structural context.
• Medical education continues to perpetuate harmful biases—such as using race as a proxy for genetics—without addressing Social and Structural Determinants of Health (SSDoH).
• Bisinformation is systematically disseminated through online platforms, including search engines and social media, where AI-driven recommenders amplify biased narratives.
• There is a significant gap in large-scale computational studies of bisinformation, especially in health contexts, despite its documented impact on minority health outcomes.
• The Fairness via AI framework offers a more impactful alternative to traditional Fairness in AI by targeting the root causes of inequity rather than just correcting AI outputs.
• The framework is supported by emerging research grants, such as one from the National Board of Medical Examiners’ Stemmler Fund, validating its feasibility and relevance in real-world medical education settings.