Articles

When Algorithms Shape Access: Designing AI for Equity in Healthcare

When Algorithms Shape Access: Designing AI for Equity in Healthcare

By Racheal Hernandez, MAS

Introduction

Artificial intelligence is transforming industries at breakneck speed, from healthcare and finance to hiring and law enforcement. One of the most pressing ethical questions about AI in healthcare isn’t technical, but structural: Will AI help reduce inequities in care delivery, or unintentionally reinforce them?

Because AI is trained on massive datasets composed largely of existing human knowledge from books, articles, internet content, and social media its outputs are only as objective as the inputs. And much of that data reflects a historical bias shaped by a dominant cultural lens, often white, Western, and male.

For those working in diversity, equity, and inclusion (DEI), the implications are profound. If left unchecked, AI could not only mirror but reinforce systemic inequities. So how do we navigate the risks and unlock the potential of AI as a tool for equity rather than exclusion?

Bias In, Bias Out

At its core, AI works by learning from patterns. If those patterns reflect biased hiring practices, underrepresentation of minority voices in media, or criminal justice disparities, the AI system learns and perpetuates those patterns (Buolamwini & Gebru, 2018).

Healthcare algorithms are a stark example. For instance, an influential study found that an AI system used to allocate healthcare resources was less likely to refer black patients for additional care than white patients with the same medical conditions. This disparity occurred because the algorithm used healthcare cost as a proxy for health needs which reflects the systemic underinvestment in black communities rather than actual need (Obermeyer et al., 2019).

Who Writes the Rules?

Most AI systems are designed by a narrow demographic of technologists. According to a 2020 study by the AI Now Institute, over 80% of AI researchers are men, and the overwhelming majority are white (West et al., 2019).

When design teams lack lived experience with racism, ableism, or gender bias, they may fail to anticipate harmful impacts. What’s seen as “neutral” is often just a reflection of dominant norms.

The Myth of Objectivity

One of the greatest dangers in AI is the illusion of objectivity. When a system makes a recommendation on a job applicant or a loan approval it may seem more trustworthy because it's data-driven. But if the underlying data reflects decades of exclusion or prejudice, then the algorithm simply automates injustice (Noble, 2018).

Steps Toward Equity-Driven AI

While the risks are real, there are also opportunities. With intentional design, community input, and transparent oversight, AI can be made more equitable. Here’s how:

1. Diverse Data Curation
Ensure that training data includes a broad range of voices, experiences, and cultural perspectives. This includes language data, imagery, and behavioral models. Diverse data leads to more equitable algorithms.

2. Inclusive Design Teams
Recruit, hire, and empower technologists from marginalized communities. Inclusion at the table leads to better, fairer technology.

3. Audit for Bias
Conduct regular algorithmic audits to test for disparate impact. Independent, third-party reviews can identify risks that internal teams may miss.

4. Center Ethics in Development
Ethics should not be an afterthought. It must be embedded in the design process from ideation to deployment. Organizations should develop frameworks for responsible AI governance (Whittaker et al., 2018).

5. Listen to Affected Communities
Those most impacted by biased technology must have a voice in shaping and governing it. Participatory design models are critical.

6. Transparency and Accountability
Organizations must disclose how AI systems make decisions and allow users to challenge or appeal outcomes. Without transparency, trust breaks down.

Conclusion

AI is not inherently racist or inclusive but reflects the data, values, and choices of its creators. The question is not whether AI will shape society, but whose vision it will serve.

If we’re serious about equity, we must build technology that challenges historical bias rather than codifies it. This requires systemic change in how we train, hire, and hold accountable the institutions developing AI.

Used with intention, AI can help us identify and undo patterns of discrimination. But only if we lead with justice—not just code.

About the Author

Racheal Hernandez, MAS, is a Healthcare Administrator with over 20 years of experience in ambulatory care and healthcare leadership. Passionate about creating strong operational teams that deliver high-quality care.

References

• Buolamwini, J., & Gebru, T. (2018). Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification. Proceedings of Machine Learning Research, 81:1–15.
• Noble, S. U. (2018). Algorithms of Oppression: How Search Engines Reinforce Racism. NYU Press.
• Raji, I. D., & Buolamwini, J. (2019). Actionable Auditing: Investigating the Impact of Publicly Naming Biased Performance Results of Commercial AI Products. AAAI/ACM Conference on AI Ethics and Society.
• West, S. M., Whittaker, M., & Crawford, K. (2019). Discriminating Systems: Gender, Race, and Power in AI. AI Now Institute.
• Whittaker, M., et al. (2018). AI Now Report 2018. AI Now Institute.