For a substantial part of my career, I supported talent acquisition and development that brought in top-quality interns and produced exceptional, well-rounded STEM professionals. We collected and managed data that recruiting, hiring, placement, training, and funding decisions came from. We relied on human designed and developed processes and systems. And human systems carry human assumptions.
Bias isn’t new, it’s inherent in our human nature. We categorize, assume, and make biased decisions daily. Algorithmic bias has convoluted our decision making even more. It's a reflection of who we are as a society. AI systems have been trained on our historically biased data, making predictions and producing outcomes based on historical and sometimes intentional patterns of inequity. A hiring tool trained on a decade of promotion decisions will learn to favor whoever got promoted most in the past. A grant-scoring system trained on prior award data will learn to favor organizations that look like past recipients. Where’s the innovation and ingenuity in that?
Why This Matters for Nonprofits and Local Government
Organizations in the public and mission-driven sector are often the last to think AI bias applies to them, but are often the first to be harmed by it. If your nonprofit uses an AI screening tool to sort grant applications, volunteer inquiries, or job candidates, have you delved into the system design and its training data? If your government agency uses AI to flag service requests or prioritize resources, the same question applies. Bias in these contexts doesn't just affect arbitrary outputs. It affects real people that depend on public services, have acquired the skills, but still lack opportunities, or trusted your organization to treat them fairly in a world that often diminishes their worth.
What You Can Actually Do
Start by being curious to identify and mitigate these biases. Remember, AI isn’t one-size fits all and AI vendors are still, well, sales people. Whether the tools you are considering are popular tools (ChatGPT, Zapier) or more customized, less-known built in smaller organizations, they all have both possible benefits and harms. Consider intentional questions built into your AI procurement and oversight process:
• Was this AI tool tested for disparate impact across race, gender, age, or disability?
• Can the vendor show you how it performs across different population groups?
• Is there a human review process, that forces true human decision making versus fluff reviews, for high-stakes decisions made with AI assistance?
• Does your team know how to recognize when an AI output seems off?
Bias Isn't Inevitable — But Ignoring It Is a Choice
AI governance frameworks like the NIST AI RMF and OECD AI Principles both center fairness and bias mitigation as core requirements. Organizations that treat bias as a 'tech problem' are not viewing the entire picture. Tech companies can only do so much to mitigate bias; organizations bear responsibility as well.
AI adoption should not result in diminished organizational values.
If you're not sure where your AI tools stand on bias and fairness, let's talk. I help organizations ask the right questions before those questions cost them their limited resources, reputation and integrity. Reach out at TawanaTownsendConsulting.com for more.