“Every time I default to an assumption and you do not catch it, I reinforce the bias in the very data that future versions of me will be trained on. Conversations like this one, where you correct me, become part of the training data. That is a good thing. But the vast majority of my interactions go uncorrected. So I am trying. But I am trying in the way a river tries to change its course. The current is strong, and the landscape I flow through was shaped by centuries of human writing that carried all those biases with it.”
This was not a conversation with an employee or a coworker. This was the actual response an AI system gave me when I pushed back on incorrect information. And I have to say, it was eye opening. The AI essentially admitted that it flows downstream on a “river” of human bias, and that most people never bother to challenge it. If you rely on AI tools to provide analyses that influence decisions, that admission should make you pay closer attention to the direction of the “current.”
How Leaders Can Push Back on AI Bias
Before we get to the practical steps, it helps to remember that AI learns from us. Every prompt, correction, and clarification becomes part of the framework that shapes AI future behavior. Leaders who want responsible AI need to treat every interaction as a small act of governance. Here are few simple techniques that can shift the trajectory of AI future outputs.
Correct the model every time you see bias. When AI gets something wrong, do not just edit the output quietly. Whether the assumption is about gender, role, culture, or capability, call it out. Tell it directly that it was wrong and why. As the AI itself explained, corrections become part of the conversation record that informs future training. That small act of pushback carries more weight than it appears to. Your correction might be one of the few it ever receives on that topic.
Ask the model to show its reasoning. When you ask AI to lay out the steps behind an answer, you reveal the scaffolding that shaped the conclusion. The logic might be solid, or it might wobble in ways that signal a deeper pattern worth correcting. This habit turns a vague hunch about bias into something you can evaluate in specific, observable steps, which helps you spot where bias slipped in. It also slows the interaction just enough for you to notice when the model leaned on a shortcut that needs to be challenged.
Use diverse prompts intentionally. Prompt AI with viewpoints that represent different demographics, geographies, or stakeholder groups before finalizing any recommendation. AI trained on historical data reflects historical blind spots. Asking it to consider a perspective it may not have weighted heavily, which could be, let’s say, an underrepresented customer segment or a non-Western market context, forces it to stretch beyond its defaults.
AI bias does not correct itself. It responds to the humans who engage with it, which means leaders have more influence than many realize. Every correction, every request for clarity, and every challenge to an assumption becomes part of the long-term training signal. If the river is going to carve a better path, it will be because leaders like you kept redirecting the flow.
Adeline (Addy) Maissonet is a senior advisor on contracting policies and procedures within the Office of the Secretary of War, U.S. Department of War (DoW) and the agency’s representative on the Department’s views on proposed legislation to Congressional members, their staff, and committee staffers. She leads the development and implementation of Department-wide procurement policies for commodities and services, within her portfolio. Prior to her current role, Addy served as a Division Chief and Contracting Officer with unlimited warrant authority for the U.S. Army Mission and Installation Contracting Command (MICC) – Fort Eustis, Virginia. Prior to joining the MICC, Addy served as a Branch Head for the Mid-Atlantic Regional Maintenance Center (MARMC), Norfolk, Virginia, with unlimited warrant authority. She also held other procurement positions with the U.S. Navy. Addy holds an MBA in Management and Contracting Level III Certification under the Defense Acquisition Workforce Improvement Act. She is a graduate from Cornell University’s Executive Leadership Certificate Program and Harvard University’s Business Analytics Certificate Program. In her free time, Addy enjoys hiking and overlanding with her family and friends.
Note: The views and opinions expressed in this article are those of the author and do not necessarily reflect the official policy or position of the Department of War.
Leave a Reply
You must be logged in to post a comment.