InstagramLinkedIn
Creative Impact

Designing AI Differently: Course-Correcting Bias

19 March 2025

Featured image for Designing AI Differently: Course-Correcting Bias

History has a habit of being written by the loudest voices in the room. For centuries, those voices have largely belonged to men. But as AI shapes the way we learn, decide, and govern, we have an opportunity to shift the narrative and design a more balanced future.

The thing is, AI doesn’t have unconscious biases—it has unintentional ones, picked up from patterns in its training data. And right now, those datasets are skewed. Lack of diversity in AI training has real-world consequences; facial recognition tools have been called out for racial bias, and even AI-generated leadership advice for women often leans into outdated stereotypes.

Here’s the kicker: AI is already the first stop for information for 68% of U.S. adults. These models are trained on vast amounts of data—articles, videos, podcasts—but the question is, whose voices are shaping that knowledge? If AI continues learning from the content that dominates today’s digital space, the future will be built on a foundation of male-driven perspectives.

Why Makers Should Step Up

In business, we already know that diverse teams are 75% more likely to see ideas become commercially viable (World Business Council for Sustainable Development). If AI is becoming an integral part of how teams function, we need to apply the same logic—more diversity leads to better outcomes.

But there’s resistance. Some argue that making AI more inclusive isn’t commercially viable. The truth? Research already shows that training AI on more diverse datasets doesn’t just benefit underrepresented groups—it improves overall accuracy and performance for everyone.

Designing the Next Wave of AI

So, how do we fix this? In design, we embed inclusivity from the ground up—whether it’s in branding, UX, or campaigns. AI should be no different.

One approach is to build ethical frameworks directly into AI systems. Imagine AI models that preface responses with acknowledgments of potential bias, or actively correct for it by pulling from a broader range of perspectives.

Then, there’s content creation. Right now, 90% of publicly available leadership speeches come from men. Machines don’t know what’s missing; they just learn from what they’re fed. If we want AI to recognize female leadership, we need to change what goes into the system. That means curating datasets with speeches from women across industries, cultures, and backgrounds—not just adding more of the same voices.

At Your Creative, we’ve been working on exactly this with Madam Speaker, a digital archive housing over 200 speeches by women across eight disciplines. Initially curated by archivists, it’s now a crowdsourced initiative, aiming to gather 100+ hours of diverse leadership speeches to train AI models. The goal? To make sure AI reflects the real world—not just a slice of it.

A Future Built on Better Data

AI is learning from us, so let’s design it with intention. Let’s move beyond the outdated notion that power sounds a certain way. The future of AI—and by extension, the future of decision-making—shouldn’t be shaped by a single dominant voice. It should be rich, diverse, and undeniable.