When designing AI-driven software, one of the biggest challenges isn’t purely technical—it’s human. Specifically, confirmation bias, the tendency to seek out information that supports existing beliefs, can silently kill creativity and trap teams in endless cycles of flawed designs. At voiceofreason.ai, we’re acutely aware of this danger and have developed a design process that actively combats it, ensuring that our AI-driven solutions for people with Alzheimer’s, autism, and cognitive challenges remain innovative, user-centered, and effective.
As Alan Kay, one of the pioneers of modern computing, once said, “A change in perspective is worth 80 IQ points.” That’s exactly what we aim for: shifting perspectives by leveraging AI as a thinking partner, not an echo chamber of our initial ideas.
Our AI-Driven Design Process
Our software design process starts with defining a base set of architectural requirements. These requirements include technical considerations like speed, maintainability, and cost, but most importantly, they reflect our core priority: the user experience. Since we build for users with unique cognitive needs, getting this right isn’t just important—it’s essential.
Here’s how we approach it:
Start with a Wish List of Outcomes:
Rather than jumping straight into technical specs, we begin with a clearly prioritized outcome-based wish list. This list outlines what we want the end-user to achieve. For example, when designing a feature for guiding a lost user back home, the top priority is simplicity and seamless interaction, not technical complexity.
Collaborate with Multiple AI Models:
Once we have our wish list, we engage several AI models—Claude, ChatGPT, and others—to critique and refine our initial requirements. We ask them questions like:
What are the pros and cons of this design?
How can we simplify the flow while preserving functionality?
What alternative approaches might achieve the same outcomes with less complexity?
This multi-model approach helps expose blind spots, but only if we avoid the trap of feeding the AI overly detailed or rigid requirements.
Iterate and Evolve the Design:
Through iterative interactions, we push the AI to generate new ideas. We ask it to create diagrams (like mermaid flowcharts) and propose alternative data flows. Sometimes we provide class specifications upfront; other times, we let the AI propose them from scratch.
The key here is balance. We want AI to challenge our thinking, not just expand on our original ideas. As Steve Jobs said, “Innovation is saying no to 1,000 things.” That includes saying no to ideas that feel safe but won’t move the needle for our users.
The Danger of Confirmation Bias
While this process sounds straightforward, there’s a lurking danger: confirmation bias. When we provide the AI with highly detailed requirements, it tends to stick too closely to those inputs. It won’t stray far from the original design—often simply expanding it, adding unnecessary complexity, and reinforcing flawed assumptions.
This mirrors a common problem in human teams. How many times have you seen teams stuck in endless debates, redesigning the same fundamentally flawed system because no one wants to abandon it? Confirmation bias keeps everyone locked into old thinking, and the result is wasted time and stifled innovation. As Charles Kettering, an inventor and head of research at General Motors, once said, “A problem well stated is a problem half solved.” The trouble is, confirmation bias prevents us from properly stating the problem in the first place.
How We Combat Confirmation Bias
At voiceofreason.ai, we’ve learned that the best way to combat confirmation bias is to anchor every design decision in customer outcomes, not technical preferences. By framing the problem in terms of what the user needs rather than what the technology can do, we force ourselves—and the AI—to think differently.
This outcome-driven approach leads to more creative solutions. Instead of endlessly iterating on favorite designs, we focus on delivering something that truly helps the user. If a proposed solution doesn’t meet our top-priority outcomes, we discard it, no matter how clever or technically impressive it might be.
It also requires us to embrace being wrong—a lot. Innovation demands a willingness to throw out cherished ideas in pursuit of something better. As Jeff Bezos once said, “If you double the number of experiments you do per year, you’re going to double your inventiveness.” That’s our philosophy. We experiment, fail, and refine quickly, always keeping our users’ needs front and center.
The Result: Better, Simpler Solutions
This approach has already paid off. Our two-sandwich teams, using this iterative, outcome-driven process, have built and refined prototypes in record time. More importantly, these prototypes aren’t just technically sound—they’re designed with empathy for the user. Whether it’s helping someone with Alzheimer’s navigate safely or assisting a child with autism in communicating more easily, our products reflect a deep understanding of the people we serve.
Looking Ahead
Designing AI-driven solutions isn’t just about coding or architecture—it’s about thinking differently, questioning assumptions, and staying focused on the end user. At voiceofreason.ai, we’re committed to avoiding the trap of confirmation bias and embracing new perspectives. That’s how we build products that truly make a difference.
We’re excited to take the next steps, gathering feedback from real-world testing and continuing to innovate. Because, as Albert Einstein said, “We cannot solve our problems with the same thinking we used when we created them.”
By staying open to new ideas, challenging our biases, and keeping our users’ needs front and center, we’re confident we’ll continue to create groundbreaking solutions for those who need them most.
Want to learn more about how we design for innovation? Check out our blogs at voiceofreason.ai.