Many people assume AI will make their lives easier—and in many ways, it does. It can summarize articles, draft emails, predict outcomes and surface trends we might otherwise miss. But easy isn’t the same as efficient or effective. What we actually need to work well with AI isn’t greater ease, but more creative friction.
Creative friction means inserting deliberate tension into your work with AI to counter two different types of biases: human cognitive biases and AI’s mechanistic ones. Cognitive biases—those well-worn mental pathways that the human brain relies on to conserve energy—make us see the world through the past and through comfortable patterns. AI develops biases based on how it’s trained. AI doesn’t “know” anything; it aggregates data and information from across the internet with limited judgment.
Both biases are persuasive and can be pernicious, sometimes leading us astray. The consequences are both technical and personal. They shape the jobs we get, the care we receive, the paths we pursue and the risks we take or avoid. Often, we don’t even notice it happening.
When you’re tackling a challenge, it might seem like the hardest part is gathering facts or weighing options. But the real difficulty lies in the combatting those hidden influences steering our thinking without our awareness as we search, sort and decide.
Cognitive biases have a purpose: They help us move quickly through daily life, automating decisions we don’t need to spend precious brain space on, whether it’s reviewing the weekly customer data or driving to work. But they often lead us astray when accuracy really matters.
Take confirmation bias for example: Once we’ve formed an opinion, we tend to predominantly notice information that supports it and ignore what doesn’t. Availability bias leads us to give more weight to what’s most recent or most memorable, even if it’s not the most relevant. These biases quietly influence the questions we ask, the evidence we trust, and the options we consider—often without us realizing it. They don’t just distort what we see. They distort how we think.
AI, on the other hand, doesn’t think the way we do—and that’s a good thing. It doesn’t have wants or needs or opinions. Its flaws come from how it’s built. AI systems learn from past data, and that data can be biased, incomplete, unrepresentative or false. If an AI tool is trained on hiring data that reflects years of favoritism toward one type of candidate, for example, it will likely continue that pattern—not out of malice, but because that’s what it was shown.
These are mechanistic biases, or errors that come from the system’s inputs and logic, not its intent. AI tools don’t know what’s fair or right; they just optimize for patterns they’ve seen before. So when those patterns are flawed, the outcomes will be too.
The real trouble begins when these two kinds of biases—ours and the machine’s—start feeding off each other. We ask AI questions that are subtly shaped by our assumptions. AI gives us answers shaped by its training data. And when those answers confirm what we already believed, we trust them even more. That’s the feedback loop. Human bias shapes how we use AI, and AI bias reinforces what we already think. It’s a cycle that can quietly magnify errors, limit creativity and lead to poor decisions that feel efficient and objective—but aren’t.
To steer around these pitfalls, I’ve found that creative friction can be used at key moments where human and AI biases tend to take hold before the question is asked, while the answer is being reviewed, and as the final decision is made.
Apply creative friction in these three moments to ensure that what moves fast also moves in the right direction:
1. Before asking AI for help. First, examine your own outcome expectations. Ask yourself: What am I hoping AI’s input will confirm? What assumptions am I bringing in? This simple step slows the rush toward a pre-formed conclusion. It might lead you to reframe the question you ask to be more neutral or to share disconfirming data. The friction here can help you spot when you’re using AI to validate what you already believe. Friction here isn’t obstruction—it’s clarity.
2. Once the AI gives you an answer. This is when you need a second layer of friction: curiosity. Instead of accepting the output as fact, you can ask, “Why did the tool recommend this? What might be missing? What data did it rely on?” These questions enable you to challenge the relevance and factualness of the information provided.
3. Before locking in a decision. In this moment you can cross-check—either with a second tool, a colleague or a different lens entirely. Even if AI is providing quality information, it doesn’t know your circumstances and context.
These small moments of intentional resistance and friction help surface blind spots, challenge defaults and reconnect you with your own judgment. When added at just the right places, creative friction doesn’t slow you down; rather, it sharpens your thinking and strengthens your choices.
The goal is to surface what would otherwise stay hidden. It might reveal a confirmation bias where you look to find data that agrees with a favored hypothesis. For AI, it might flag a data bias that happens when the information used to train an AI system reflects existing inequalities or gaps.
Think of creative friction like rubbing two logs together. When you apply the right pressure at the right moment, you don’t just uncover what’s beneath the surface. You spark something new: a better insight, a smarter question, a clearer path forward. It’s not about doubt for doubt’s sake; it’s about creating heat that leads to light.
In decision-making, especially with AI, creative friction helps you resist the pull of smooth, fast answers. It invites you to examine the assumptions baked into your prompt, the reasoning inside the tool and the conclusion you’re being handed. Without it, you risk making decisions that are smooth, efficient and quietly flawed.
