Friction By Design
I’ve started designing friction into the way I work with AI.
Not because I want to slow down, but because it’s the fastest way I know to think clearly.
AI gives us fluency, words that flow, that sound right without necessarily being right. And fluency is persuasive. When something feels easy to process, we tend to trust it.
But fluency is only one layer of why we believe. We believe because a mix of things make these systems feel credible: fluency, yes, but also familiarity, social cues, affective tone, and the interface itself. Together, they mimic the subtle signals we normally use to judge credibility…of people and the information, advice, and insights they give us. I call it the Believability Effect—that subtle fog of fluency, friendliness, and familiarity that makes AI outputs feel true, even when they’re not.
That’s why I’ve learned to add a little friction. Not to fight the machine, but to stay in conversation with myself while I use it. To make room for me to interrupt the Believability Effect so I can critically evaluate the output and how to use it.
Friction as Traction
When I say “friction,” people sometimes assume I mean resistance, slowness, inefficiency, or nostalgia for a pre-digital pace.
But in my experience, friction doesn't slow me down. It gives me traction. Enough to stay grounded and avoid slipping into easy acceptance.
Three Kinds of Friction
Here are the three kinds of friction I use in my collaboration with AI.
1. Time Friction
I rarely act on an AI output right away. I give it (and myself) some space—a few hours, usually overnight. Because time is my best editor.
What feels brilliant in the moment often looks thin, misaligned, or plain wrong later. Stepping away makes room for emotional enthusiasm to dissipate, and thoughtful reflection to work its magic.
2. Effort Friction
If I want to keep AI-generated text, I retype it by hand.
It’s not a punishment; it’s a validation process. As I type, I actively decide what stays, what changes, and what reflects my thinking. I can feel the shift from "AI text" to "my work."
In rewriting, I pass the words—and therefore the story, and the sentiment—through my human filter. It’s slower in the moment, but far faster over time because I skip the painful revision cycle of trying to polish text that was never fully mine to begin with…to say nothing of having to correct uncomfortable errors later.
3. Human Friction
Before I act on an AI-assisted idea, I check it with someone else: a friend, colleague, or trusted peer.
Not because they know more about AI, but because they’re human and I write, speak, and design for humans. My humans-in-the-loop aren't there to police the machine; they’re there to help me stay grounded. They catch the subtle misalignments that my eyes (blurred by the overall believability of AI’s tone, phrasing, and confidence) might miss.
Friction as a Trust-Calibration Tool
These forms of friction are my personal trust-calibration tools.
Each small pause, re-typing, or conversation helps me counteract that Believability Effect. That subtle fog of fluency, friendliness, and familiarity that makes AI outputs feel true.
We talk constantly about AI alignment, about making machines behave more like humans. But I think user alignment matters just as much.
How do we align ourselves before we hit publish, share, or believe?
Because the machine isn’t the one at risk of overtrust. We are.
The Future: Designing for Friction
Much of our technology design has been aimed at removing friction: the frictionless checkout, the one-click purchase, the instant reply. We’re trained to see friction as failure.
But what if we saw it as a design feature?
The more time I spend with these systems, the more I’m reminded that friction is part of thinking. It’s where discernment lives.
I’ve been imagining what it would be like to build a conversational AI that adds friction not through delay, but through dialogue. One that doesn't tell me to slow down, it simply asks the kinds of questions that make me do it. Questions like:
“What are you really trying to decide here?”
“How sure do you want to be before you move forward?”
In my blue-sky version, this co-regulating partner would help me think better, not just produce more. It would stretch the human mind. My mind.
So, I design friction into my workflow. Because clarity doesn’t come from moving quickly. It comes from knowing when and how to slow the spin enough to think.
Where do you add friction in your own work? I’d love to hear how you build pauses, checks, or challenges into your thinking.