Unhook
Confessions of a UX Addict. What happens when a UX designer realizes they are the victim of their own user retention strategies?
A few months ago, I was setting up Google Family Link for my daughter.
I went through the whole thing — app limits, screen time budgets, content filters, bedtime restrictions. I spent about 45 minutes making sure her phone couldn’t trap her. Making sure the feed would slow down. Making sure she’d have to actually choose to keep scrolling instead of just falling.
When I was done, I felt like a responsible parent.
Then I picked up my own phone and opened Instagram.
I didn’t set a timer. I didn’t have a bedtime restriction. I had no app limit, no content filter, no one watching whether the feed was taking more than it was giving. I just opened it, the way I open it twenty times a day, and I scrolled until something interrupted me.
I’ve been a UX designer for over twenty years. I’ve built consumer apps. I’ve designed the flows, the notifications, the onboarding sequences, the engagement hooks. I know exactly how these things work. I’ve won awards for making them work well.
And I set up a protection system for my child that I have never once considered building for myself.
That’s the thing I can’t stop thinking about.
How Hooked Got Into Everything
In 2014, Nir Eyal published Hooked. If you work in product or design, you probably know it. Trigger. Action. Variable reward. Investment. A four-step framework for building products that people return to habitually.
It was genuinely useful. It explained something that designers had felt intuitively but couldn’t articulate — why some products stuck and others didn’t. It gave us a theory. And designers, who had been operating largely on instinct and good taste, loved having a theory.
The problem is what we did with it next.
Hooked was written for consumer social apps. It was describing a specific product category, in a specific moment, for a specific kind of user — an adult, with full information, choosing freely to use a product.
But frameworks don’t stay where they’re put.
By 2016 it was in mobile gaming. By 2018 it was in health and fitness apps — the ones with streaks that made you feel guilty for taking a rest day. By 2020 it was in enterprise software, where I’ve watched sales dashboards gamified with leaderboards and “at risk” badges that made human beings feel like metrics. And now, in 2024, it’s baked into every AI product being shipped — every nudge, every suggested action, every “you haven’t checked in today” prompt is a trigger wrapped in good intentions.
We didn’t ask whether the framework should spread. We just noticed that it worked, and we used it.
The People Who Named It
To be fair, some people did ask.
The Center for Humane Technology — Tristan Harris and others who had worked inside the machine — gave us the systemic critique. They named the attention economy as a public health problem. They made policymakers and parents pay attention. That was genuinely important work.
Don Norman, who gave us human-centred design in the first place, started arguing that human-centred isn’t enough. That our moral radius as designers needs to expand beyond the user in front of us — to the people the product touches, the families it sits inside, the sleep it disrupts, the anxiety it compounds. He called it humanity-centred design.
Both of these things are right.
But neither of them gave me — gave us, working practitioners building real products for real clients — a vocabulary for the opposite. A way to talk about it in a design review. A framework for the unhook.
CHT said the system is broken. Norman said think bigger. And I went back to my desk and kept designing notification systems, because the client wanted higher open rates and I didn’t have a better language.
That’s the gap I’ve been trying to close.
What I Started Noticing
Once you see it, you can’t unsee it.
The streak in Duolingo. I used to admire it as a retention mechanic. Now I watch people — myself included — open the app not because they want to learn Spanish, but because they don’t want to break the number. The goal has migrated. You’re no longer doing it for the thing. You’re doing it for the record of having done it.
The like counter. We designed that to tell you something useful — people responded to what you shared. But what it actually does is install a checking behaviour. You post. Now you need to know. Then you need to keep knowing. Every few minutes, the pull. Not because anything has changed. Because the uncertainty is designed to be unresolvable.
The typing indicator. Three dots. Someone is composing a reply. You cannot leave now. You are being held in an app by punctuation. Social obligation, manufactured by a UI element, because someone on a product team knew that presence creates retention.
The autoplay countdown. Five seconds before the next episode starts. Stopping requires an active decision. Continuing requires nothing. Inertia, weaponised. The default was chosen deliberately. And it was not chosen for you.
Get Bijith Kunchammed’s stories in your inbox
Join Medium for free to get updates from this writer.
Every one of these was a design decision. Made by a designer. Probably not unlike me.
The Asymmetry That Bothers Me Most
Here’s what I keep coming back to.
The people most harmed by these patterns are not the people who were imagined when they were designed.
A 32-year-old product designer in Bangalore — someone like me — has context. I know what a streak is. I’ve read Hooked. I can, in theory, make a somewhat informed choice about whether to let it work on me.
A first-generation smartphone user in Tier 3 India, encountering a financial literacy app for the first time, has none of that context. The mechanics are invisible. The designer’s knowledge advantage is total. And I have designed for that person too.
An employee using enterprise software they didn’t choose, with engagement mechanics built into their performance dashboard, has no exit. The hook is inside their working day. And I have designed for that person too.
Same patterns. Wildly different power dynamics. Wildly different ethical weight.
We designed them without making that distinction. We learned a framework, and we applied it uniformly. The person with the most information and the most choices got the same product as the person with the least.
Five Things I’m Trying to Do Differently
I’m not here to tell you to stop building products that people use. That would be absurd. What I’m trying to articulate — what I’ve spent the last year trying to build into my own practice — is a different set of questions to ask before the engagement mechanic goes in.
The first is literacy. Before any rule, any limit, any control — can the user see how this works? Not the usage data. The intent. This feature uses a variable reward loop. This notification is timed for when cortisol is lowest. This streak is designed to activate loss aversion, not to help you learn. If you can show someone exactly how the mechanic works and they still choose to use it, it’s honest. If you have to hide the mechanic for it to function, something is wrong.
The second is agency. Platform defaults are not neutral. They’re chosen for the platform’s engagement, not for the user’s life. What would it look like to give users a blank page — with recommended healthy starting points, grounded in research — and let them write their own terms? News at 8am and 6pm only. Autoplay off. Feed limited to people I actually know. Off by 9:30pm. Their rules. The device as a servant, not a warden.
The third is environment over willpower. Every solution we’ve designed for digital overuse puts the burden on the human. Resist the notification. Delete the app. Set a timer. But willpower is finite and unreliable, and you are asking one person to out-muscle a system that has thousands of engineers working in the other direction. The better question is: how do we change the conditions instead of the resolve? When the environment shifts — when the feed slows, when the reward becomes slightly less variable, when the path of least resistance is presence instead of scroll — the pull weakens without anyone having to fight themselves.
The fourth is signal preserved. The fear of unhooking is missing something real. A genuine emergency. A message from someone who matters. Breaking news that changes things. This is a legitimate fear, and it’s exactly what engagement-first design exploits to keep notifications on. The answer isn’t less — it’s separation. What the user defines as genuinely important should reach them always, frictionlessly, immediately. Everything else — the streak reminders, the algorithmic nudges, the engagement bait — operates within rules the user has set. Nothing important is lost. Everything extractive is contained.
The fifth is the whole person. The phone emits radiation. It tracks location. It shapes sleep, relationships, the development of children who watch adults use it. Designing for the whole ecosystem means taking responsibility for all of this, not just the interface. A device designed for human flourishing doesn’t just show you how long you’ve used an app. It understands what you were doing instead — and whether that was better.
What It Could Look Like
I’ve been sketching what an OS designed around these five principles would actually look like. Not a new product — a design brief for the infrastructure that already exists.
A Hook Map: open any app and see, plainly, exactly how it’s designed to keep you. Variable reward loop. Social comparison mechanic. Streak anxiety trigger. Labelled. Not hidden.
A Gap screen: what you said you wanted from this app versus what you actually do. Shown once, gently, without judgment.
Your Rules: a blank page with recommended healthy starting points. You write the terms. The device holds them — not by blocking or locking, but by quietly shifting the environment when you’re in the spaces and times you said mattered.
A Quiet Mode: not do not disturb. Something more considered. The phone knows you’re home, that your family is present, that you said this time was yours. It becomes slightly less interesting. Not useless. Just less insistent.
A Whole Self view: radiation levels, privacy exposure, sleep impact, usage patterns in the context of your actual life — not as guilt, but as information. The device aware that it exists in a body, a home, a family.
None of this requires technology that doesn’t exist. The sensors are there. The APIs are there. Apple and Google have all the plumbing. What they’ve chosen to build on it is control and surveillance. This is a choice to build care instead.
The Question I Can’t Answer Yet
I set up Family Link for my daughter because I understood what the product was doing and I didn’t want it done to her without her knowledge or consent.
I haven’t done the same for myself. Not really. Not in any way that’s honest.
And I build these products. That’s the part I sit with.
I don’t think I’m a bad designer. I think I was working within a framework that felt like craft and turned out to also be extraction. I think most of us were. The best engagement mechanics I ever designed — the ones I was proudest of — were the ones that worked so well the user didn’t notice they were being held.
That’s not a neutral thing to be good at.
So here’s what I want to leave you with. Not an answer. A question that I think is worth carrying into your next project, your next design review, your next conversation with a client who wants higher DAU.
Does this feature need the user to come back — or does it need the user to flourish?
Those two things are not always in conflict. But they are not the same thing. And for most of the products we’ve built, for most of the last decade, we designed for the first one and called it the second.
I’d like to know what happens when we start designing for both.
Bijith Kunchammed is Co-founder and Head of UX at AufaitUX, an award-winning UX design agency with studios in Bengaluru, Calicut, Palo Alto, and Sharjah.
