It's August 9th. Welcome back to The Weekender read.
The Judgment Gap
In nearly every leadership meeting, someone will confidently say, “the model’s basically as smart as a person, so let’s let it decide.”
It sounds reassuring and straightforward, but this is where we make a mistake, because it’s the wrong idea.
We act as if being “smart,” “articulate,” “wise,” and “trustworthy” go together, like height and reach in a person, where having one means you automatically have the other.
Yet patients who have lost all language skills, who can’t name objects or describe their day, can still solve sudoku or logic puzzles without any trouble. Why?
Because it turns out that thinking doesn’t rely on language, but on something else.
Once you see this, you start to see it everywhere, even in the tools that write your marketing copy or quietly decide what a customer sees next.
Large language models are great at reckoning: they quickly and fluently manipulate patterns with statistical confidence.
But judgment is a different skill. It means understanding what a specific situation needs when something real is at stake, and caring about getting it right for the person involved.
Reckoning can seem like judgment, especially when it sounds smooth and confident. And that’s where the problem starts.
I’ve written before about this feeling in Reckoning with the Machine, describing the unease of watching systems change government, law, and identity faster than anyone can keep up.
Once you know it, you start to notice the gap everywhere it can quietly cause harm: in the ‘let the agent decide’ feature on your product roadmap, in assistants that finish a customer’s sentence before they’ve finished thinking, and in every ‘recommended for you’ module that confidently chooses for someone without caring if it got it right.
That’s the main point of my next piece later this week: The Judgment Gap. Because fluency that looks like competence is now a real strategy problem for anyone running a brand, a customer experience team, a product roadmap, and so on. This isn’t just an abstract topic for philosophy seminars.
Here’s the key point: the more ‘thinking time’ you give these systems, the more confident they become, but not more careful.
That’s the opposite of how real judgment works under pressure.
You can manufacture confidence.
But you can’t manufacture trust.
So when you say ‘the AI is deciding,’ is it really making a judgment call, or is it just reckoning and pretending to be judgment? I think I know the answer, and the two seem the same until something goes wrong for the person using the system. By then, the real work should have happened earlier, not after the fact.
This is what happens when decisions are quietly handed over, again and again, and we don’t remember agreeing to it.
That’s what this Weekender edition explores.
1
At a Talks at Google conversation, someone asked Mo Gawdat, the former ex-Chief Business Officer of Google X, and someone who has spent more time than most of us thinking hard about where this is heading: as AI takes over most of the thinking and deciding for humans, how do we deal with people becoming passive consumers who quietly surrender their own ability to analyse and decide for themselves?
His answer surprised me, and we should consider it fully.
His view is genuinely utopian.
He believes we’re heading toward a societal reset, a moment when complexity outgrows itself, and that after this, most analysis, negotiation, and decision-making will be handled by machines. Instead of panic, he feels relief.
If machines take over decision-making, he argues, humanity can stop defining itself by output and start focusing on something older. As he put it, “our ability to love and connect” is what truly makes us human, not our ability to analyse a spreadsheet or optimise a decision.
I don’t disagree, and I want to believe him. He may be right in fifty years. But no brand, CX function, or product team operates on a fifty-year horizon.
We’re focused on the next quarter, where the reset hasn’t happened, abundance hasn’t arrived, and every “recommended for you” module, autofilled form, and AI-drafted response is quietly deciding something for a real person with a real stake in the outcome.
This is happening today.
So until that reset comes - if it ever does - I believe the now work is to decide, on purpose, how much of someone else’s authorship you’re willing to let an interface quietly take over.
Gawdat’s utopia assumes the handoff is complete and safe to make. But right now, we’re in the middle of making it, one convenience, one designed experience, at a time, without ever agreeing to the terms — which is exactly the territory the next two sections in this Weekender are describing.
2 The Great Delegation.
You open your phone to a feed that has already decided which fifteen things in the world deserve your attention.
You start typing, and three suggested words appear ahead of your cursor, offering to finish your thought before you’ve even finished thinking it. You draft something important, and a tool gives you a cleaner, faster version, so you accept it, because most of the time, it really is better, isn’t it?
None of these feels like a loss of anything.
Each one, by itself, feels helpful. It saves you effort you didn’t really want to spend.
But that’s the catch.
We don’t lose our autonomy all at once. We lose it through repeated small, seemingly reasonable handoffs, none of which feel like we’re giving up control. Then we realise that many decisions in our lives are happening in places we never really agreed to.
It’s delegation—the quiet handover of decisions, one convenience at a time, from a person to a system built to anticipate them.
This happens because of a gap between reckoning and judgment.
Interfaces are built mostly on reckoning, but they’re so smooth that it’s hard to notice, in individual moments, that real judgment isn’t there.
As I’ve written about before, we live as two selves at the same time.
There’s our legible self, built from our click patterns, watch time, and purchase history—the ‘recommended for you’ version that’s easy to measure and profitable.
And then there’s our wild self: intuitive, unpredictable, and able to change its mind in ways no past pattern could predict. Sometimes, it even surprises you.
Almost every interface you use today is designed to serve only your legible self, because that’s the only part a syI wrote a whole piece on this split in ‘AI and Your Two Selves’ if you want the longer version.t the longer version).
What’s new is that researchers can now measure what this actually costs, instead of just talking about it in theory.
Researchers find that those who accept AI suggestions without resistance report lower confidence in their own reasoning than those who challenge the output before accepting it.
It turns out that you keep your judgment by using it regularly, by exercising it, and not just because you’re the one pressing ‘send.’
It’s the Autonomy Trap, and it’s all about design choices, made by designers constantly trying to reduce friction, but that quietly decide how much of your own authorship remains.
This raises a question: if you’re one of the people designing these systems, not just using them, how do you build for someone’s wild self when the system only lets you see their legible self?
3 Consent Rails
Here’s my thinking if you’re the one building the feed, the checkout flow, or the recommendation engine, instead of just using what someone else created.
Every default, every pre-filled field, and every ‘recommended for you’ module is a choice about how much of the other person’s authorship you allow to remain.
The design question isn’t whether to influence someone’s decision—every interface does that, and always will. It’s about how much thinking you’re comfortable letting the interface quietly take over, and whether the person using it even notices.
It comes down to three questions.
Reversibility: Can someone easily undo what the system just did for them, or does it lock them into something hard to change?
The Hand-Back: Does the system know when to give a decision back to the person, based on what’s at stake and their situation, instead of quietly handling everything itself?
And the Vibe Check: Is the interface aiming for what someone actually said they wanted, or just reacting to their most impulsive click?
All three questions are based on one rule: the bigger the impact of a decision, the more friction there should be, not less. This goes against what most growth teams are used to, and that’s the problem.
Some of us are Architects, but most of us are Approvers.
As Architects, we design the choice architecture; As Approvers, we just click through whatever is handed to us.
Most of us are acting as Approvers of decisions we didn’t realise we stopped making ourselves. Most teams, often without meaning to, are building interfaces that quietly train customers to become Approvers, one convenience at a time.
4
Try this over the weekend—not as a productivity hack, but as a test: turn off one automated default, like the algorithmic feed, autocomplete, or an auto-renewing service you’ve never really checked.
Notice the extra effort it takes to do it yourself?
The irritation you feel as your own sense of agency starts returning?
Then ask yourself which of the ones you left on would actually give you a different result if you did them from scratch.
Most won’t, and that’s fine.
But a few will surprise you. Those are the ones worth taking back, not because the machine got them wrong, but because getting them right was never really the point.
The important thing is that you get to decide. This is one of the main arguments running through much of what I’ve written this year in The Age of Entanglement: identity, autonomy, and judgment have always been connected to our tools.s.
The discomfort many of us feel now isn’t just nostalgia for effort, but a fair question about how much of our life, identity, or brand relationship is being shaped. And the honest answer for most of us is: a lot.
I’ll be writing the long version of both halves of this argument: the judgment half in The Judgment Gap, and the autonomy half in The Autonomy Trap, so you can dive deeper into each next week instead of just this Weekender’s quick overview.
Have a great weekend, and just try turning something off.
I’m Michael Cooper, and I think about entanglement — our identities, using personal agents, AI, Intelligent Interfaces, culture — and their impact on our behaviours, choice-making, autonomy, and brand engagement.
If you like this newsletter and want more:
My private work, where I share original thinking with brands, strategists, and futurists, and have provocative conversations.
My LinkedIn, where I post my ideas.
You can also connect or follow me on X and Medium.
I speak at conferences, in-house company events, and with enterprise teams — anywhere I’m allowed to “think aloud” about the entanglement of our humanity, culture, strategy, tech, brands, and the future.
Ready for more?
Thanks for reading,
Michael