It Didn't Take Your Choice. It Shrank the Menu.

People worry that AI will start making their choices for them, but mostly AI does something quieter. It lets you choose, but from a list AI wrote.

Offered three replies, you pick the second. That was a choice, just about.

But you didn't choose from everything you might have said. You chose from three things the AI guessed you'd say, based on what people like you have said before.

The reply you'd have written yourself, the slightly odd one with the joke in it, wasn't there to weigh up. Nobody took that reply away, because it was never offered.

Add up every suggested reply, every recommended next show, every "people also bought" and every sentence finished for you. You get a day in which you chose all the time and decided almost nothing.

Jacob Ward, a technology journalist, wrote a book about this called The Loop, and his argument has three rings. First, humans evolved shortcuts, the quick instincts that got our ancestors through the day. Second, businesses learned to cash in on those shortcuts. Third, AI learns from what we did before and predicts what we'll do next. Then AI narrows the options until we pick what was intended, and it all feels like freedom.

His picture for this is the loop, and his phrase for where the loop ends up is "convenience into control".

His most useful point is about the decision muscle. Choosing without a menu is a skill, like finding your way without a map, and skills you don't use fade.

Someone who has picked from suggestions for five years is worse at the blank page than they used to be, because they haven't practised.

Ward doesn't think you can beat the feed with willpower. He compares relying on willpower to managing an allergy by being brave. His answer is safeguards, which are small, fixed habits that put one choice of your own back into the day. None of them asks you to use AI less.

Off the menu

Once a day, when something offers you three options, write a fourth. Write your own reply instead of picking one, or choose the film nobody recommended. Your fourth option can be worse than the suggestions, as long as it's yours. Do this daily, because that's what practice is.

Or you can change what you ask AI for.

Don't give me suggested replies. Instead, ask me one question that would help me work out what I want to say, and then wait. I'll write the reply myself.

Ward borrows the next habit from avalanche experts, who know that the moment they most want to ski the slope is the moment to re-read the snow report. So when a suggestion feels exactly right, pay attention. When a recommendation lands with a small rush of relief, that's your cue to check what the suggestion is based on.

I'm inclined to go with this. Before I do, show me what it rests on. Show me the assumptions, the data, and what would have to be true for it to be the wrong call. Be plain, not reassuring.

The last habit turns your habits into rules, because willpower runs out at six in the evening and a rule doesn't. For example, "No suggested replies to people I know."

Or "I write the first sentence of anything before I ask." Write your rules where you'll see them, and keep them small enough to follow.

Most menus are fine, so I'd save these habits for the choices that shape you. It's a kindness to get three suggested answers to "are you free at 3?"

The loop matters most in what you read and what you say to people. Even there, you could argue it's harmless. A shorter menu is just what convenience is, a good recommendation is a useful service, and nobody wants ten thousand options. And "make your own choice" can be a way of pretending effort is good for its own sake.

Ward accepts most of that, and so do I, and we only disagree at the edge. If every choice is narrowed for you, is the person making those choices still someone you'd recognise?

A fourth option once a day is a cheap way to find out, though the test is an uncomfortable one. How often, this week, did you choose something that wasn't on offer?

AI offers three. You can always write the fourth.

Credit. The ideas of "the loop", "convenience into control", the decision muscle and the avalanche-expert habit are Jacob Ward's. They come from The Loop (2022) and his 2026 interviews. Read the book for the full argument.