The Death of Knowledge? Part 4 - Cognitive Offloading - When the machine does the thinking

Previously on The Death of Knowledge?

We’ve seen how expertise is being devalued in a world of instant answers. Now we turn to another equally important shift: the way we offload our own mental effort onto tools. AI is about to take that habit further than ever.

A long tradition of shortcuts

Humans have always looked for ways to lighten the cognitive load. We tied knots in string to remember debts. We carved tallies into wood. We wrote shopping lists and kept diaries. The calculator took away the grind of long division. Google replaced our need to remember who originally played Mark Fowler in EastEnders.

This isn’t laziness. It’s sensible. Offloading frees the mind to focus on other things. You don’t need to memorise every phone number if your device does it for you. You don’t need to recall the whole London Tube map if you can check an app. The danger is not in offloading. The danger is in forgetting to do any of the thinking ourselves.

The AI leap

While traditional offloading merely stored information, AI now takes care of the processing. You no longer just record the shopping list, you can ask the AI to plan the menu or check your fridge to see what's running low. You don’t just save numbers, you let it run the analysis. You don’t just check a definition, you ask it to write the essay.

The machine is not just a memory aid. It is a thinking substitute. And because it delivers its work fluently and fast, it feels easier to accept than to challenge.

The effect is amplified by tone. Machines do not signal hesitation. As we saw before, unlike real experts, they rarely admit “I don’t know.” The neat sentences and apparent authority short-circuit our scepticism. We find ourselves nodding along, less inclined to dig in and check for ourselves.

Cognitive Atrophy by convenience

Skills erode quietly. You don’t lose them overnight. You simply use them less and less. Rely on GPS for every journey, and your spatial memory weakens. Let the model write all your first drafts, and your ability to structure arguments may dull. Stop doing mental arithmetic, and even basic sums feel harder.

It’s like skipping leg day at the gym. You get away with it for a while, until one day you find the strength is gone. Cognitive habits need exercise too. If you always hand them over to a machine, you may not notice the decline until you need them and they’re not there.

We should be careful here. No one wants to romanticise drudgery. Few people pine for the days of memorising long multiplication tables or carrying every phone number in their heads. But skills are not just about utility, they shape how we think. Mental arithmetic builds number sense. Handwriting improves recall. Wrestling with a draft makes your ideas clearer. Remove the struggle and you risk losing the insight it produces.

Support versus substitution

The distinction is simple. Use AI as support and you keep control of the thinking. Use it as a substitute and you risk losing it. Support means generating options, checking work, and providing 'ideas' which you still interrogate. Substitution means accepting the output as the finished product without engaging with it yourself.

One practical discipline is the two-pass rule. First, let AI draft. Then put it aside and write your own version from memory. Compare the two. Notice where you lean too heavily on the machine. That process strengthens understanding rather than eroding it.

You can flip the order as well. Write your own version first and then ask the AI for improvements, additions, or blind spots. Or take the AI’s draft one and step into the editor’s chair. Humans are often sharper editors than drafters. Editing allows you to maintain the rigour of the review while still saving time on structure and style.

The point is not to outsource the whole job. It is to use the machine as a copilot. And while Microsoft Copilot may not be the best chatbot on the market, it has probably chosen the most accurate name. The tool is there to assist, not replace, and the metaphor is worth taking seriously: if the copilot suddenly took full control, you’d worry about the landing.

A dietary choice

Think of AI like fast food for the brain. Quick, tasty, and fine in moderation. But if it becomes the staple, the intellectual nutrition suffers. You can end up with more words, more outputs, more volume and less insight.

The problem isn’t the occasional Big Mac. It’s the daily dependence. A balanced cognitive diet means carrying some core knowledge yourself, practising problem-solving, and keeping curiosity alive, while still using AI to save time on the boring bits.

A healthy balance also means deliberate practice. Rehearse what you need to know without the machine. Stretch your own reasoning before you check its version. Build habits of curiosity, one of the most underrated human traits that AI cannot replicate. Ask why an answer looks the way it does, not just whether it sounds convincing. These small disciplines keep you in charge.

Why it matters

The stakes go beyond individual memory. A society that outsources too much of its thinking risks collective fragility. If no one remembers, checks, or questions, then errors multiply and bad ideas spread unchecked. True democracies (and let's not examine that one too carefully!) depend on citizens being able to weigh claims critically.

The paradox is that AI could strengthen this capacity if used well. By generating counter-arguments, by surfacing forgotten perspectives, by offering quick drafts to challenge. But that requires us to remain active users, not passive recipients. Once we stop questioning, we give up the very thing that makes knowledge meaningful.

The question for next time

If cognitive offloading reshapes how we think, education is where the consequences will hit first. What should schools and universities teach in a world where machines can handle the essays, the revision, and even the feedback? That’s the subject of the next part.

Go to https://www.glasshalo.co.uk/death_of_knowledge_2025/ for the full essay with more depth, more jokes, and more philosophers than you can shake a stoical stick at.

I train senior people how to use AI well in their work without the hype and the tech talk. If that sounds interesting for your company, please connect with me on LinkedIn or DM me. Or just download the full paper and I'll be in touch.