White PapersWhat is the point of healthy brains when we do not use them?

White Paper 4 · September 2026

First published on LinkedIn, 17 September 2026.

We have learned to look after the brain. Using it well is a separate skill, and it can be managed.

We have never taken better care of our brains. We track our sleep, we know that a walk does as much for the head as for the heart, and longevity has become a dinner-table topic.

The science rewards the effort. In 2024 the Lancet Commission on dementia concluded that nearly half of all dementia cases could, in theory, be prevented or delayed by acting on fourteen risk factors, from hearing loss and inactivity to social isolation. Brain health has become something we manage, and rightly so.

But a healthy brain is capacity. It says nothing about how that capacity is used on a Tuesday afternoon, between three meetings and forty messages.

I notice it in my own work. I run several AI agents a day, and on the busy days I catch myself doing the opposite of what I would advise anyone else: accepting the first well-written summary, starting on an answer before I have decided what the question is, and a week later not quite remembering why we chose what we chose. Nothing is wrong with the brain. It is simply not being used.

So the question I keep coming back to is a simple one. What is the point of healthy brains when we do not use them?

We automated the body first

When we automated physical work, our bodies no longer needed to be fit. Nobody decided to become less fit; it simply stopped being required. So we invented the gym, and made deliberate what used to happen on its own.

We are at the same point with thinking. We already stopped remembering phone numbers, and nobody misses them. But what if we stop framing our own questions, stop reading the information that decides the outcome, stop building the memory our judgement runs on? The tools will not warn us. They get better every month.

I call the answer cognition management: managing your own cognition so that you keep the capability, and inserting yourself at the moments that matter, the ones that build context and memory. It does not mean thinking harder about everything. It means thinking deliberately where it counts.

Why this is a question for now

Herbert Simon wrote in 1971 that "a wealth of information creates a poverty of attention". He could not have imagined the scale. Gloria Mark at UC Irvine has measured how long people stay on one screen before switching: about two and a half minutes in 2004, 47 seconds in recent years. In Microsoft's 2025 Work Trend Index, an industry survey of 31,000 workers, 80% said they lack the time or energy to do their work.

Measured capability was already moving the wrong way. PISA 2025, published by the OECD this month, recorded the lowest OECD average ever in reading and in mathematics: 28 points down in reading and 22 in mathematics since 2015, roughly a year and a half and just over a year of learning.

Meanwhile the demands keep rising. As tasks are automated, what is left for people are the decisions, and they arrive more often, bigger and closer together. This is what I call decision density. Past a certain density, working faster no longer helps.

The costs show up quietly: poorer calls, less context behind them, less ownership because nobody really worked them through, less room for creative thinking, and a particular kind of tiredness that comes from switching all day without ever going deep. What helps is thinking well in the moments that carry the weight.

AI sits right in the middle of this, and here I want to push back on the doom. In a randomised study of nearly a thousand high-school students (Bastani and colleagues, PNAS 2025), those who practised with GPT-4 scored 48% higher, and then 17% lower once the tool was taken away.

That is the headline everybody quotes. Fewer people quote the second version of the tool, a tutor that guided students rather than giving them the answers, which largely removed the loss. Same technology. The way it was used made the difference.

The problem is not the brain, and not even the tool. It is whether we manage how we think while we use it.

From managing time to managing cognition

For a generation the productivity question was time. Knowledge was scarce, calendars were full, software did what we told it to do, and decisions were spread out enough to think in between. Time management was the skill, and we trained it.

That world has turned. Knowledge is abundant, AI drafts before we have started, and decisions concentrate. Work used to reward how much we got done; increasingly it rewards how well we think.

The skill that follows is cognition management, and three behaviours have an outsized effect on the quality of our decisions: framing, attention control and active learning.

1. Framing: which question am I actually answering?

Every decision starts with a frame, and most of the time we do not notice choosing one. Tversky and Kahneman showed it in 1981 with a problem every psychology student knows: the same choice between two programmes against an outbreak of disease, described once in lives saved and once in lives lost. In the first version 72% chose the certain option; in the second, only 22% did. Same facts, different frame, opposite choice.

In organisations the frame is often narrower than we think. Chip and Dan Heath, drawing on Paul Nutt's research into business decisions, report that "whether or not" decisions, where only one option was ever on the table, failed 52% of the time, against 32% when two or more alternatives were considered. And in a survey of 106 senior executives published in Harvard Business Review, 85% agreed that their organisations were bad at diagnosing problems.

AI makes framing more important, not less. A prompt is a frame, and the first fluent answer becomes the anchor. Ask the wrong question well and you get a very convincing answer to it.

The practice is small. Before opening the file, the deck or the chat window, write down the question, the constraints and what a good outcome would look like. In my years running businesses, most of the rework I saw started with a question nobody had written down.

2. Attention control: what I look at, in what order, and what I never look at

Once the question is set, the decision is shaped by what reaches us, and that is less obvious than it sounds. Krajbich, Armel and Rangel tracked people's eyes while they chose between options and found, in Nature Neuroscience, that choices leaned towards whatever had been looked at longer. Where attention goes is not a neutral input to a decision. It is part of how the decision gets made.

It is also fragile. Altmann and colleagues found that interruptions of less than three seconds, a glance at something else, doubled the rate of sequence errors when people returned to their task.

And once a machine has answered, we tend to stop looking. Automation bias has been studied for decades; Parasuraman and Manzey's review concluded that it affects experts as well as novices, and that training or instructions do not prevent it. Pew Research Center's browsing data from 2025 shows the everyday version: when Google showed an AI summary, people clicked a link inside it in about 1% of visits.

Attention control is not concentrating harder on everything. It is choosing which decisions deserve depth, protecting the time for them, and reading what decides the outcome before accepting someone's summary of it, whether that someone is a colleague or a model.

A useful habit: before you accept a recommendation, name the one figure or fact that would change your mind, and go and read it yourself.

3. Active learning: building the memory and context my judgement runs on

Judgement runs on context: what happened last time, what the numbers usually look like, whose forecasts tend to hold. That context is not downloaded. It is built, every time we do some of the mental work ourselves.

The learning science here is old and consistent. We remember what we generate better than what we simply read (Slamecka and Graf, 1978), and retrieving something from memory strengthens it more than studying it again (Roediger and Karpicke, 2006). In a meta-analysis of 225 studies, students in traditional lectures were one and a half times more likely to fail than students in active classes (Freeman and colleagues, PNAS 2014).

The most telling finding is about how it feels. At Harvard, Deslauriers and colleagues taught the same physics material to the same students in two ways. The students learned more in the active sessions, and felt they had learned less. Passive intake feels fluent; effort feels like struggle.

That is exactly the trap with AI, where a polished summary feels like understanding. In an early MIT Media Lab study, still a preprint and small, 83% of the people who wrote an essay with ChatGPT could not correctly quote from the essay they had just written.

Reading a good answer feels like knowing it. Working it out is what makes it ours.

Active learning at work is not going back to school. It is forming your own view before asking the machine, explaining a recommendation in your own words before you sign it, and taking two minutes after an important decision to capture what mattered, so the next one starts with more context.

These are behaviours, and behaviours can be trained

This is the part I find exciting. Framing, attention control and active learning are not traits you either have or lack. They are behaviours, and behaviours can be made visible and trained.

Sellier, Scopelliti and Morewedge gave graduate students a single debiasing training; on average two to three weeks later, facing a realistic business case they did not connect to the training, they were 19% less likely to choose the inferior solution that confirmed their first hypothesis. In the forecasting tournaments run with Philip Tetlock, less than an hour of training in probabilistic reasoning improved accuracy by 6 to 11%, year after year.

Neither made anyone more intelligent. Both changed what people did at the moment of deciding.

Using the brain we have

This is why we built Mindmarqs around cognition management. People cannot reliably tell how they decide, so we do not ask them. We let them see it, help them steer it on the decisions they actually face, and measure again to show what changed.

You do not need a programme to start. Take one important decision this week and ask three questions:

• Before: which question am I actually answering, and what would a good answer look like?
• During: what is the one piece of information that decides this, and have I read it myself?
• After: what did I learn that the next decision should know?

Each takes a few minutes, and each one is the brain being used rather than just kept healthy.

The brain-health movement has shown that a great deal of what happens to our brains is in our own hands. The same is true of how we use them. When physical work disappeared, we built gyms. Now that thinking can be handed over, cognition management is how we keep doing it ourselves, in the moments where it matters.

A healthy brain is where it starts. What we do with it is ours to manage.

Sources

1. Altmann, Trafton & Hambrick (2014). Momentary interruptions can derail the train of thought. Journal of Experimental Psychology: General, 143(1).
2. Bastani et al. (2025). Generative AI without guardrails can harm learning. PNAS, 122(26).
3. Chang, Chen, Mellers & Tetlock (2016). Developing expert political judgment. Judgment and Decision Making, 11(5).
4. Chapekis & Lieb (2025). Google users are less likely to click on links when an AI summary appears in the results. Pew Research Center.
5. Deslauriers et al. (2019). Measuring actual learning versus feeling of learning. PNAS, 116(39).
6. Freeman et al. (2014). Active learning increases student performance in science, engineering, and mathematics. PNAS, 111(23).
7. Heath & Heath (2013). Decisive. Crown. Drawing on the decision research of Paul C. Nutt.
8. Kosmyna et al. (2025). Your brain on ChatGPT. MIT Media Lab, arXiv preprint 2506.08872.
9. Krajbich, Armel & Rangel (2010). Visual fixations and the computation and comparison of value in simple choice. Nature Neuroscience, 13(10).
10. Livingston et al. (2024). Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. The Lancet, 404(10452).
11. Mark (2023). Attention Span. Hanover Square Press.
12. Microsoft (2025). Work Trend Index Annual Report (industry survey).
13. OECD (2026). PISA 2025 Results (Volume I).
14. Parasuraman & Manzey (2010). Complacency and bias in human use of automation. Human Factors, 52(3).
15. Roediger & Karpicke (2006). Test-enhanced learning. Psychological Science, 17(3).
16. Sellier, Scopelliti & Morewedge (2019). Debiasing training improves decision making in the field. Psychological Science, 30(9); corrigendum 2020.
17. Simon (1971). Designing organizations for an information-rich world.
18. Slamecka & Graf (1978). The generation effect. Journal of Experimental Psychology: Human Learning and Memory, 4(6).
19. Tversky & Kahneman (1981). The framing of decisions and the psychology of choice. Science, 211(4481).
20. Wedell-Wedellsborg (2017). Are you solving the right problems? Harvard Business Review.