In 1956, cognitive psychologist George A Miller wrote a paper that was to become a cornerstone of cognitive science: the “magical number seven, plus or minus two”. It suggests that our working memory, i.e. the number of things we can juggle in our minds at once, hovers around five to nine. It is why phone numbers are broken up into two to three units at a time. It is why landline numbers were easier to remember than mobile numbers. The latter just has too many numbers for us to carry many of them around in our head. Miller’s magic number is why we struggle with too many listed choices in restaurant menus, why questionnaires with too many options and response choices fail us and why medical leaflets listing dozens of contraindications or potential side effects simply turn into white noise in front of our eyes.

Human cognition has limits and that has always been the case. It explains the limits of working memory, the narrow bandwidth of our mental juggling space. That is why, when we overload that space, it sets the stage of what is now called “popcorn brain”. That is when your attention is jumping, flickering and overloaded by rapid information inputs and tasks competing to be done. It is cognitive scatter. It is what happens when we keep trying to shove more into a system that is built for five to nine discrete things to hold in your head.

Then there is the Google Effect. This is also known as digital amnesia. A study (Sparrow, Liu and Wegner 2011) indicated that the magic number of five to nine may have reduced to as few as 3 items that we can juggle in our working memory. When we know information is stored externally, such as on Google, we are less likely to make efforts to remember it ourselves. So, the brain has adapted in this way to online information-seeking behaviours. Why memorise when you can search? Why hold all this stuff in your head, the burden of recall, when the online world remembers everything?

What we have started to outsource is not just working memory but thinking itself. Our brains no longer struggle to remember, they just reference things instead. We do not compare ideas, engage in critical thought. We want summaries, skim reading, visual infotainment. We do not weigh choices; we react to whichever is most convenient.

This reduction in our working memory capacity has real implications for how patient-facing information is designed. Can we deal with shared-stem questions with multiple sub-items? Can we process multiple response choices? Can we make sense of a long list of bullet-pointed possible side effects in a leaflet?

Making sense of this has also made me ask myself something personal.

I have a library.

Not a metaphorical one, but a real one. Over 7000 books and I have a further 4000 in my audio library. What does that mean for me?

Let us say a typical book of 300 pages with 500 words per page contains around 150 000 words. If I finish one book every 10 days, that is 36 books a year. At this pace, it would take 194 years to read them all.

For audiobooks? A 10-hour title a week, 52 a year. This would be about 70 years of listening.

It would take two centuries to get through them, if reading and listening in parallel.

But the strange and very human thing about this is that I do not want to finish them all.

A library is not a checklist. It is a landscape. A place to live in. A map of myself, a one I wander in.

Even amongst the books I have read, a sad truth remains that I, or any other human, remember very little of them. We retain around 10-30% immediately after reading a book. A few weeks later, it drops to around 5%. After a few years? This falls to under 1%.

But some things do stay. These are not facts or plots. They are images, feelings, metaphors: emotional residue. Something you needed to hear right at the time of reading it. A single paragraph can reorient an entire worldview. Stuff that does not stick to memory, but clings onto the soul. That is what remains.

So, if machines can remember everything, why shouldn’t we let them? Why not outsource our working memory? Why not let the machine carry the weight?

Our capacity is small. Their capacity is massive. Why pretend otherwise?

There are definitely advantages: Offloading data lightens our load. Machines do not sleep. They are accurate and consistent. They can filter options, retrieve facts and hold entire libraries: just think, 4000 audiobooks on my Audible. Supposing one book weighs a kilo. That’s 4 metric tonnes I would otherwise be carrying about.

On the other hand, outsourcing too much takes something vital away from us.

Working memory is not just short-term storage, it is where intuition forms, learning, judgement. It is where thought is turned into insight.

If a machine holds my choice, how can I know which one is right?

If a machine recalls the details, will I stop noticing what is worth remembering?

That is the cost.

Maybe there is a middle way:

Let the machines hold the noise, while humans focus on the signal.

Let them store the map, but humans choose the road.

Let them remember the facts, but humans hold onto what the facts mean.

It is not about working memory. It is about identity. Thinking is a muscle. A coming into being.

I will not finish my library because that was never the point of it. It is there to walk through, to return to, again and again. You never read the same book twice, so the saying goes. And after me, maybe those books will be scattered donated, forgotten or dumped in a landfill. In the meantime, they remind me of something technology never can do. That to think is not just to know. It is how you feel your way forwards, under the weight of so many physical and cognitive limitations, with the light of a being that is alive.

Thinking in the age of outsourcing is about what we choose to do with knowledge.

The real threat is not forgetting, but that we stop engaging. That we stop thinking about ideas, comparing viewpoints, not rushing to conclusions. Not simply recalling facts, it is fashioning them into a form of something that lives, something that changes us.

Thank you for reading,

Mark Gibson

Leeds, United Kingdom, April 2025

References:

Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review, 63(2), 81–97. https://doi.org/10.1037/h0043158

Sparrow, B., Liu, J., & Wegner, D. M. (2011). Google effects on memory: Cognitive consequences of having information at our fingertips. Science, 333(6043), 476–478. https://doi.org/10.1126/science.1207745