Cognitive offloading

You tilt your head to read a sideways label instead of rotating it in imagination. You write the shopping list instead of memorising it. You let the phone remember the number. Each is a case of cognitive offloading: getting something outside the head to do work the head would otherwise do.

Evan Risko and Sam Gilbert, who reviewed the research, define it as the use of physical action to alter the information-processing requirements of a task so as to reduce cognitive demand. It is ancient, since writing is the great example, and it is usually a good bargain.

What it does to memory

Betsy Sparrow, Jenny Liu and Daniel Wegner found that people who expected to be able to look information up later remembered the information itself less well and remembered where to find it better. The result became known as the Google effect. It can be read as loss or as sensible economy: memory is being spent on the index and not on the contents.

Risko and Gilbert stress that offloading is a choice, and that the choice depends on a person’s judgement of their own abilities. People offload more when they doubt their memory, whether or not the doubt is accurate.

The philosophical question

Andy Clark and David Chalmers argued that when an external resource is reliably available and automatically relied on, as a notebook is for someone who cannot form new memories, it is not merely an aid to the mind. It is part of the mind. On that view offloading extends cognition and does not diminish it.

Offloading to AI

A notebook stores what you put in. A calculator performs an operation you could specify. A language model does something further: it produces the reasoning, the summary or the draft. What is offloaded is no longer storage or a fixed procedure but the thinking that would have generated the content. The worry is that skills not exercised are not kept, and that a person who has never done the work cannot tell when the machine has done it badly.

Where I stand

I think the right question is not whether to offload but what must stay with the person for the result to be theirs to rely on. On my account of reading an instrument, the warrant for a belief formed with a tool comes from the user’s own competent use of it: knowing what it is good for, when it fails, and how to check. That competence is the one thing that cannot be handed to the tool. Offloading that preserves it is ordinary good sense. Offloading that erodes it leaves a person with outputs and without the standing to trust them.