What is cognitive offloading?

a glossary entry · 17 october 2026

To offload is to put a load down somewhere else. Cognitive offloading is reducing the mental work a task demands by acting on the world outside the head, such as writing a list or setting a phone reminder.

the term comes from cognitive psychology. in a 2016 review, Risko and Gilbert define it as “the use of physical action to alter the information processing requirements of a task so as to reduce cognitive demand.” a shopping list is the plain case. the list holds the items, so memory does not have to.

What does it look like?

these are everyday examples, not findings from a study.

writing a list so the items do not have to be held in mind.
setting a reminder for something that has to happen later in the week.
saving a document before starting on the next one.
counting on fingers, or tilting the head to read a label printed sideways.
asking a model to summarise a long thread or suggest a next step. this one can hand over part of the judgement as well as the memory.

What does the research show?

offloading usually improves performance on the task in hand. the cost, when there is one, tends to show up later, in what a person can still remember without the aid.

in experiments by Storm and Stone (2015), people who saved one file before studying a second remembered more of the second. the benefit did not appear when saving was made unreliable. the authors read this as saving reducing interference, so the saved material stopped getting in the way of what came next. later experiments have replicated the effect, with some limits.

in online experiments by Gilbert (2015), people set more reminders when there was more to remember or a greater chance of being interrupted, and setting reminders improved performance.

the cost falls on memory. in three experiments by Grinschgl and colleagues (2021), people offloaded less when offloading was made more costly. they then did worse on the task but remembered more in a surprise test afterwards. the harm was not inevitable. in the third experiment, people who were made to offload but knew a memory test was coming almost entirely avoided it.

the research on unfinished tasks, and on what writing them down does, is in the brain dump guide.

Where does AI fit?

a list stores what a person already decided. a model can produce the draft or the recommendation itself, so offloading to it can include part of the thinking.

in a 2025 survey of 319 knowledge workers by Lee and colleagues at Microsoft Research and Carnegie Mellon, more confidence in generative AI went with less reported critical thinking, and more confidence in their own ability went with more. Gerlich's 2025 survey of 666 people in the UK, a convenience sample, found that heavier use of AI tools went with lower critical thinking scores, with offloading accounting for part of that link.

both rely mostly on self-report and show associations. neither shows that using AI causes a decline, and neither tested any particular product. in Lee and colleagues’ own words, “Our analysis does not establish causation.”

the wider picture is mixed. a 2026 systematic review of 89 studies of university students by Alubthane found the effects of generative AI on thinking “neither uniform nor unconditional”. its author proposes that AI tends to amplify thinking when its use is structured and to substitute for it when it is not.

What it is not

forgetting: the information still exists, held somewhere else. the risk is relying on a store that is missing or wrong.
a sign of a weak memory: the choice to offload follows confidence as well as ability. in the studies Risko and Gilbert review, people with less confidence in their memory set more reminders even after their actual ability was taken into account.
handing over a decision: a reminder holds a choice already made. an answer from a model can contain a judgement the person did not make. the checking guide shows how to inspect one.

A small check you can try

this is an informal habit, not a validated test.

pick one task that now goes to a tool or a model.
before looking at what it gives back, write one line of your own answer or plan.
compare the two. note what the tool added, and what got accepted without a check.
keep what is worth keeping somewhere it will be seen again.

What does nenspace claim?

nenspace is built on the difference between keeping a record and handing over a judgement. /space takes the remembering, with notes, tasks, habits and a logbook. nen is built to hand the thinking back.

that is a design aim. it does not establish that nenspace protects or improves anyone's thinking, and none of the studies above tested it.

try a brain dump · read the extended mind essay · start free