Tue Mar 10

It’s tempting to offload your thinking to AI. Cognitive science shows why that’s a bad idea

Written by: Misia Temler, Research Affiliate, Psychology, University of Sydney

With so many artificial intelligence (AI) products on offer now ^[1], it’s increasingly tempting to offload difficult thinking tasks to chatbots, agents and other tools.

As we chart this new technological terrain, more and more we’re exposed to vast amounts of information and highly sophisticated software that offers to do the thinking for us. In just a few seconds, tools such as ChatGPT, Claude or Gemini can draft your emails, generate a caring birthday message for a friend, or even summarise the plot of that novel you haven’t gotten around to reading.

Such increased offloading has raised the fear that people will become overly reliant on AI. This could have unintended consequences, such as eroding our critical thinking skills and declining our overall cognitive ability ^[2].

This fear is not unfounded. Research from our lab ^[3] suggests the online environment exploits our cognitive tendencies – individual differences in how we think, perceive, pay attention and remember. In turn, some people end up taking more mental shortcuts and only engaging with information superficially. Other studies have linked high AI use to increased laziness ^[4], anxiety ^[5], lower critical engagement and feelings of dependence.

Yet it may be how we use AI ^[6] that’s the problem, rather than the fact we do it at all. Generally, relying on external sources is fine – we do this constantly. But it’s important to remain in control of what we choose to offload, and why.

How do we even know things?

We all constantly rely on each other’s knowledge to function as a society. Doctors provide medical information, engineers are in charge of construction, financial advisers give investment tips, and so on.

All this spread of expertise ^[7] provides each of us with more knowledge than we can individually hold. In other words, we constantly balance offloading (letting someone else do the thinking) with scaffolding (relying on external knowledge sources to enrich our own thinking).

Scaffolding often happens when we learn ^[8]. For example, a teacher doesn’t write an essay for their student – instead, they provide feedback so the student can connect, integrate, and grow their knowledge base.

Crucially, we also don’t offload all thinking tasks to one specific person. Instead, we carefully consider the person’s trust and expertise ^[9] before accepting their advice, tools or support. We also check how the new information fits in with what we already know.

As our knowledge grows in a certain area, we rely less on outside support, just as a student relies on a teacher until they learn enough to stand on their own.

It’s not just our brains doing the work

Cognition ^[10] (our thinking skills) is the central concept in all of this. Our minds engage in three fundamental tasks:

encoding information (taking it in so the brain can parse it)
storing information, and
retrieving information.

Cognition relies on how well these three mental tasks work together. When we’re overwhelmed with information, distributing tasks to outside sources ^[11] lessens that mental effort.

Research shows when our attention is strained, our minds focus more on encoding information while sacrificing storage and retrieval ^[12], which are more taxing.

Intuitively, it’s easy to assume all our cognition just happens in the brain. But our cognitive processes are sometimes extended to things in the environment ^[13]. These external sources can be people, physical objects and digital tools. A diary is an extension of your mind if you use it to retrieve memories you’ve written down.

However, flippantly offloading your knowledge acquisition and storage to external sources – such as asking ChatGPT any question that pops in your mind – can have an impact on your critical thinking skills. This is because acquired knowledge actively interacts ^[14] with newly encoded information in our minds: we convert information we come across in a way that makes sense to us ^[15].

And the more knowledge we hold, the greater our capacity to encode and critically interpret new information. For example, knowledge of Hitler and Mussolini in the context of the second world war helps us to better understand the modern dangers of dictatorship.

Hard work can be rewarding

To restore balance, we need to perform the more difficult cognitive tasks ourselves, not just offload them whenever it’s convenient.

The faster and easier option isn’t always the best – just like choosing to walk to your friend’s place provides better exercise ^[16] for your body and mind than driving there does.

Sometimes hard work ^[17] can be rewarding. When faced with using AI tools, you can either choose to control them, or let them control you.

One way to balance your relationship with AI tools is to use reflective practices ^[18]. Ask yourself: how do you feel after using AI? Do you feel proud and satisfied, or do you feel more anxious and more overwhelmed? Have you replaced or scaffolded your cognition today? What tasks can you do to expand your mental capabilities tomorrow?

For a successful relationship with AI, we need to exercise all our mental skills – otherwise we really do risk losing them.

This may not always be easy, but it remains in our control.