How AI tools can affect critical thinking

With the growing range of artificial intelligence (AI) products available, handing demanding thinking tasks over to chatbots, agents and similar tools is becoming ever more appealing.

As we navigate this unfamiliar technological landscape, we encounter increasing volumes of information alongside exceptionally capable software that claims it can think on our behalf.

Within seconds, tools including ChatGPT, Claude and Gemini can compose emails, create a thoughtful birthday note for a friend, or summarise the storyline of a novel you have not yet found time to read.

This greater tendency to delegate work has prompted concerns that people could become too dependent on AI. The possible unintended effects include weakened critical-thinking skills and a reduction in overall cognitive ability.

These concerns have a basis. Research conducted in our laboratory indicates that the online environment takes advantage of cognitive tendencies: the individual ways in which we think, perceive, focus our attention and remember.

Consequently, some people take more mental shortcuts and interact with information only at a surface level. Other research has associated heavy AI use with greater laziness and anxiety, less critical engagement, and a sense of dependence.

However, the issue may lie in the way we use AI rather than in using it at all. Depending on outside sources is generally acceptable, and something we do continually. What matters is retaining control over what we decide to offload and the reasons for doing so.

How do we know what we know?

For society to work, we continually depend on one another's knowledge. Doctors supply medical information, engineers oversee construction, financial advisers offer investment guidance, and so forth.

This distribution of expertise gives every one of us access to more knowledge than any individual could retain alone. Put differently, we constantly strike a balance between offloading - allowing somebody else to do the thinking - and scaffolding - using external sources of knowledge to strengthen our own thought.

Scaffolding commonly takes place during learning. A teacher, for instance, does not write a student's essay; they give feedback that enables the student to link, integrate and expand their knowledge base.

Importantly, we do not delegate every thinking task to a single individual either. Before taking on someone's advice, tools or support, we carefully assess their expertise and trustworthiness. We also consider whether new information accords with what we already understand.

As our expertise in a particular field develops, our reliance on external support declines. This resembles a student depending on a teacher until they have learned enough to manage independently.

AI tools and the work of cognition

Cognition - our ability to think - sits at the heart of this issue. Our minds perform three essential functions:

• encoding information, meaning taking it in so the brain can process it;
• storing information; and
• retrieving information.

Cognition depends on how effectively these three mental functions operate together. When information becomes overwhelming, allocating tasks to external sources reduces the mental effort required.

Studies show that when attention is under strain, the mind gives greater priority to encoding information, at the expense of storage and retrieval, which demand more effort.

It may seem obvious that cognition happens entirely inside the brain. Yet our cognitive processes can sometimes extend into our surroundings. Those external sources may include people, physical objects and digital tools. If you use a diary to recall memories you recorded in it, that diary functions as an extension of your mind.

But casually delegating the acquisition and storage of knowledge to external sources - for example, asking ChatGPT every question that enters your mind - may affect your critical-thinking abilities. Knowledge we acquire actively interacts with newly encoded information: we interpret information we encounter in ways that are meaningful to us.

The more knowledge we possess, the stronger our ability to encode and critically interpret fresh information becomes. Understanding Hitler and Mussolini in the context of the Second World War, for example, helps us recognise the present-day dangers of dictatorship more clearly.

The rewards of difficult thinking

To regain balance, we must take on the more demanding cognitive tasks ourselves rather than automatically offloading them whenever doing so is convenient.

The quickest and simplest choice is not necessarily the best one. Walking to a friend's home, for example, offers more exercise for both body and mind than travelling there by car.

Effort can sometimes bring rewards. When deciding whether to use AI tools, you can choose to direct them, or allow them to direct you.

A useful way to balance your relationship with AI tools is through reflective practice. Consider the following: how do you feel after using AI? Are you proud and satisfied, or more anxious and overwhelmed? Have you replaced your cognition today, or scaffolded it? Which tasks could you undertake tomorrow to broaden your mental capabilities?

A healthy relationship with AI requires us to exercise every mental skill; otherwise, we genuinely risk losing them.

That will not always be straightforward, but it is still within our control.

Misia Temler, Research Affiliate, Psychology, University of Sydney

This article is republished from The Conversation under a Creative Commons licence. Read the original article.