Watch Waveriding or Get Wet?

Does use of artificial intelligence necessarily pose the risk of diminishing critical thinking or thinking skills? The answer might well depend on how AI is used.

But that is true of many human endeavors. People rarely stop engaging in activities they find intrinsically rewarding simply because technology makes the outcome easier to obtain.

As much as I enjoy this clip of waveriding at a favorite spot, I'd much rather be doing it.

Humans experience deep satisfaction when engaged in challenging activities that match their abilities. The reward is not merely the finished product but the experience of making it.

That is why people:

• climb mountains despite vehicle access

• bake bread despite supermarkets

• garden despite grocery stores

• play golf despite television coverage. 

The activity itself provides enjoyment. If people only cared about outcomes, very few would actually play sports.

And AI may affect creative work in much the same way.

Calculators changed mathematics, for example. They reduced arithmetic effort but did not eliminate the need to formulate problems or interpret results. 

There is a reasonable argument to be made that outsourcing “writing” to a language model poses a risk. 

For many writers, that is a relatively negligible risk, since many writers compose because they enjoy the process of writing, and it makes no sense to outsource the “fun” of writing at all. 

For many writers, writing is enjoyable because it combines:

• discovery

• exploration

• self-expression

• problem solving

• craftsmanship.

The point is that technological advances rarely eliminate hobbies.

• Photography did not eliminate painting

• Recorded music did not eliminate amateur musicians

• Power tools did not eliminate woodworking

• GPS did not eliminate hiking

• Word processors did not eliminate writing.

For many forms of knowledge work, AI is not replacing thinking so much as changing where the thinking occurs. 

Instead of spending most of one's effort locating information, more of the cognitive effort shifts to:

• Framing the problem before asking AI

• Evaluating and challenging the response afterward

• Synthesizing the results into an original conclusion.

So, in many research and knowledge-work settings, AI functions less as a replacement for thinking than as an accelerator for information retrieval and synthesis. 

The important intellectual work frequently occurs before the AI interaction (defining the problem, framing the question) and after it (evaluating, integrating, and applying the results).

That is similar to how experienced researchers have long used search engines and databases, for example, and suggests that “how” AI is used matters.

The important question is therefore not "did AI do part of the work?" but " which parts of the thinking did the human still perform?"

Researchers may spend less time collecting information and more time deciding what the information means. They might ask more or different questions.  

And experts might benefit from using AI more than novices, as they are able to formulate better questions, based on:

• relevant mental models

• domain knowledge

• intuition about good questions

• ability to detect errors. 

Stage	Traditional research	AI-assisted research	Where critical thinking occurs
Define question	Formulate hypothesis	Formulate prompt/problem	Very high
Gather information	Library, databases, Google	AI search or LLM	Moderate
Evaluate evidence	Read sources	Verify AI claims and sources	Very high
Compare viewpoints	Read competing authors	Ask AI to generate opposing views	Very high
Draw conclusions	Human synthesis	Human synthesis	Very high
Communicate findings	Human writing	Human writing (possibly AI-assisted editing)	High

Perhaps there is an analogy to the use of “search.” When Google became dominant around 2000, educators raised similar concerns about search making us dumber. 

Instead, search shifted the balance of cognitive work away from memorization and toward higher-level reasoning.

AI might be similar, in many instances.

Thought traditional search asks users to perform much of the information synthesis themselves (find sources and evaluate them), 

AI saves time by summarizing results.

But it does not eliminate the need to ask:

• Is this correct?

• Is something missing?

• What assumptions underlie this answer?

• What evidence contradicts it?

• Are these the strongest sources?

Skeptics will note that many users will not take the time to do so. But that arguably was the case beforehand. 

Rather than replacing reasoning, AI often expands the range of questions that can be explored within a fixed amount of time. 

Also, there are some techniques that encourage broader and deeper exploration.

Recent research describes phenomena such as "cognitive offloading," "epistemic atrophy," and an "illusion of understanding," where users mistake fluent explanations for genuine comprehension. These effects appear strongest when AI substitutes for independent evaluation rather than supporting it. (Business Insider)

For experienced researchers, AI is often best understood as an unusually capable research assistant rather than an autonomous thinker. 

But the researcher still bears responsibility for asking the right questions, verifying sources, weighing evidence, and integrating insights into an original conclusion.

In that sense, AI resembles an evolution of search rather than a replacement for thought. 

The cognitive work shifts away from locating information and toward framing, evaluating, and synthesizing it. 

AI substantially reduces the effort required for search, retrieval, summarization, and drafting, but arguably does not eliminate the need for problem formulation, judgment, skepticism, synthesis, or decision-making.

Whether critical thinking declines depends less on the technology than on whether users treat AI as an answer machine or as a research collaborator whose outputs require evaluation.

For writers who enjoy the process of writing, AI is not a replacement, anymore than watching surfing is a replacement for surfing.