A student is learning to cook a new dish. They could follow each step themselves, noticing how the ingredients change and adjusting when something looks wrong. Or they could ask an AI system to plan the meal, set the timings, and tell them exactly what to do next.
Both paths might lead to dinner. But only one necessarily helps the student become a better cook.
That distinction matters in education. AI can help students finish work more quickly, explain a difficult idea, or suggest a next step. Those can be valuable supports. But when a system does the thinking that an assignment was meant to help students practice, completing the assignment may no longer mean the student has developed the intended capability.
Effort by itself is not proof of learning. Some effort is wasted: searching through confusing directions, repeating a task without understanding it, or struggling with a barrier unrelated to the learning goal. Good teaching helps reduce that friction.
But some effort is part of learning. Recalling an idea from memory, trying a solution, noticing a mistake, and revising an approach can help students build knowledge they can use later. Learning can feel slower when students have to do these things themselves, even when the added effort supports longer-term retention.
Researchers Robert and Elizabeth Bjork call some of these conditions “desirable difficulties.” Their research describes how approaches that make practice more challenging, such as spacing study sessions or retrieving information from memory, can improve later retention and transfer, even if they make learning feel less fluent in the moment. Springer Nature Link
The word desirable is doing important work. Difficulty is useful when students can make progress and the challenge supports the learning goal. Confusion without support is not automatically productive. Neither is making every assignment harder to prove that students did it themselves.
AI makes this distinction more important. A student could ask for a full solution and submit it without doing the reasoning the assignment was designed to develop. Or the student could ask for a hint, attempt the problem, and use the response to check their thinking. The same tool can help remove a barrier or remove the learning opportunity, depending on how it is used.
Early research is beginning to examine this tension. In one study, 91 university students used either ChatGPT or a search engine to research a complex science issue. The researchers found that ChatGPT reduced the mental effort students reported, but the students’ recommendations and justifications showed less depth than those produced by students using search. It is a single study with a specific task and an older version of ChatGPT, so it does not establish that AI use always weakens learning. It does show why task completion and cognitive engagement should not be treated as the same thing. ScienceDirect
The goal is not to preserve every bit of difficulty. It is to keep students engaged in the parts of the work that help them grow. AI might explain a concept in simpler language, provide practice questions, or offer a hint after a student has made an attempt. A student can still be responsible for deciding whether the explanation makes sense, showing how they reached an answer, and applying what they learned to a new problem.
This changes the question we ask about effort. Instead of asking, “Did the student do this the hard way?” we can ask, “Which parts of this task were meant to build capability, and did the student have a chance to practice them?”
That question is useful for faculty designing assignments, and for students deciding how to use AI. If the purpose is to learn how to make a judgment, students need to make and explain judgments. If the purpose is to develop a skill, they need practice doing it. If the purpose is to understand an idea, they need opportunities to retrieve, explain, and apply it, not only read a polished answer.
AI can help students spend less effort on tasks that are distracting or routine. The educational challenge is to use that efficiency in service of learning, rather than treating less effort as the goal by itself.
Effort matters when it helps students become more capable. If AI becomes part of how people think, work, and make decisions, what does it mean to learn and what should learning help a person become? That is the question for the next post.
Continuing the Conversation
Series 1: AI Is Exposing Existing Problems ✓ Completed
Series 2: What We Do About It ✓ Completed
Series 3: Cultivating Human Thinking in an AI World ✓ Completed
Series 4: Learning Alongside AI ✓ Completed
Series 5: Preparing Students for an AI World ✓ Completed
Series 6: The Purpose of Education in an AI World
Current Post (6 of 8): If AI Can Complete Coursework, What Is the Purpose of Effort?
Next Up: If AI Becomes Everyone’s Cognitive Partner, What Is the Purpose of Learning?