When does learning become work?

In a recent piece, I wrote about the contradiction that emerges when faculty distrust student work because AI may have contributed to it, then turn to AI to help evaluate those same students. The responses to that piece were useful, including the disagreements. One reader offered a straightforward explanation: faculty already know what they are doing, so they can take shortcuts. Students are learning, so they need to do it the long way. The central issue, they suggested, is skill development and effortful practice. It’s a good distinction and one I use in my own teaching.

In my grant-writing course, there are points when I ask students to do the work the long way. They need to wrestle with how a documented need connects to a program design and whether the proposed activities might plausibly produce the intended outcomes. A generative AI system can produce polished language for those sections in seconds. It can also produce a proposal whose individual sections sound credible but do not hold together as a program. If students have not worked through the model themselves, they may not be able to see the problem. Elegant prompting doesn’t always fix this. The course is constantly doing two things at once in terms of writing and design, and my feedback flows between the two.

That seems to support a clear rule: students are learning, so they need to do the work in long form. Faculty and other professionals have already learned, so they can use tools to work more efficiently.

The problem is that I cannot identify the point at which learning becomes work.

My students are learning to write grants, but they are also producing the same type of document I produce as a grant-writing consultant. When I write a grant, I am doing professional work. I am also routinely learning while I do it. I have to learn the agency I am working with and figure out how to present what they’re trying to accomplish. My professional status does not mean the learning process is over. In fact, I often have to learn and adapt quickly. I distinctly remember presenting a first pass on a grant application for a community agency and having half the team be completely turned off by the proposed strategy. I thought I had gotten it right, but we had missed each other in the development process. I had to adapt and learn anew.

Now, learning about a new agency is different from learning how to write a grant in the first place. My experience matters. But being experienced doesn’t settle which parts of a particular task I can hand over to AI. Both students and professionals may use it to deepen their thinking, and both may use it to avoid thinking.

Colleagues at BetterUp use the terms “Pilots” and “Passengers” to describe different responses to AI in the workplace. I find the metaphor useful, although I use it somewhat differently here. A pilot does not need to manually perform every function. The pilot does, however, need to understand the destination and check the instrumentation. They are ultimately responsible for the flight itself. Passengers just kinda ride along. This has less to do with how frequently someone uses AI than with whether they remain present in the work.

A student can use AI and remain intellectually engaged. They can ask the system to challenge an idea and decide whether its suggestions make sense. We ask them to do this in class when we give feedback and ask for revisions. We’re humans, but the student still has to think through what to do with the feedback. A student can also ask AI to generate an answer they do not understand and submit language they cannot meaningfully defend. Stating that students shouldn’t engage with AI at all risks cutting them off from useful tools to improve their work.

I can hear my critics already: “But students might produce better work without actually learning anything!” Yep, that’s true, no argument there. As Brian Perron keeps saying, the incentive system around the tool matters. Our primary job is to develop students’ thinking. A tool can augment that process or impede it.

I use AI in my own work. Sometimes it helps me organize ideas. I also use it to smooth audio-transcribed feedback after I have evaluated student work. That can save time, but I still have to make sure the feedback says what I mean and accurately reflects the student’s work. If I stop doing that, my status as a faculty member does not somehow make the use responsible. I have simply become a passenger with more authority.

This is why I do not think “students are learning while faculty are working” can carry the full ethical weight we sometimes place on it. It explains why a particular task may need practice to be learned. It does not establish that student use is inherently suspect or that professional use is inherently appropriate.

The more precise question is: What part of this work does the person need to think through for themselves?

Sometimes the answer is basically all of it. A student cannot learn to formulate a clinical impression if the system does the reasoning and they never have to work through it. They need enough direct experience to recognize when something only sounds convincing. However, doing something without AI is not automatically educational. Inefficiency is not the same as learning.

At some point, competent use of available tools also becomes part of the skill. A future grant writer needs to know what feels like it’ll “land” with a funder and whether the model actually represents the program’s intentions. They also need to do this at speed. Preventing all AI use may make it easier to evaluate individual authorship, but it does not necessarily prepare students for the conditions under which they will actually work.

This leaves faculty with a more difficult design problem than simply allowing or prohibiting AI. Before setting a rule, we need to be clear about what an assignment is supposed to teach and where students need to do the work themselves. How will they demonstrate that they understand what they submit? This has been true for a while, but AI has foregrounded the topic in a way few of us were fully prepared for.

Those questions should apply to faculty use as well.

Some assignments should require students to work without AI. Others should teach them how to use it thoughtfully. The important part is that we can explain the difference. Our goal should not be to produce students who can complete a task only under classroom conditions that disappear once they enter professional life. Nor should we be satisfied with polished products without the competence behind them. We are trying to prepare people who can use powerful tools without surrendering their personal authorial responsibility.

Disclosure: AI tools supported the development and editing of this piece. I developed the core argument, wrote the work, and take responsibility for the final text.