Why AI Can't Save Your Graduation Project / AI 为什么救不了你的毕设？

I was a student from 2014 to 2022. When I think back to those years, I remember small, specific things, like takeaway meals that cost ten yuan. On a lucky day, they might even be free. Those subsidized meals are part of how I remember student life.

I also remember the educational atmosphere as fairly open. I got along well with my teachers, and those relationships have stayed strong since graduation, right up to today.

I don’t think the graduation projects back then were better than they are now, either.

At least in terms of technical execution, visual impact, and level of finish, much of the work I see today has surpassed what we made. When it comes to the choice of subject, the expression of ideas, and the depth of thought, I can’t rank the two generations. Good relationships between teachers and students don’t directly translate into better work, either.

So when I discuss AI and graduation projects today, I don’t want to begin with the claim that students are worse than they used to be. What interests me more is this: as the work gets better, how much can it really tell us about the education behind it?

A piece of work can only tell us so much. It may reflect training at school, but also knowledge the student acquired elsewhere, help from peers, and capabilities provided by tools. When all these come together in the same image, video, or interface, it becomes hard to separate their contributions by looking at the finished result.

AI has lowered some of those barriers further. People without a background in design may be able to use it to improve a poster or try effects that were previously difficult for them to produce.¹ When a school presents work like this as evidence of its teaching, it needs to be even clearer about what students actually learned there.

This can certainly be good for students. Spending less time held back by software or limited production resources may leave them more time for what they really care about. But it also raises a question for schools: when some of these results can be achieved through other means, what exactly do four years of formal study provide?

Another, more impressive graduation exhibition cannot answer that question on its own.

As I understand it, a graduation project should give students a chance to bring together things they have learned separately and try to make something they are willing to take responsibility for. In the process, they need to make choices. They will also discover that some of their original ideas don’t hold up, discuss them with others, acknowledge failures, and decide how to continue.

The final work matters. How well the ideas have been realized, how the materials have been handled, and whether others can make sense of the work all deserve serious assessment. But a polished final product does not reveal everything that was learned along the way. Some important changes only become visible in how students make choices, revise their work, and explain their decisions.

Letting students explore for themselves means allowing them to take wrong turns. An unsuccessful attempt may still help them understand a problem. But if none of this counts when the work is graded, and only the polish of the final product matters, then some of what the school wants students to learn never enters the assessment.²

At that point, AI can keep improving the product, while the learning that was overlooked remains overlooked.

Using AI to compare options or check an explanation that doesn’t make sense still involves thinking. But if students also let it answer why a question is worth pursuing or what makes a result good, they may never confront the things they most need to work out for themselves.

They may do better on the assignment in front of them without becoming better at dealing with the next problem they encounter. Whether AI hands over an answer or helps students find one for themselves can also affect what they retain from the experience.³

If a school asks students to express themselves, then criticizes their expression for being clumsy or immature, what are they supposed to do? The safer choice may be to keep themselves out of it and make something familiar and orderly, with no obvious major flaws.⁴

It is easy to set requirements for a graduation project. Independence, originality, completeness, depth: these words take up very little space in an assignment brief. Helping students gradually develop those abilities takes sustained teaching, specific feedback, and time to make mistakes and revise their work.

All of this has a cost.

How many students a teacher supervises at once, and how much time they can give each project for discussion and revision, both affect what a graduation project can achieve. When schools set their requirements, they also need to make time for this work, rather than simply relying on teachers and students to try a little harder.

A shortage of teaching resources is not always unavoidable. Keeping costs down through teacher pay and staffing can be a deliberate choice by those running a school. Such practices were documented in earlier research on private higher education institutions.⁵

Poor quality therefore cannot always be explained as a teacher’s inability to teach or a student’s unwillingness to learn. Decisions about how to allocate resources have consequences too. If reduced investment harms teaching and administrators are still willing to accept that cost, responsibility needs to be traced back to those making the decisions.

Good teaching is not simply a matter of spending more, and schools do face real limits on resources. But those limits should be honestly acknowledged when judging educational outcomes. A school cannot cut back on the conditions needed for supervision while leaving teachers and students to meet the same expectations, or even higher ones.

If gaps in training have to be filled at the end through students’ independent learning, paid services, or tools, then the more impressive the graduation work looks, the more important it becomes to explain how it was made. A school can be happy about its students’ growth, but it should also answer a question: what, specifically, did it contribute to that growth?

Students can learn to make things outside school, but their degrees still require approval from the institution awarding them. Schools set specific standards and then use them to decide whether students qualify.⁶ Each requirement for a graduation project is therefore more than a teacher’s opinion about the work; it may also affect whether a student receives a degree.

Since schools both educate students and help judge whether that education has succeeded, they should explain how these requirements help students learn, and on what basis they demonstrate that the students have learned.

For those running institutions that depend on tuition fees, continuing to attract students is a real business need. Graduation work may also be used to demonstrate a school’s value. These interests do not necessarily conflict with education. The concern is whether the need to showcase results begins to outweigh the need to educate students. Once a school cares more about whether the results look respectable, while being unwilling to invest in the process that produces them, the graduation project can easily become a task where the students’ job is to make the work and the school’s is to explain what it means.

AI may even make this tension less visible for a while. The work gets made and the exhibition goes smoothly, making it easier to lose sight of the guidance, discussion, and practice that were missing from the teaching. The problem is still there; the finished work may simply not reveal it.

Still, what students take away from school is not all contained in that one piece of work.

Over the years, news reports and experiences around me have increasingly left me with a sense of wariness between teachers and students. That feeling has been especially strong for me since the pandemic. I don’t have data showing how much it has increased nationwide, and I don’t think complaints or reports of misconduct necessarily mean that relationships have deteriorated. Necessary oversight should exist. But if every disagreement first makes both sides think about protecting themselves, it becomes difficult to build the trust needed to discuss the work.

Students need to be able to say, “I haven’t figured this out yet,” and teachers need to be able to say, “I don’t know about this part.” If neither can say these things openly, both may become preoccupied with looking as though they already meet expectations. The discussion then struggles to reach the places where help is actually needed.

These relationships are part of why I still care about graduation projects. Many years after graduation, a piece of work may become an important link between a student and their teachers, classmates, and school.

I would call this a “memory dividend.”

Some works are not particularly good when they are first completed. Yet when we return to them years later, there may be more in them that matters. An effect that now looks dated can remind us why we once liked it. A detail that barely worked may be tied to a discussion or someone’s help. Keeping the work gives us another way back into that experience.

This value will not be fully reflected in the grade awarded at graduation. It may emerge slowly after leaving school, as a person changes. When someone is willing to talk about their work again and get in touch with the people they worked alongside, the school continues to be part of their life in a very concrete way.

This cannot excuse poor teaching. For an experience to be worth remembering, someone needs to have taken the student seriously, made a real investment in them, and allowed the work to reflect their own choices. Time does not turn every experience into a good memory. Indifference and humiliation can be remembered for a long time too.

AI can certainly be part of a work with this kind of value. Students may use it to try things they previously lacked the resources to make, or it may lead to new discussions with teachers and classmates. Those experiences are just as real. The question is whether the tool is part of a learning experience in which people are involved and care about what happens.

When I say AI cannot save a graduation project, I mean that improving the final product cannot automatically deliver the learning the project is meant to support, the test it offers of how well education has worked, or the connections it may leave years later.

If these things are missing, a more complete and beautiful piece of work will not be enough to make up for them. At graduation, schools need to see what students have made. Whether, years later, those students still want to bring out that work and talk about what they learned there and whom they met should also be a reason for schools to take graduation projects seriously.