🎓 Are University Capstone Projects "Obsolete"? What Are We Really Grading?
🏛 In university engineering faculties, we are facing a staggering situation regarding "capstone projects"—often seen as the pinnacle of a student's educational life and where they are expected to put in serious effort for one or two semesters—that we can no longer ignore.
Engineering education, by its nature, aims to impart skills in problem-solving, analytical thinking, and designing a system from scratch. Traditionally, the capstone project is the ultimate test where a student puts all the theoretical knowledge acquired over 4 years into practice and proves they can stand on their own feet. I certainly exclude our students who focus on their projects ethically, with genuine hard work, turning night into day. However, the general landscape in the field and during jury defenses shows that we are frighteningly rapidly moving away from the core purpose of engineering education.
The Next-Generation Landscape Juries Face
When you sit down to evaluate a project, the anatomy of the work presented to you now looks very similar. We can group the flaws of this new era under three main headings:
Frankenstein Projects
When you open the hood and look inside projects that seem highly complicated and appear to use modern architectures from the outside or on the presentation screen, you encounter utter chaos. In reality, there is no "engineering design"; you see that many ready-made infrastructures are somehow "stitched together" using Large Language Models (LLMs) or API requests. There is no logical coherence between code blocks. Systems that fully meet no technical requirements (performance, security, scalability) and run almost by pure luck are brought before the jury as a "success."
Candidates Alien to Their Own Reports
In the past, even if projects were incomplete, the student knew where they made a mistake in the code they wrote or the circuit they built. Today, however, when we ask students about the most fundamental components of their project or their engineering choices at critical decision points, they cannot explain them. Questions like "Why did you choose this algorithm?" or "How will this database architecture handle this load?" remain unanswered. The saddest part is realizing they haven't even read or understood the statements, numbers, and tables in the brilliantly academic-sounding report (written by AI) they claim to have "prepared themselves."
Passing the Buck in Defenses
An engineer is someone who puts their signature under their work and takes responsibility. Yet, the answers we get to our questions during the jury usually point to a "third party": "It wrote it like this", "It gave these results", "When I fed the data to the system, it generated this graph". We frequently and sadly see that AI has ceased to be a "tool" that accelerates the learning process and has become a "subcontractor" onto which all cognitive action and responsibility are dumped.
We could multiply these examples for pages with anecdotes from the field. Artificial intelligence will not leave our lives; on the contrary, it will become even more integrated. Rejecting this is nothing more than Luddism. However, there are crucial questions that academia and the industry need to ask themselves and urgently open up for discussion.
❓️ The Big Questions
We must approach this problem not with outdated rules like "ban artificial intelligence," but by completely changing our evaluation metrics. Perhaps capstone projects should no longer be months-long coding assignments; they should transform into System Design interviews where the candidate defends an architecture by drawing it on a board live in front of the jury and explains how they used AI "correctly and ethically."
I am very curious about the observations, experiences, and most importantly, the proposed solutions of my esteemed colleagues in academia and the industry regarding this issue. How do you think we will get out of this situation? Let's meet in the comments! 👇
