Han van der Maas | The bar can be raised for student+-
Technological innovations are creating a different type of student, writes Han van der Maas. “Just a little longer and we will invent assignments using ChatGPT which students create using ChatGPT and which we in turn review using ChatGPT.”
What do we really want students to learn? I ended a column on AI in January 2023 with this question, a bit of a lame ending. This question has taken on a new charge with the development of ChatGPT and competitors, but concerns many more tools, including the Internet itself. Despite all these innovations, the final requirements of my educational program, which I myself took before the rise of the Internet, never fundamentally changed.
Relevant to answering this question are the concepts of assisted or augmented intelligence, I prefer to call it IQ+. The operational definition of that is: your score on an IQ test with your phone. My IQ+ was not remarkably high while I was in college, sharing a rotary dial phone on a cord with other students, but has been increasing ever since. With my first cell phone, I suddenly became incredibly good at calculations, my memory took Wikipedia-like forms, and soon I could produce flawless English sentences. With ChatGPT, I can now unleash my phone on all parts of the IQ test. My IQ+ now falls off the regular IQ scale.
Some tools, like the calculator, are just a momentary addition. But sometimes IQ+ flows into IQ, such as when I learn from ChatGPT answers or memorize the Wikipedia information looked up. This IQ, I call it IQ+-, has also increased. In my opinion, IQ+- is just IQ again but many of my peers think otherwise, who believe in “true” intelligence.
Teachers now have a lot to do with the student+, especially when we test without supervision. Today, to promote study success, we give students all sorts of midterm assignments to complete at home. Just a little longer and we will invent assignments using ChatGPT that students make using ChatGPT and which we in turn review using ChatGPT. Bachelor’s and master’s theses, and other individual reports, are also the work of the student+. This has long been a problem, think of the help of highly educated parents and the rampant, expensive dissertation agencies. In that sense, ChatGPT does create a level playing field. For less than a euro a day, you can get rid of your highly educated parents.
When it comes to testing, we can no longer ignore this development: either we accept the student+ who produces a read worthy essay on any topic in an afternoon, or we go back to very old-fashioned forms of testing, such as the oral exam, for example.
I can see some merit in the old-fashioned supervised essay at the end of each academic year. The student, not the student+, is given five hours to read a scientific article and write an essay about it on the spot. In the first year, this may be a summary, but eventually we expect a well-written critical analysis of the scholarly work. Throughout the year, the student has been able to practice and learn with AI tools, and that essay is proof of that. We test the student+-.
We can also have a second five-hour session in which the student+, with phone, is given the same essay as a test. After all, the average employer primarily wants to know what the student+ can do. That I use ChatGPT in my research does not interest the UvA, as long as I advance science.
Thus, the assessment must be radically changed. We better do away with grading work that is not supervised (and limit ourselves to passed or failed). The work only serves to prepare for the real test in the old-fashioned exam hall. There the student+- demonstrates what AI has really brought us in terms of improved learning. I think the exit qualifications can be raised substantially. The modern student has access to the best lectures ever, the best Q&A system ever, the best practice material ever, can program in any programming language, etc. What do we actually want students to learn? More and better. The bar can be raised!
