Artificial intelligence has entered the classroom: what is philosophy teaching losing?

In universities, generative artificial intelligence is no longer the technology of the future. It is a condition of this term's teaching.

One of the fields most affected by that shift is, paradoxically, the field best equipped to interrogate technology: the teaching of philosophy.

On 27-28 August the American Philosophical Association (APA) and the American Association of Philosophy Teachers (AAPT) held a joint online forum. Its name carried a Socratic allusion: “The Examined Algorithm”. The subtitle was blunter: strategies for navigating AI in the classroom.

The wrong question#

The discussion usually opens with one question. What should be done if a student has had an essay written by a machine?

The question matters, but it is secondary. It is a question about policing, and on the policing side matters are unsettled. As the dispute over accusations we reported last week made clear, detection tools generate both false positives and false negatives. One philosopher recounted that a text he had written himself was flagged by such a tool because its subject was artificial intelligence.

The real question lies deeper, and a piece published on the APA's own blog on 26 August states it cleanly.

Is writing merely the record of thinking?#

P. D. Magnus argues that philosophical writing is not the setting down on paper of a conclusion already reached. In philosophy, writing is the process of discovering what we ought to think.

His title gives the thesis an image: moving about in the space of ideas. To hand a piece of thinking over to a machine, on Magnus's account, does not make it easier to map that space — it obscures the landscape.

The analogy works like this. When you build an argument yourself, you do not merely arrive at a conclusion; along the way you also see the dead ends, the assumptions that do no work, and the objections that turn out to be stronger than you had supposed. That seeing is the capital your later thinking draws on.

A ready-made text hands you the destination and withholds the route.

What is actually learnt in philosophy#

The purpose of a philosophical education is not to memorise correct answers. These are the skills it tries to impart.

Distinguishing one concept from another. Questioning the grounds of a claim. Stating the opposing view in the form its defender would accept. Noticing the weak link in one's own argument. And, where the case requires it, changing the opinion one began with.

None of these is a matter of outcomes. Every one is a skill of process.

This is why artificial intelligence faces the philosophy teacher with more than a problem of honesty. It brings a more fundamental question: can the act of thinking itself be contracted out?

Two arguments on the other side#

It should be recorded that the argument does not run in only one direction.

The first objection is instrumental. The pocket calculator did not abolish the learning of mathematics; it abolished the labour of calculation and released the mind for higher-order problems. Artificial intelligence might likewise take over the literature search, the summary and the first draft, and leave the philosopher to the real work.

The reply runs thus. A calculator produces a result, and the correctness of that result can be checked independently. A philosophical text is not exhausted by its result, and its correctness cannot be tested in a single operation.

The second objection is egalitarian. For a student who cannot handle academic English as a native speaker does, who is the first in the family to reach university, or who works long hours, artificial intelligence may be a levelling instrument. The objection deserves to be taken seriously, and it is the least discussed side of the argument.

The students' own objection#

A further piece of evidence complicates the picture in an unexpected way.

In a recent article two philosophers reported that a substantial proportion of their students object on moral grounds to using generative artificial intelligence: the environmental cost, the use of data without permission, and the devaluing of labour.

The assumption that every student wants to use this technology and that teachers are the ones trying to hold them back may therefore be false. Some students are conscientious objectors, and they expect their teachers to respect the position.

In conclusion: the question is shifting#

In the years ahead the real question in philosophy classrooms will not be whether artificial intelligence ought to be used.

That question has already been passed; the technology is in the room. The real question is this:

What does it mean, in the age of artificial intelligence, for a person to produce their own thought?

And that is the kind of question philosophy can answer. Nor is it an accident that it has surfaced in the classroom. Like Annette Baier's proposal for a slow philosophy, which we noted in this column earlier this month, philosophy generally has something useful to say once it begins to treat its own professional conditions as a philosophical problem.