Suppose you witness an accident and have to report it to the police. “Well, the red car drove with high speed towards me and then took a sharp left turn”, you exclaim. You try hard to find the right words to capture the precise sequence of events and people you’ve noticed. Being a bit fussy by nature, you keep correcting yourself. “To begin with, it seemed as if the driver wanted to run me over.”, is the formulation you eventually settle on. – Now imagine you try to refine your impressions using ChatGPT. Obviously, there is always room for improving on style and grammar. But can you expect ChatGPT (or any LLMs or AI) to improve on the accuracy of your factual statements? No. But what does it mean if it is used to that end anyway?

Given the way ChatGPT works, it has no experience of the world in any sense. Rather, it generates sentences by predicting the most likely sequence of words based on the input it receives and the vast amount of text it was trained on. Thus, it cannot improve on the content of any statements relying on experience. While this is no surprise, the repercussions it has for teaching contexts deserve careful attention because such contexts thrive on the correction of statements. Especially now that people not only use this device to hide their plagiarisms, but also to “decide” all sorts of questions. I’ve been wondering repeatedly what precisely it is that goes wrong in teaching contexts with the use of AI and now I begin to think that it comes down to a loss of corrigibility, a loss of understanding what corrigibility even means. Put in a nutshell, seeing that this device improves the form of (written) texts in amazing dimensions, it makes us blind to the fact that it impoverishes our relation to empirical content. In what follows, I’d like to explore this loss with regard to teaching philosophy.  

What is corrigibility? Corrigibilty means that a statement or text can be corrected. If I state that it’s raining you can correct me by pointing out that it’s in fact not raining. We offer and receive corrections all the time. We improve the way of phrasing something we’ve seen by finding more adequate phrases. We can differentiate between grammatical and stylistical corrections as corrections of form as opposed to content, but often the two are difficult to keep apart. The phrase “it’s raining” is formally correct when used among English language speakers, but what makes it correct for these users is how it’s applied to a shared experience of the world (in which it happens to rain). If I ask you to refine your phrasing, suggesting for instance that it’s really pouring and not just raining, I can mean at once to pay attention to your experience and the conventional way of expressing such an experience in the English language. When you think about your experience and modes of expression, you’ll likely involve linguistic sources (your language conventions, literature, the sociolect your audience is supposedly expecting etc.) as well as non-linguistic sources (whatever you can gather from other sense-modalities). Most importantly, you’ll involve relations of applying linguistic resources to non-linguistic experiences. In other words, we relate linguistic conventions to (non-linguistic) facts. ChatGPT, by contrast, doesn’t do that. Having no relation to the world, it is confined to linguistic resources; it has no other sense modalities and it has no way of relating linguistic to non-linguistic facts. In other words, while it can improve on formulations, it cannot be corrected. Put in Wittgensteinian terms, whatever seems correct to ChatGPT is correct – and that means that there is no sense of distinguishing between correct and incorrect. (There is an intriguing piece about learning this the hard way.) Thus, we shouldn’t even say that it’s “hallucinating” when it’s “making things up”. There is no meaningful distinction for this device between hallucinating and getting things right in the first place.

Now I doubt that I’m spreading any news here. So why is this worth saying? Because both the language of ChatGPT and of the merchants of this technology constantly suggests that this device is learning, being corrected and improved. Yes, it’s being improved at what it does already, but it’s not improved in any other sense. This lingo tricks many of us into thinking that the improvement is of the kind that we are familiar with. Just like AI is now increasingly taken to be a meaningful, sexy or caring interlocutor, it tricks many of us into assuming that it could “learn” by being “corrected”. But learning, for humans, always involves a relation to the world. The great confusion about ChatGPT, then, is that it would be improved in any way that we would try to improve our own way of expressing ourselves.

How does this affect teaching (philosophy)? There are many pieces about the decline of the humanities in the face of ChatGPT and related devices. Given how this technology diffuses our sense of authorship and our reading culture, I’m inclined to think that our whole way of cherishing text production and reading will go out of fashion and become a nerdy niche. Just like long electric guitar solos or keyboard solos, which seemingly were ubiquitous in the 70s and 80s, are now a thing for a few nerds on youtube. So as I see it, the problem is not that students are faking texts; the problem is that most texts are considered irrelevant. Along with the skills and features that go into their production. Being able to write good texts is already irrelevant in world where so-called leaders get by without even glancing at their briefings. But let’s stick to the current story. My hunch is that the loss of corrigibility ingrained in ChatGPT is the outcome of a larger trend that was clearly recognised in Harry Frankfurt’s On Bullshit as early as 1986: Once you realise that you can convince without sticking to techniques of truth-evaluation, you can disregard truth altogether. After all, the question is not in what way ChatGPT is incorrigible. We can figure that out quickly. The question is why are we letting ourselves be corrected by a device that is incorrigible.

But that’s a question for nerds. Mastering long written texts, let alone writing them, then, doesn’t seem to hold much of a promise for anything now. This is not just because students have incentives to fake their work; it’s because there are hardly any incentives to produce such work in the first place. Why do you need to learn to play the piano if you have keyboards with automatic accompaniment? Of course, you might get sick of their sounds quickly. But who cares if that’s all that’s on offer?

So again: the problem is not cheating; it’s irrelevance. Writing this, I feel like a fossil decrying the loss of its natural habitat. And that’s probably what it is: An old man whining that no one recognises the beauties hidden in the art he cherishes. So what? So what indeed?

So what’s left for teachers? If you don’t worry too much about plagiarized texts, you might adjust your energy towards getting people to think, not by by analysing texts, but by coming up with good prompts for ChatGPT or by enhancing your techniques of video editing. In other words, while certain products (such as well-written essays) will simply be done by ChatGPT in the future, you can support students in improving “their” work by focussing on helping them to use this and the AI devices to come as a good tool. The remaining question is, though, what this tool is good for, once we admit that writing texts is irrelevant?

A question for scholars. – How can we spend a lifetime on a chapter in Aristotle and think we’re done with a student essay in two hours? Both can be equally enigmatic.

Writing this little joke back in 2018, ChatGPT was still unheard of in my part of the world. My point was that our teaching practice results in our unlearning to read and encourages mindless writing and reading. Back then, people responded by emphasising that, contrary to texts of “proper” philosophers, student work is being produced and read to be judged with regard to specific skills, so it doesn’t merit further attention. With the advent of ChatGPT, the judging part of this kind of exercise went down the drain. But even back then, the thought that sparked the worry behind my joke was that we have students produce texts that no one wants to read and, basically, that we train forms of writing that no one wants to read. After all, we now know that it’s not only student papers that often get no more than a quick glace, but equally work of peers. As I see it, then, ChatGPT did not alter this situation but just made our practice of mindless reading and writing more visible. At least, we talk about it now.

If this contains at least a grain of truth, then we knew very well before the advent of ChatGPT that our exercises weren’t very promising. Why? Of course, writing is a great thing and should be practised, but grading writing is another matter altogether. Either our responses would have to be very formulaic or they would have to be so time consuming that no one could serve larger classes. So the problem is not that students now have better ways of cheating. The problem is that we don’t and didn’t act well as readers of our student work. No matter whether we act like cops to catch cheaters or just keep rushing through masses of work: we’re acting as a bad role model for good reading and writing. If we rush through student papers, we demonstrate that we only care about grading. Students learn that they should mainly care about grades, too. It’s no surprise, then, that what gets perfected is not the writing but the techniques of cheating.

But this doesn’t mean that students don’t want to write or learn writing. Rather, they probably don’t want to write for readers who spend two to five seconds on a paragraph that took two days to compose. Perhaps what we (should) really feel, now that ChatGPT makes it almost impossible to distinguish real from hallucinated work, is relief – relief that student essays can’t be graded as they used to be. It should encourage us, not to abolish this kind of exercise, but take it more seriously and stop grading it in the way we used to. While we can focus our common grading practices on other kinds of exercises, we could encourage student essays designed as longer projects for those who really want to go through the effort.

I keep noticing that one of the greatest worries of students is finding a workable research question for their essays – one that allows them to structure their work and keeps them in line with what they intend to promise. I’ve written on this before at length on this blog (here, a Groningen student reviews three of my posts), but I think I managed to break down the issue even more. – I’ve just finished an exciting three-day seminar on “Spinoza’s Ethics. An introduction to doing philosophy systematically for beginners.” The idea was to bring out the systematicity of the Ethics to such a degree that students could continue working with the text independently without loosing grip of the general framework. On the last day, I wanted to turn the tables and have my roughly fourty students work, not primarily through the text, but through potential essays on this book. Hence, I divided them into six groups and gave them pertinent tasks. In a feedback round, most students and I agreed that this was a surprising success. Therefore, I simply want to share the set-up of the tasks:

At the beginning, I asked the students to:

• find some friction (in the text),
• provide concrete evidence for this friction by providing at least two passages or sentences from the text,
• provide a motivation of why the friction arises,
• provide at least one possible answer as to how to state the friction and how to amend it.

By asking for a friction, I could rule out explorative works that have no natural boundaries and can keep students on the preparatory reading path forwever. By asking for concrete passages of text, students will have a constant fall-back place when themes would have them meander into the wilderness of the space of reasons. By asking for friction in the text, I also wanted to make sure that the friction arises from an immanent reading, rather than from an external criticism (e.g. of the sort: I don’t like this kind of philosophy, so here is why I prefer something else – which doesn’t really engage with the material). This approach mostly yields ambiguous uses / understandings of particular terms. Not everyone is interested in that kind of focus, but the promise is that this provides a workable way in and out, while one still has time to draw in all the pertinent aspects related to working through the friction.

Eventually, the student groups came up with six workable research questions. The evidence and motivations made for a pertinent structure, the possible answer for a fairly clear hypothesis. In the discussion of these approaches, we then tried to establish what the chapters / sections should look like: i.e. how they should implement the friction, motivation, and the answers. (In fact, this idea for planning essays is derived from my account of questions.)

Especially for beginners, it’s also important to counter the feeling that they don’t yet know enough to come up with a proper question. I addressed this worry by making clear that any philosophical thesis can (and indeed will be) countered at some point. So one shouldn’t waste their time by trying to immunise their work against criticism. Whatever you’ll say can be criticised. So you might as well get started immediately.

Of course, this is not waterproof. But my sense was that students now had an idea about how to move from reading to planning their writing – and that’s all I want.

***

Many thanks again to my great students in this course!

A: Marriage is a speech act.

B: Why do you devalue traditional rituals?

A: I don’t! I just made a point about the constitutive role of language for marriage.

B: Oh, but why did you say that marriage was nothing but talk?

…

It’s common to paraphrase texts, whether written or uttered, with our own words. In fact, students are often encouraged to “use their own words” when asked to restate or summarise an argument. Such instructions are usually intended to allow for checking whether something has been understood correctly or for translating technical terms into common language. The misunderstanding in the example above can serve as an illustration: While A made a point about the act or marrying requiring the people getting married to actually say “yes”, B took A to mean that marriage reduces to talk and is thus devoid of any further value. So B took A’s utterance in a reductive sense, while A meant it in a constitutive sense. These different senses are brought out in the paraphrases. B specifies to have taken the initial sentence in the sense of “nothing but”; A clarifies to have meant “the constitutive role”. Paraphrases are ubiquitous and yet very difficult to master, at least in philosophical contexts. Attempting to outline a simple doctrine often forces me to re-write for hours on end. But what precisely is it that makes paraphrasing so difficult? As the little example shows different paraphrases can already come with different interpretations, which in turn entail different evaluations.

In what follows, I hope to gesture at an answer that shows how paraphrases (tacitly) depend on interpretations and thus also determine evaluations of the paraphrased positions. What’s more: I hope to give a reason for dispelling the common myth that you can present a position without already being committed to an evaluation. I’ll close with some thoughts about how paraphrases also validate the paraphrased thoughts.

Problems with paraphrases in philosophy

While exercises of paraphrasing might be “basic tasks”, they are generally highly contentious and often even lie at the core of academic disputes. On the one hand, paraphrases can be historically problematic in that they introduce ideas unheard of at the time of the original expression; on the other hand, paraphrases can be systematically problematic in that they introduce unwanted (metaphysical) commitments. Historically speaking, the early modern use of the term “man” commonly has a wider scope (including women) than the twentieth-century use that renders it identical to the expression “male human (being)”. At the same time, this etymology is complicated by the fact that the pre-twentieth-century use of “man” is often tacitly referring to human males as the standard. Thus, depending on the context, the paraphrase of “man” with the term “human being” might count as anachronistic, although it is etymologically apt. Metaphysically speaking, we might wonder whether the expression “the present King of France” commits us to non-existent objects. Whichever side you take on such matters, a crucial function of the paraphrase lies in directing the attention or focus of the interlocutor. Thus, it clearly affects the philosophical approach to a given thought or content and also the direction a conversation about it might take. Hence, the worry arises whether different paraphases tacitly commit interlocutors to contrary interpretations. Arguably, such worries are rooted in a holistic understanding of sentence meaning. Assuming that the meaning of a given sentence is not atomistcally determined, but by a set of other sentences that the given sentence is related to, I will worry whether get the implications right. Going back to the initial example, speakers A and B construe the meaning of the initial sentence via different sets of related sentences or implications.

This kind of problem becomes clearly palpable in teaching contexts. A helpful example is Anselm of Canterbury’s Proslogion. Confronted with the phrase “something than which a greater cannot be thought” (aliquid quo nihil maius cogitari possit), many students begin by asking whether that renders God’s greatness subjective. Given the phrasing in Anselm, this question reveals a certain kind of modern understanding of the term “thinking”. Reading “… cannot be thought”, they render “thought” as an activity that they (or even an ideal thinker) can perform and thus as the individual mental act of a subject. This reading drastically changes when students are exposed to an understanding of “than which a greater cannot be thought” as “greatest possible”, such that it can be seen as a metaphysical modality rather than a subjective act. Again, it’s paraphrases, both tacit and explicit ones, that bring out different commitments. Thus, the seemingly simple task of “saying the same thing in your own words” requires a careful interpretation of the phrase that is to be paraphrased. As I see it, then, giving a paraphrase depends on a specific interpretation of the initial phrase. Where does this leave us?  

The interpretation-ladenness of paraphrases

If our paraphrases are guided by specific interpretations, then this means that there is no such thing as a neutral report or presentation of a position. If I present Anselm’s formula in my own words and mistakenly say, for instance, “God is the greatest being, according to Anselm” (as opposed to “God is the greatest possible being”), then I’ll be mistakenly implying that God is the actually greatest thing (and something greater could be thought). However, most problematic paraphrases are not owing to such obvious blunders. Rather, they can depend on quite nuanced understandings. Now, my point is not that there are problematic interpretations; my point is rather that the paraphrase we choose commits us to a limited set of possible interpretations (as opposed to a different set of possible interpretations owing to a different paraphrase) and that there are no paraphrases that come without implications (and thus specific interpretations contrary to others). If this is correct, then there is no innocent or neutral paraphrase of a given expression. This, in turn, allows us to rid ourselves of a persistent myth: the myth that one could paraphrase a position and only then decide what to make of it. This myth often translates into a common thesis or essay structure, suggesting that you can structure your work by first presenting a position neutrally and only then evaluating it. This is impossible because the paraphrase already commits you to a specific interpretation – whether you know it or not.

The validating nature of paraphrases

But why, you might ask, is this so? Why can’t I simply paraphrase a position neutrally, leaving it open for various possible commitments? I have to admit that I have attempted this for a long time. But it doesn’t seem to work. The reason is that paraphrasing is a strangely bi-directional activity. On the one hand, a paraphrase is a bit like an (indirect) quotation, trying to convey what someone (else) has said. (Of course, we also continuously paraphrase our own expressions.) On the other hand, a paraphrase is like an appropriation, trying to convey what you have understood. These two aspirations can come apart, of course, both in historical and current readings. But what is even more important is that the appropriation often carries with it a sort of validation. In trying to appropriate someone’s form of words, I validate what has been said – by embedding it into my own thoughts. In fact, we often present and paraphrase claims that we take to be commonly accepted without specific references to any particular author. People now constantly say that mariage is a speech act, without particular reference to Austin or a theory of performative utterances. This thought has become part of a fairly common way of thinking about the role of language. Its ubiquitous paraphrases have made it part of the public domain, as it were. This way the initial thought gets validated in various formulations.

To see this, it’s vital to realise just how ubiquitous paraphrases are. In fact, most of the things we say are paraphrases of others’ words. In fact, we learn to speak and practise our daily interactions by constantly saying, in slight variations and paraphrases, what we hear and read others say. And since it often doesn’t matter who precisely said what, the exact authorship of the paraphrased sentences fades – until we fully embrace the thoughts ourselves and think we are original.   

So paraphrasing is a continuous and indeed necessary activity we practise in our daily lives, often used to validate thoughts in new contexts. This feature of validation, I submit, carries over when we, as philosophers, present the position of someone else. To what precise extent remains to be seen. But it’s clear that the concomitant validation plays a crucial role in the way we learn and pick up thoughts both in daily interactions and as philosophers who appropriate thoughts of others into our understanding of the world and the claims of current or past interlocutors.

An afterthought:

If this is remotely correct, using ChatGPT or other tools for paraphrases (instead of learning and constantly practising the validation of thoughts ourselves) might have highly problematic consequences for our (linguistic) interactions with others and indeed ourselves.

I suppose I’ll never quite forget how clueless I was when reading Wittgenstein’s Tractatus for the first times. Why, to name one confusion of many, did he not bother to provide more motivation for the very first sentence: “The world is everything that is the case”? Was there some common or perhaps divine intuition that I just failed to have? For me at least, things changed drastically when I read somewhere that Wittgenstein was greatly inspired by gestalt psychology. According to the approach he might have had in mind, then, we start grasping stuff from a holistic totality and analyse our way into details rather than from atoms from which we piece together an initially fragmented world. I’m not saying this is the whole story. But for me it provided a possible motivation that helped me understand a sentence that seemed to have been written without justification. Crucial context had been provided by something that seemed utterly absent from the text itself. However, once I took gestalt psychology into account, many things began to add up. By contrast, when texts are sufficiently current or explicit about their inspirations, their motivations seem fairly obvious. So much so that you can predict an author’s response to a given question as soon as you know some of their basic claims. Given demands of consistency, you sort of know what Ruth Milllikan would respond to the question of “whether ChatGPT can think”, even if she has not done so explicitly (just for the record, I’d say she’d point to her considerations of Swampman). But when texts are sufficiently remote in time or cultural conventions, it’s vital to take into account their sources of inspiration. In fact, I think scholarly forms of reading largely consist in re-establishing contexts of this sort. In what follows, I want to motivate this approach to reading and provide some further distinctions along the way.   

Imagining the author’s desk. – One of the first things you learn as a student of medieval philosophy is that you have to start by reading much of the Aristotelian corpus. Even if you don’t follow this advice, you’ll soon find that Aristotle is all over the place. Explicit and indeed implicit references to his works are woven into the fabric of most medieval philosophical texts. When you look at critical editions of medieval works, you’ll learn another important thing: Many medieval texts are full of often unacknowledged quotations from other authors. When you read William of Ockham, you’re often faced with a jumble of chunks from Henry of Ghent, Duns Scotus, and others. This reading experience changed my attitude to texts generally: When I read a text, I (also) want to know which books were lying on the author’s desk when writing the text. The reason for searching such sources is not just a curiosity about the author’s inspirations or the phenomenon of intertextuality; often the material that inspired the author helps you understand the text at hand. Ockham, for instance, will presuppose that his readers also read Boethius or Aquinas. So he wouldn’t bother explaining an issue when a brief reference to a theory of an earlier author would do. While Ockham’s own brief reference can leave you clueless as a modern reader, looking at his sources might provide just what you need to understand where he is coming from. This means that you can figure out what Ockham was trying get at when looking at earlier stuff. Accordingly, the more material on Ockham’s desk you can identify the better you’ll understand his frames of reference and – perhaps – ways of thinking.     

Leaning from Jenny Ashworth. – For students of philosophy, (early) modern philosophy is often introduced as an era in which philosophers shook off the reverence and references to prior authorities. Accordingly, these texts (seem to) encourage a mode of reading as if they were written straightforwardly “for you”, i.e. without the need to recur to earlier, especially scholastic, sources. As I see it, such authors were basically just better at hiding their sources. Jennifer Ashworth’s work on post-medieval scholasticism, even in figures such as John Locke, debunk this myth of textual autonomy, pushing the contextualisation to an instructive extreme. While Locke seems to pretend, even at the time of writing his Essay, that he is completely out of touch with Aristotelian and scholastic sources, Ashworth and others have shown clearly that he was very much inspired or at least wrestling with this material (here is one of Ashworth’s papers on this issue). What helps, then, in understanding such authors is the diligent study of contemporary and earlier texts and trying to get a picture of the books in their libraries.

What to look for. – Studying an author’s sources or gathering them from scrupulous critical editions is a good starting point for getting at the ‘material basis’. But you’re not doing a plagiarism check. (Of course, you might do, and conclude that all authors were less original than you thought, but that would merely betray a lack of understanding intertextuality.) So in what way can you exploit such sources? As I see it, imagining the author’s desk can get from very concrete kinds of inspiration, i.e. the very words someone actually quotes, to fairly abstract modes of thinking: so you’re looking for quoted words, imitated styles, related kinds of arguments, common principles, terminology, leading concepts or models. One thing that made me apply this strategy of reading more explicitly as a proper method was the realisation that I use unacknowledged forms of inspiration much of the time myself. When I wrote my PhD dissertation on Ockham’s philosophy of mental language, for instance, I was greatly inspired by my studies of linguistics, especially text linguistics. Even though hardly any of that made it into my text, certain ways of linguistic reasoning continuously served as a backdrop for my reading and writing. So if I write like this myself, it’s not entirely outlandish to assume that other people were and are inspired in similar ways.

Figuring out how an author ‘thinks’. – Eventuallly, this approach to reading might get you into very elusive interpretive territory. Going from what other texts might have inspired an author’s writing, you might get a feel for more abstract kinds of inspiration. Does John Locke think like a mechanist or does his medical background have a bearing on his thought, such that he might be said to think like a biologist avant la lettre? While this kind of issue is very elusive indeed and very hard to argue for, you might try and find some evidence in the way an author construes or exploits examples, thought experiments or analogies. While elusive, certain styles of reasoning preclude certain forms of consideration and might provide insights into what enables discoveries or inventions (or what might have blocked them).

In other words, trying to make ‘the context’ of a text concrete by imagining sources of inspiration re-establishes the conceptual space in which you can see an author moving within the boundaries that provide both consistency and limitations.   

______

* Here is part one of the series on reading.