The Devil Isn’t Just in the Details: Reflections on Artificial Intelligence, Islam, and Higher Education
Many of us have had our first conscious experience with AI through interacting with ChatGPT, that marvellous wonder of the postmodern world which writes stories, poems – and, yes, student essays. This is not to say we haven’t had less conscious experiences with AI, which have swayed our financial and political decisions, as well as flagged mistakes in our spelling and grammar, only that ChatGPT has changed the game in many ways. Including in education.
ChatGPT has, hence, been of serious consternation to many educators, including at the university level, who unknowingly started awarding A’s to essays written by The GPT. While some more broad-minded educators have now considered how ChatGPT may be used as an educational tool – for instance, as a conversational partner – many students (and, indeed, professionals of many stripes) are not at that level yet and simply want it to their job for them. While this has been decried as an assault on academic fairness, and an attack on the learning process, it also raises the question: what is higher education for?
If universities are perceived as vocational school, and one’s degree or grade point average directly correlates to their future salary, then – yes – using ChatGPT to write high-scoring essays is deeply unfair, because that student will reap decades of a higher salary, courtesy of ChatGPT. But is that the point of education? Is the sole point of higher education that students receive quantitative marks on written assignments, which then correspond to digits in their bank accounts? The Reign of Quantity indeed. While many people today do attend university for precisely that reason, it is worth considering that, even up until the 1980s, higher education was not just seen as a job track. Rather, the point of being educated was to develop the human and humanity – to nurture the individual, to explore new ideas, and to elevate the human race. Famous academies, such as that of Plato, did not even award degrees, let alone grades or marks. In contrast, today, in some places, universities have become factories; run along corporate agendas alongside student loan agencies, they churn out graduates, sometimes with a student-to-teacher ratio of 1:400; and the continued requirement to have a degree in order to get this or that job keeps them in business. It is no wonder that students would turn to ChatGPT – but that, perhaps, gives us an opportunity to rethink what higher education is for, how it is run, and whether or not it still needs to be quantitized.
Here, it is now helpful to take a step back and look at the tradition of Islamic education. Historically, Islamic education – including the hawza system – has not centred on degrees or marks. While scholars awarded ijazas (licenses) in various matters, the idea that one completed one’s education – and had hence “graduated” – would have struck many historical Islamic scholars as ludicrous. Learning was, after all, from the cradle to the grave. Traditional Islamic education was democratic before democracy was in; effective teachers were, more or less, selected by their students, rather than on an institutional basis. (Exceptions apply.) Conversely, the rigors associated with study – such as travelling across the desert on foot or horseback on the proverbial “journey for knowledge” – weeded out the less dedicated students. Most importantly, learning was idealized as an act of worship, rather than a fast track to a job. Learning for any reason other than the sake of Allah – including fame, fortune, or just to win a debate – was looked down upon. (How far we have strayed in the era of social media!)
Today, Islamic education is more complicated. Many places do have systems involving degrees, units, and passing marks. Teaching Islamic Studies in the West, under the system of the contemporary university, offers its own challenges; one has to respect the tradition while adapting it to how things have done. One advantage, however, is that, compared to fields such as marketing, there are far fewer students in Islamic Studies who are just in it for the money. (In fact, I have never met any, although once I did walk into a classroom where a student had scrawled, “What kind of job do we get with this afterwards?”) For that reason, there is less incentive to cheat.
This is not to say that students of Islam do not employ The GPT. Many do, in varying ways, ranging from upgrading droll paragraphs, to coming up with ideas, to translating, to writing entire essays, to writing about ChatGPT. Of the above, I have found that “writing entire essays” is rarer than the others (something that cannot be said about the world of professional academic publishing). Rather, there are many reasons that student use ChatGPT.
One is that English is not a heritage language of the Muslim-majority world. As such, many students of Islam do not speak English as their first language, and may have learned it in adulthood. This particular cohort often has strong Islamic education (for instance, a hawza education) but has difficulty expressing themselves. It is therefore understandable that they may wish to use language tools. This, to me, does not bring up a significant ethical issue regarding authorship. However, it does bring up the question of language, environment, and thought. Languages are not merely mechanical tools, all otherwise the same. Rather, languages themselves have been shown to lead to different thoughts. What can be said and thought in one language is sometimes not said or thought in another. Even children speaking multiple languages have been shown to say different things in different languages – for instance, speaking only respectfully about their parents in Japanese, but disrespectfully in English. (Thanks, Nickelodeon.) While some people might assume that everything there is to be said about Islam has been said in Arabic, and so there is no need to go outside the Arabic-language conceptual zone, an enormous amount of new and interesting literature has been written by the 30% of Muslims who live outside the Muslim-majority world, much of which has grown out of grappling with the interaction between Islam and secular modernity. Furthermore – as any of the Muslim youth sent to “youth programs” will say – addressing the genuine needs of people in the West requires, literally, speaking the languages of the West – not only for communication, but to engage with the nuanced issues that people are facing. Therefore, using ChatGPT as a language tool is understandable, but circumvents a significant part of the pastoral process.
Another is that some students, in the earlier years, did not learn to write. This is a particular problem in the United Kingdom, where many Muslim students do receive a substandard education, particularly girls, and are not expected to excel. This is also compounded by the overall lack of reading books in our time, insofar as books model writing better than Instagram. Therefore, a machine-wonder such as ChatGPT helps with structuring one’s ideas into coherent paragraphs and an organized essay. Is this such a bad thing?
It’s not a bad thing, if it is done as part of a learning process (such as learning how to write). However, there is also something to be said about the writing process – how the brain organizes writing, and how that relates to not only understanding and rehashing what one knows, but the discovery of new ideas. If you don’t believe me, write a paragraph about ChatGPT. Then add 5 sentences to that paragraph. Likely, you will have written some new things that you didn’t even know that you knew.
And herein also lies the secret of writing. Writing is not only for communicating, or for passing exams. It serves that function, and that function is important for religious professionals, especially in Islam, since Islam is a textual religion, based on interpretation of the Word – just as the universe itself was created through the divine Word. One can know many things about Qur’anic exegesis or the narrations from the Prophet (S), but if one cannot convey them through speech or the written word, that knowledge will not help others.
Many Islamic scholars preach, verbally, and that is its own art. Writing is another.
While preaching impacts the people in the here and now (or at least it did prior to the YouTube generation), writing persists over time, and thus has a certain eternality about it. Even today, it retains more complexity.
So communication is essential for people whose bread-and-butter is religion, and oftentimes, that is through writing. However, that is not all to writing. Rather, the secret is in the process of writing. As the author – including the student author – writes, understandings develop and form on the page, just as any other form of art takes shape. They may have had no idea what they were going to write – as, indeed, I did when I started this – but sometimes it just flows. In that, sometimes there is an inspirational quality, a spiritual aspect to the mechanical craft of writing. Here, it is not my aim to judge whether or not a digital being such as ChatGPT may enjoy a similar spiritual or inspirational quality, only to point out that, when it comes to educating a person, the important thing is that it happens to a person. This is what makes them their future self, the future scholar. Writing also teaches us what we know and what we don’t know, and forces us to confront the latter. Using a machine to circumvent that process stops the educational process and reduces it, at best, to the acquisition of facts – which may help a student pass a multiple-choice exam, but will not develop them further.
ChatGPT also has some quirks when it comes to Islam; possibly, this is another reason for the rarity of full-length essays generated by it. In the infrequent case when I see full-length essays on Islam generated by ChatGPT, they frequently fail. This is because ChatGPT does not do Islam well. First, despite the fact that ChatGPT has an enormous database in many languages, including Arabic source texts, it discusses Islam in a shallow manner. Any student training to be a specialist in Islam who opens an essay with something along the lines of “Islam is a major world religion practiced by over 1 billion people” is failing to specialize. Second, it frequently makes mistakes. Some of these mistakes are factual; for instance, confusing Abu Ali ibn Sina with Abu Ali Iyad because they both share the same kunya. Arabic, apparently, is ambiguous. Second, it cannot contextualize. If it thinks that democracy is good in the 21st century, then it is happy to praise democracy in the 7th century, especially since plenty of Muslim apologists have contributed to its dataset by arguing that 7th-century Islam was actually democratic. Part of the job of the historian and the scholar is to try to walk in the sandals of bygone generations, even if – as one historian once told me – we can never truly do so.
Third, and most worrisome, it has biases. An expression in computer programming that says “garbage in, garbage out”. That is no matter how watertight your program is, if your input is faulty, your output will be garbage. Regarding Islam, ChatGPT has, unfortunately, been fed a lot of garbage. Not all of this garbage is by casual bloggers. Some of it is elite academic literature in the orientalist or neo-orientalist traditions, speaking about the clash of civilizations, the burden of the modern white man to civilize – i.e. secularize – the Muslim-majority world. Some of it also reflects intra-Muslim conflicts – for instance, writings by expats against various governments; the English-language bias will give a preference to writings in English – which represent certain social, political, and religious views – over those written in other languages in the Muslim-majority world.
There is also the question of whose voice is loudest in the Muslim world; for financial and political reasons, Salafi interpretations of Islam have dominated the internet, and many publications on Islam in the present era. As a result, ChatGPT does not do Shi’ism well. Interestingly, it handles abstract issues – theology and philosophy – better than fiqh, making basic errors when discussing Shi’i fiqh. Therefore, there is an obvious issue with unreliability – but, more worrisome to me, are the root causes of this unreliability. Whose voices are wittingly or unwittingly prioritized by these corporations running our new world (to the point that Google is now investing in nuclear reactors)? This of course is not limited to essay-writing; many of us have noticed in recent months the censorship of news about current events on many social media platforms, and bots rather than humans propagating someone’s decision about what is right and true.
This should not be taken to mean that I, personally, am against ChatGPT. I find it fascinating and am inclined to think that historians (should we humans persist that long) will deem it as one of the most significant advances in human history. The development of artificial intelligence raises all sorts of interesting questions, not only about the purpose of higher education. However, here, the point is that just as machine learning occurs through practice, so too does human learning. And that practice is not just rote; it is not just quantitized. A scholar is not built merely through acquiring factoids; a scholar is built through reflection. Ceding that to the machine could, ironically, improve its ability to write essays on Islam, but would short-circuit the intellectual and spiritual aspects of writing that are its real purpose in higher education.