Large Language Models (LLMs) have shocked academia in general and the humanities in particular. If we rely too much upon machines to perform demanding tasks, such as generating essays, summarizing complex documents, and translating sources in unfamiliar languages, we can produce quite a bit but learn little or nothing. I do not see any simple approach that will suit every academic discipline and, far from having any definitive answers, I am still framing questions. That said, I can offer some preliminary thoughts on what I have observed, both as a student of the human record and as someone who has labored for more than 40 years to use digital technologies to transform the human record’s role in the intellectual life of society.
While I am a professor of classical studies, I describe myself as a philologist, and in defining that term, I paraphrase a Latin definition from an 1822 birthday oration for the King of Prussia by August Boeckh: philology is the use of the human record to advance “the lived understanding (cognitio) of the past (antiquitatis) in its entirety (universae).” I first heard this definition when I was a first-year student a half-century ago, and I have found it extraordinarily useful ever since. First, this definition of philology is based on goals rather than methods. If methods such as digital technologies and quantitative analyses advance our understanding of the past, then those methods belong to philology. It has thus seemed natural to me to devote myself to exploring how the study of the past can flourish in an increasingly digital age.
With LLMs, cognitio, the first term in this definition, rises to prominence. If we define philology as lived understanding, then no information technology—not writing, not books, not databases, and not automated systems—has any value except insofar as it enhances the intellectual lives of real human beings. Or, to paraphrase Protagoras from the 5th century BCE, we measure the value of things by their impact upon human beings.
Our first task in humanities must focus on developing our abilities to read and to write. We can, if we are careful, learn more effectively with LLMs if we exploit them to answer questions, to pose questions to us, and to challenge us to think. In fact, the greatest challenge my students face when learning a historical language with no native speakers is that they have very few opportunities for feedback: each student gets limited immediate feedback in class and must wait days (usually) to get feedback on written assignments. If LLMs can provide immediate and inexhaustible feedback (and they certainly can with training), the next generation of students will be able to internalize languages such as Greek and Latin far more effectively.
If we are, however, not satisfied with simply acquiring core skills and are committed to a full understanding of a field such as classical studies, LLMs can ameliorate multiple challenges (and I might say crises).
First, students who focus on Greco-Roman culture as part of a major in Classics (or, as at Tufts, Ancient World Studies) do not, for the most part, have the time to master Greek or Latin and must therefore depend upon translations without any ability to engage with the source texts in the original language. LLMs now not only can provide translations of Greek and Latin but can, more significantly, provide word-by-word grammatical and cultural explanations for the source texts, allowing readers to engage directly with source texts in unfamiliar languages. If we combine LLMs with methods from corpus linguistics, we can enable readers to explore cultural concepts and literary genres at scale.
Second, if we are to use a term such as “classical” in 21st century America, we obviously need to embrace languages and cultures beyond Greece and Rome and support, insofar as possible, languages such as Sumerian; Akkadian; the various forms of Ancient Egyptian; Sanskrit; classical forms of Chinese, Persian, Arabic, and Japanese; and the K’iche’ Maya of the Popol Vuh. India alone now has eleven classical languages. If we cannot even manage the languages of European cultural heritage, how are we to engage with the world as a whole? With LLMs, we can begin to engage with those languages that we simply would not otherwise be able to master and, in so doing, begin to realize the goal of studying world literature in a way that was simply not possible.
Third, even if we were only working with sources in English, we still cannot read all the scholarship that is of potential relevance to us. Every professional working in a field of any scale depends and always has depended upon a network of reviews that summarize the many documents that they will never have time to read, much less digest. Of course, if we never read anything but summaries, we will never gain a detailed knowledge of any document and accompanying argument, but we will always depend upon indirect sources for our knowledge of scholarship in any developed field. LLMs give us a fundamentally new tool and a chance to identify for closer inspection ideas of potential importance that are now buried in print and even digital collections.
In short, if we use tools to avoid learning how to read, to think, and to write, we will learn less and our minds will be duller. But if we are committed to advancing human understanding in a field such as classical studies, LLMs allow us to mitigate cognitive constraints that limit our intellectual range.
Artwork by Cindy Qi