MESA Banner
Middle Eastern Studies in the Age of GPT

Panel XIII-16, 2023 Annual Meeting

On Sunday, November 5 at 1:30 pm

Panel Description
This Panel examines the promise and perils of GPT (Generative Pre-Trained Transformer) for Middle Eastern Studies from various perspectives. From a technological standpoint, there is considerable disparity between English and Arabic NLP (Natural Language Processing), and ChatGPT will further exacerbate this situation. The more advanced state of English NLP tends to impose modeling rules on other languages, effectively eroding their distinctive characteristics. From a historical vantage point, Middle Eastern Studies in North America have been shaped by a specific form of prose writing, standardized during the establishment of the earliest Western European universities. This particular form emphasizes objectivity, transparency, and translatability. ChatGPT's dataset, response structures, and Middle Eastern scholarship largely adhere to this scientific prose style in English. Consequently, concerns arise regarding translatability in the era of ChatGPT, as it tends to homogenize language ontologies and conflate distinct historical and cultural trajectories, thus imposing a standardized mode of thinking rooted in an overtly English-speaking episteme. This phenomenon further reinforces the dominant discourse structure of scientificity, which marginalizes alternative forms of knowledge and suppresses diverse modes of speaking and writing that have played a vital role in the richness and significance of Middle Eastern Studies in North America.
  • ChatGPT, in its general approach, follows the discourse structure of scientificity, balancing arguments from a presumably 'objective' standpoint and utilizing empirical explanations to address questions. In contrast, 'subjugated knowledges' that have been marginalized or dismissed due to their perceived lack of scientific validity will be thoroughly challenged in the emerging epistemic shift. ChatGPT marks the beginning of generalized, ubiquitous AI that builds on the foundation of scientificity. As a text-centric system, it leaves no space for non-discursive approaches— such as adab and taṣawwuf in the Islamic tradition— which are fundamental components of traditional sciences. ChatGPT's subject/object-based response structure significantly differs from the nuanced and experiential knowledge embedded in the qasida (Arabic ode), a form closely associated with lived experiences, attuned knowledge forms, environmental awareness, and ritualistic commemorations. We are now grappling with an epistemic shift of a different order, pushing cybernetics and semiotics—here, I am using semiotics as a composite function of surface, screen, and scientificity— as the dominant forces shaping human knowledge and material dimensions. This transformation has the potential to undermine non-scientific modes of knowing, deeply rooted in cultural, genealogical, and experiential domains, offering unique insights and understandings that may not fit neatly within the framework of scientific objectivity.
  • The disparity between Arabic Natural Language Processing (NLP) and English NLP is poised to exert a substantial influence on the emerging knowledge structures shaped by ChatGPT. As the dominant medium, culture, and form, English is likely to impose modeling rules on Arabic NLP. This raises concerns about the preservation and vitality of languages such as Arabic, Persian, and Urdu in the face of the epistemic configuration unleashed by ChatGPT. The paradigm shift in education triggered by ChatGPT further accentuates these concerns. Students increasingly rely on the system for their studies, including producing complete papers. In Middle Eastern studies and other humanities, this reliance often leads to the creation of artificial scholarship based on English translations of non-English source materials and topics. Consequently, the substance and authenticity of scholarship are compromised, potentially shaping the future foundations of knowledge in the Arabic-speaking world around an inherently different, English-based system. The implications of this trend are significant. It raises questions about preserving linguistic diversity, cultural authenticity, and the equitable representation of knowledge systems, particularly in natural language-based systems. How can we ensure that the epistemic configurations facilitated by ChatGPT do not overshadow or diminish the inherent value of languages like Arabic, Persian, and Urdu in constructing knowledge and preserving cultural heritage? How do we bridge the gap between non-English NLP systems and English NLP? These are some of the concerns that I will explore in my presentation.
  • Scholars of Middle Eastern studies are bound to encounter the perils of translation in various ways. The majority of people living in the Middle East do not communicate in English, and many texts pertaining to the region are not originally written in English either. Along with Arabic, a language that has remained dominant there, texts have been composed in Turkish, Persian, Syriac, Aramaic, Hebrew, and Greek. These languages belong to different groups with distinct histories and sensibilities that neither overlap with English nor accede to the sovereignty of a particular form of prose writing standardized with the formation of universities in twelfth-century Europe. As people in North America and other regions primarily access information about the Middle East through English, scholars face the responsibility of translating their understanding into English while adhering to a standardized form of scientific prose. This presents both the promise and perils of translation—highlighting the necessity and impossibility of capturing the essence of Middle Eastern languages and their varied forms, be it poetic or otherwise, in English. Given that North American scholarship on the Middle East significantly contributes to ChatGPT's dataset, the rise of this technology further amplifies the significance of this scholarship. This paper aims to delve into the challenges associated with translating Middle Eastern source texts and their unique linguistic forms into English. If ChatGPT is homogenizing language ontologies and standardizing our thought processes primarily based on an English-speaking episteme. In that case, it begs the question: to what extent is this outcome influenced by the translation efforts led by English-speaking scholars of the Middle East?