11. Descriptions and Interpretations

We probably spend too much time describing. I mean: describing with little interpretation. If, for instance, it is aimless to both advocate for literary criticism and exalt the higher rational order of data mining, the pervasive tendency to turn scholarly articles and dissertations into summaries of events or doctrines is equally dreadful. In one case, we defend a hierarchy that, in advance, shrinks down our work; in the other, we exhibit competences that are easily acquired by generative AI. One could argue that no description is fully devoid of interpretation—that is, of the introduction of a specific angle, of an interaction between different registers and positions, of disputable semantic strategies. Granted, but, in history, in philosophy, in the study of the arts or of literature, one could try to do as if interpretation did not exist. This efficacious fiction delivers in the name of objectivity, or fact checking, or doxography a take on realities, artworks, and texts that is both impersonal and situated. The countersolution is not to plead for pure subjectivity (which is as absurd as sheer transgression), or for unrestrained imagination, or for relativism but to follow and displace the facticity of facts (as their etymology indicates, facts are made up) by both acknowledging their actuality and reclaiming their distortive production as they become disciplinary, linguistic, and noetic objects.

Descriptions for their own sake, if they are not interrelated with an interpretive hypothesis, form the same kind of disposable information that feeds the automated generators. A comprehensive summary of the tragedy from which I am quoting three lines in my commentary is pointless if I am not linking these verses to specific characters and elements of the plot. A qualification of the regime under which a composer lived serves no function except when I am noticing a political response in a symphony. The biography of the artist is of no importance as long as I do not involve it in my understanding of the painted work. The profusion of exact elements about the first day of the war and its archival record is purposeful to the condition of being required for the elucidation of the historical event. The scrupulous overview of the scholarly debate on a given topic is irrelevant unless I am using these minutiae to bring forward a different model or theory whose strength has to be appreciated by comparison with the extant. Giving too much weight and credit to weakly interpretive descriptions is a disposable vestige of the past: of our own years of training, when we were not yet able to fully grasp an argument or a series of facts, and of a time of sparser access to knowledge. In a context of limited literacy, repertoires of dates and events, such as the Ethiopian Kebra Negast or Sima Guang’s Zizhi tongjian, are achievements by themselves. In a situation where books do not circulate or could be easily lost, where thinkers only profess orally, doxographies are important. Still, when we have the opportunity to compare how Aristotle and Diogenes Laërtius respectively synthesize the ideas of an author, we notice the strong differences between an exposition preparing the enunciation of a theory (with the Stagirite) and one adding to the view of philosophy as a varied activity (with the Lives of Eminent Philosophers). Wilhelm Dilthey separated explanation from interpretation among epistemic gestures, and one often reads that he restricted the first to the natural sciences and the second to the sciences of the mind. This divide looks to me like an oversimplification of Dilthey’s views. Yet, and more importantly, I notice among them three immediate hurdles. One is the naming of explanation as what lies beside, or below, interpretation in knowing. Explaining a physical phenomenon is barely possible without a serious interpretation of the data—that is, implying a theoretical and formalized rendition that is more than “crushing numbers” or, in literary criticism, even a modest explication de texte is more than a paraphrase (the descriptive level I am insisting upon, the lowest part of the interpretive). Then, Dilthey is misguided when he puts as “the aim of the sciences of the mind” the faculty “to grasp the singular, the individual of the social-historical reality,” for the singular is not the individual, and it is not a figure couched in the social-historical.1 Finally, it is erroneous to suggest that the “unexplainability of the mental” (“Unerklärbarkeit des Geistigen”) is what gives way to humanistic interpretation.2 The latter is as much required by the explainable, as explications do not exhaust their object and they rarely qualify it.

The time of AI is characterized by its overwhelming access to information. It is overwhelming, and therefore obfuscating, which is also why in the first decades of the World Wide Web, the search engine became so central to the ecosystem of the internet (whereas, previously, an index or a table, as useful as they were, never acquired such a reigning function; whereas catalogs, becoming largely indispensable with the multiplication of print, had always to be complemented by the serendipity of book browsing). Scholars should not act as providers of well-polished building blocks to be added to the great wall of information. Living knowledge will be difficult to place in a smooth continuum, precisely because new thinking should modify what is already at hand. Before the rise of generative AI, Wikipedia was already indicative of the inner strife. The project of a free, democratically composed encyclopedia, changeable in real time, was undoubtedly appealing. The popularity and accessibility of the platform also made it more obvious than ever before that unrelated descriptions of the overall context are a waste of time, especially in the classroom, when, on many questions, a professor will be unable in a seminar to reach the alignment of facts one finds in the best Wikipedia entries. If we were not aware of the uselessness of the general briefing by way of a generic introduction, we now have no excuse to ignore it. At the same time, that the online encyclopedia would be built on the “no original research” motto and would extol the “neutral point-of-view,” or that editing bots would play a very large role (especially in some languages), should clearly indicate how foreign this endeavor could be to the idea of the humanities I am supporting. The ban on new knowledge coincides, in Wikipedia, with modes of writing such as pervasive paraphrase and plagiarism, both operations that are also heavily involved in computerized generations. Moreover, the supposedly democratic community of Wikipedia contributors is, in practice, a highly hierarchized group whose order does not reflect gradients of epistemic competences but the frequency of interventions in the platform and the perceived acceptability in the mastering of editing procedures. The triumph of Wikipedia is becoming overly dubious when this online repository begins to be understood not only as a possible point of entry but as a necessary one and, worse, as a point of arrival—the validation of research being outsourced to its, preferably noncontroversial, insertion in an article. The events called “edit-a-thons,” routinely organized by American centers of knowledge (libraries, universities, museums), not only disseminate a poorly coined neologism and invite to conformism (many of these collective sessions being theme- or identity-based). They also display a conception of research that is energetically undermining the humanistic perspective. How remote we are from L’encyclopédie organized by Denis Diderot and Jean Le Rond d’Alembert . . . This is not solely because the content of Wikipedia is free that our best LLMs have digested its articles; this corpus was perfectly adequate to the functions and goals of the models. The generators not only recycle the internet encyclopedia, with the appropriate level of synonymy and collage, they also magnify its practical theory of knowledge. On our end, we gain by remembering that description is an act of interpretation and that it should therefore be performed as such. Let us leave the rest to the bots.