10. The Ethical Fallacy

Humanists are no policymakers, no watchers standing in front of the gates of morality. In his fragments on philology, a century and a half ago, Friedrich Nietzsche denounces the confusion between “the human” (“das Menschliche”) and “the humane” (“das Humane”) within the pursuit of the humanities.1 According to Nietzsche, the theological roots of Renaissance humanism were enduring, and, despite all best efforts, the study of the classics was still marred by the desire to turn scholars into “priests.”2 The justification for the humanities as a civilizing tool, taking away the brutality of humankind, is as old as this vocabulary. We already found it in imperial Rome where humanitas linked literary education with refined morality. Before Cicero, going into the opposite direction but on the same line, Plato was denouncing the tragedies’ pernicious “capacity for harm,” therefore removing poetry from the formation of the philosopher—while still upholding the value of reflection in the ideal of honesty and justice.3 Later, in the second century, Aulus Gellius separated the ethical from the scholarly in proposing his transcultural definition of humanitas as the equivalent of “what the Greeks call paideia [education]” or “organized learning and knowledge in the liberal arts [the disciplines],” adding that it does not mean “philanthrōpia,” understood as “a certain propensity for benevolence toward all humans indiscriminately.”4 In a seemingly unrelated context, an important debate in 1990s mainland China concerned the “spirit of the humanities” (rénwén jīngshén), asking if the latter could be divorced from the (Confucian) ethics of well-ordered “human relations” (rénlún).5 In the face of this theoretical oscillation between the declaration of a moralizing effect through the humanities and the statement of their ethical indifference once taken en bloc, it could look like the experience of the world and decolonization wars in the twentieth century definitely weakened the thesis of a civilizing process, as sharply explained by Aimé Césaire in his Discourse on Colonialism. Césaire remarks that “it would be worth . . . revealing to the very distinguished, very humanistic, very Christian bourgeois from the 20th century that . . . Hitler inhabits him” and that “the Western humanist,” such as French historian Ernest Renan, may therefore be incapable of “laying the ground for individual morality.”6

Nonetheless, today as before, from East to West and South to North, the discussion of the moralizing aspect of the study of letters is still wide open. In North America, some researchers try to show that reading novels gives you more empathy, while others repeatedly understand artworks or poems as proofs for the existence of a social structure they disapprove of. In the latter case, the humane is a secondary derivation: by showing that Ovid is promoting rape culture, we might establish a harm and potentially push for a revision of the canon; by canceling Alison Bechdel, we could promote religious values; and, in one way or another, we would have created more moral individuals. Being confronted to what one is not, journeying through one’s and others’ otherness, is the most liminal experience of the arts, without which nothing can ever happen.7 I may conceive of new ideas, transport myself into different minds I am reconstituting, and be altered by the process. Now, how this will affect me morally is a separate issue—a better understanding of something different could also convince me to more thoroughly reject it rather than “embrace diversity.” Morality being built on personal positions through life, it is influenced by what informs the discursive much more profoundly or straightforwardly than by the study of quantum physics or cell biology. But this is a chance and not an absolute inference. The letters have an ethical dimension, which I do not seek to deny: when I say that they contribute to making life livable or to expanding thinking, I am designating a sort of reflexive cultivation of mores and behavior that stands in relation to others. This, however, is quite undetermined and not the same as putting virtue (in any guise) on the list of learning outcomes. The justification of humanistic training is often split between the gratuitous and the useful. In the sixteenth century, the professor Pierre de La Ramée (Petrus Ramus) was notably scandalous for his alliance of eloquence and philosophy (with Cicero, and not Aristotle, as a model), oriented by “utilitas” in the training of competent and respectable individuals.8 The oscillation between goodness and moral indifference responds to the other wavering between utility and gratuitousness. In this dual sway, moral edification has the added cachet of offering the image of nonutilitarian usefulness.

Then, the role of the humanities vis-à-vis AI cannot be the fighting of biases. Companies, governments, training sets, and programmers are full of prejudice. As citizens, we may wish to object to the ensuing incorporation of such biases in digital systems open for public use. As scholars, we may contribute to showing their theoretical, symbolic, social, and discursive armature. This, however, is a political and epistemic engagement that should not be gauged according to a morality scale, from good to bad. The worth of the liberal arts remains, inasmuch as personal and collective emancipation proceeds from a critical assessment of what is, and especially from the undoing of the dogmatic empire of the religious, which has often been concerned with granting as well as limiting intellectual freedom, from the Christian universitas to the contemporary appeals to the so-called Islamization of knowledge, or from the propagation of nationalism to the doctrine of the social use of research.9 Yet, this is no morality, and the liberatory dimension of knowing is not restricted to, or even uniquely lodged in, the discursive disciplines. Furthermore, we should make no promise on the subjective edification that would follow the rehashed enunciation of power relation. There are, and will be, more jobs for “humanists” who accept an ancillary position in the digital economy, their training being sold as a token of humaneness and the requisite for becoming sensitivity readers at the service of generative AI. The strong possibility of a future proliferation of specialized categorical chatboxes (the fundamentalist Christian, the LGBTQIA+, the Latinx, the white American . . .) could allow for the employment of self-proclaimed experts having received the moralistic training some claim textual studies deliver. At the general level, a varnish of “humane-ities” would be taken as a predisposition for organizing the content policies of Big Tech. Here again, if this is the future, then, this is the end, my friend.

The contemporary academic field of ethics is worth a special note. The Aristotelian legacy is complex and not without problems, especially through the doctrine of the mean (involving a computational approach) and the apparent politicization of the good life. There remains that the Nicomachean Ethics—in its description of multiple forms of existence, in its insistent focus on happiness, in the idea that this branch of philosophy should shape the life of its thinkers, and in the recognition that it deals with “the contours” (tupos) of practice in a way that is “not sharply defined” (ouk akribōs)—is situating an intellectual endeavor that is irreconcilable with the professionalized ethics we are being force-fed.10 Ethics has now become a synonym for the mechanics of social casuistry. This even gives way to so-called decision trees, particularly in the medical sciences, where what is supposedly ethical (such as treating a patient or not) is the output of a logical procedure depending on set criteria (such as age, concurrent pathologies, and the like). Even in the absence of robotic caretakers, we see the triumph of AI through triage, through automated deductive procedures, and through the overall encoding that are the main reality of medicine today, especially in the United States. The ethics therein replaces forms of existence with organic entities, happiness with social usage of resources, personal morality with decisional tools, and the fluctuating contours of practice with formalized inferences based on said objective qualities.

But what about advocating ethics for computer scientists? Besides the reduction of a vast scholarly effort to one specific field (and, quite often, the further reduction of ethics to utilitarianism), the difficulty lies in an inability to propose a stance that would not preliminarily admit its epistemic subordination. Let us take a brief look at one recent attempt, at Harvard University, trademarked “Embedded ethiCS,” in the “computer science curriculum.”11 The keywords of the promoters are “normative questions,” taken as a synonym for ethical debate, listing, as examples, “In data visualization, what should be done to avoid reinforcing harmful stereotypes?” (a question I doubt philosophy is able to actually answer in a better way than sociology or justice) and “Is it permissible to develop and deploy automated systems that will replace human workers?” (this one looks a bit moot to me). The avowed objective is to be “providing philosophical tools,” not meaning, concepts, texts, or thinking experiments, and to put only a “small” “burden” on existent computer science courses. Nothing life-shattering, for sure. The entire conception of ethics as some kind of humane supplement sustained by procedures and added to the cracks of the engineering curriculum in view of business application is everything but a strategy to show relevance. The built-in obedience of the modest ethics consultant as well as the grandiloquent fantasy of actionable censorship and moral retribution are the two sides of the same defeating, and fundamentally unethical, approach. Once the humanities are chained to normative decision-making or to behavioral-theological conditioning, their sense dissolves. With or without AI, societies probably need guidelines, policies, rules, and laws, but it is hard to see on what theoretical and practical motives the discursive disciplines would be uniquely prepared for this mission—especially if the end product is the, necessarily political, selection of a norm, the writing of an algorithm, or the disposable addition of a moralistic refrain, be it in this standardized first person that the LLMs so easily simulate.