Who Is the AI Humanist?

The question of who the studia humanitatis was meant to produce has a precise answer in the sources, and it is not “the well-read.” Pier Paolo Vergerio’s De ingenuis moribus et liberalibus adulescentiae studiis (c. 1402), among the founding statements of humanist pedagogy, frames the educational aim not as the accumulation of learning but as the formation of a person capable of exercising judgment and contributing to the life of the city. Leonardo Bruni’s letter De studiis et litteris (c. 1424) makes the same claim by a different route, treating grammar, rhetoric, history, poetry, and moral philosophy — the content Kristeller (1961) identifies as the substance of the studia humanitatis — not as separate subjects but as the integrated formation of a single capacity: the ability to understand, to argue, and to act.

That formation had, from its origin, a civic cast rather than a purely contemplative one. Baron’s (1955) account of Florentine “civic humanism” locates the movement’s most consequential early exponents — Salutati, Bruni himself — not in the cloister but in the chancery, as scholars who held public office and understood their learning as an instrument of the republic’s political life, not an alternative to it. The person the studia humanitatis aimed to produce was, by design, a scholar-citizen: formed by private study, answerable to public life.

We take up the same question under different conditions. What should the person who has undergone a modern studia humanitatis — as set out in the Institute’s companion framework of that name — actually be able to do, and for whom is this an achievable ideal? The answer requires attending to a problem the Renaissance humanists did not face in its current form: the growing share of what a person comes to believe that arrives not from another person’s testimony, but from a machine’s.

The inherited problem: dependence is not new

No epistemology of the AI Humanist can proceed as though dependence on others’ knowledge is itself the novelty. Hardwig’s (1985) “Epistemic Dependence” established, in terms that remain the reference point for this literature, that no individual possesses the time or capacity to independently verify more than a fraction of what they justifiably believe — that a physicist’s confidence in a claim outside her subfield rests, like anyone else’s, on trust in colleagues she has not personally checked. Epistemic dependence on testimony is not a corruption of rational belief; Hardwig argued it is a condition of it, for experts as much as for laypeople.

The question this raises — on whose testimony, and why — has its own literature. Goldman (2001) proposed practical criteria a layperson can use to identify trustworthy experts without possessing the expertise to evaluate their claims directly: track record, credentials assessed by relevant peer communities, and the disclosed interests and biases that might distort an expert’s testimony. These criteria work, when they work, because they are anchored in structures external to the testimony itself — a reputation accumulated over a career, a professional community with its own standards, consequences for being wrong that the expert bears personally.

Why the inherited answer does not transfer

None of Goldman’s anchors is available in the same form when the testimony comes from a machine. A model has no career and accrues no personal reputational cost for error. Its “peer community,” if the term applies at all, is a benchmark leaderboard and a set of internal evaluations largely invisible to the person receiving its output. Its disclosed interests, to the extent they exist, are the commercial and design choices of the organization that built it — several steps removed from the specific claim being tested, and not addressed to the individual relying on it.

This is not a claim that machine testimony is therefore untrustworthy. It is a claim that the ordinary machinery by which a person calibrates trust in human expert testimony does not carry over intact, and that a capacity formed to navigate human testimonial dependence is not, without further formation, adequate to navigate this one. The risk this creates has been named precisely, if not for this exact case. Nguyen (2020) distinguishes an epistemic bubble, in which relevant information is simply missing, from an echo chamber, in which a person has been given active reasons to discount outside sources — and argues the second is far harder to escape because it is structurally self-reinforcing. Fluent, confident, endlessly available machine testimony creates a related risk by a third route: not bubble, not chamber, but the erosion of the impulse to question at all. Nguyen (2022) calls the resulting posture “trust as an unquestioning attitude” — trust that has stopped operating as a considered judgment and become a default.

The distinguishing capacity

The AI Humanist, as this Institute uses the term, is the person formed to resist that default — not by refusing machine testimony, which would forfeit its genuine value, but by retaining the specific capacities that ordinary testimonial trust, transplanted uncritically from the human case, does not supply on its own: the ability to ask what would count as evidence for a claim, to notice when a fluent answer has substituted for a supported one, and to know when a question exceeds what any current system’s testimony can settle.

This is a narrower claim than “AI literacy,” and the distinction is the substance of the definition, not a footnote to it. Fluency with a tool — knowing how to prompt it, how to chain its outputs, how to use it quickly — is a technical skill, genuinely useful, and orthogonal to the capacity in question. A highly fluent user can still fully surrender judgment to the tool’s output; a comparatively unpracticed one can retain it. The Institute’s guiding formulation for its Human Agency pillar — delegate execution while preserving human authorship — names the distinguishing mark precisely: not whether a person uses machine testimony, but whether the person using it remains the author of the conclusion, in the sense of being able to say what grounds it and to revise it when the grounds change.

This essay, and this issue of the publication, appear without individual bylines — a convention explained elsewhere in this edition and worth reconciling here directly, since retained authorship is the very capacity this section defines. The two senses of authorship are not the same one. Public attribution names who wrote a sentence. Retained authorship, in the sense this essay uses the term, names whether someone — named or not — can say what grounds a conclusion and revise it when the grounds change. An unsigned institutional statement satisfies this second sense exactly to the degree that a named human editor stands behind it, can account for its claims, and is bound to correct them when they fail; it fails this sense only if no one, in fact, could do so. The convention adopted here rests on the first condition holding, not on its absence being a virtue. A reader entitled to ask “who can say what grounds this, and would revise it if shown wrong” should always have an answer, whether or not that answer appears on the page as a name.

A note on the humanist premise

This essay, and the Institute it serves, take for granted that a specific capacity — the ability to form purposes, weigh evidence, revise a conclusion, and bear the consequences of a choice — is worth cultivating and protecting. That is a narrower claim than it might first appear, and worth distinguishing carefully from two things it is not.

It is not the claim, often associated with transhumanism, that human beings should merge with or be superseded by more capable machines in order to keep pace with them. Wolfe (2010) draws this distinction sharply: much of what calls itself posthumanist thought, in his account, is transhumanist rather than posthumanist, because it remains committed to the very ideals of rational agency and self-perfection that Renaissance humanism first articulated, merely relocating them into engineered bodies. Wolfe traces this genealogy, following Bostrom’s own account of transhumanism’s intellectual history, back through Pico della Mirandola specifically — the same Oration this Institute draws its founding proposition from is also claimed as an ancestor by the tradition arguing humans should engineer their way past their present limits. Pico’s argument that human beings have no fixed nature, only the capacity to shape what they become, does not by itself settle which shaping counts as fidelity to that capacity and which counts as its abandonment. This essay takes a position on that question — that cultivating the capacity itself is the more faithful reading than replacing it — without claiming Pico’s text forces that reading on any careful interpreter.

What distinguishes the two readings is not, in the end, a matter of textual interpretation alone. It is a matter of what Pico did with the argument. The Oration was not written as a contribution to a literature; it was written to open a disputation, in public, in which an unlicensed twenty-three-year-old proposed to defend his theses against the established theological authorities of his age, before any of them had granted him standing to do so — a posture of contesting entrenched authority on the merits, at real risk, rather than a claim about what human nature ultimately permits. That posture, not only the argument it introduced, is the inheritance this Institute claims directly: the Disputation Protocol that governs this publication, open to challenge from anyone without credential, is a working instance of it. Transhumanism’s own development has, on the whole, proceeded through and alongside existing centers of technological and economic power rather than against them — a generalization with exceptions, but a fair one — extending what power can already do more often than contesting it. Two things, not one, are named by this Institute: a rebirth of the capacity for human thought under conditions artificial intelligence has changed, and the contest against entrenched power that a rebirth of that kind has always required to become more than a private conviction.

Nor is this the claim that human interests should override every other consideration as a matter of species privilege. Academic posthumanism, in the tradition running from Haraway through Wolfe, is substantially a critique of exactly that kind of privilege. What this essay defends is narrower: not that humans matter more, but that the capacity for reflective, revisable, consequence-bearing choice is worth cultivating wherever it is found, and that as far as can currently be established, that capacity resides most fully in human beings. This essay does not take up, and does not need to resolve, what humans owe to animals, ecosystems, or the machines increasingly built among them; recognizing that human welfare is entangled with all three is compatible with everything argued here and is a separate claim, better made by the literatures built to make it.

That formulation has its own vulnerability, and it deserves to be stated rather than left for a challenger to find. Grounding the case for cultivating this capacity in the ability to participate in a debate about it risks quietly excluding everyone who cannot argue their own case — infants, people with severe cognitive disabilities, and, on most accounts, animals — from the moral consideration such an argument is meant to protect. Nussbaum (2006) built the capabilities approach substantially to correct this problem in the philosophical tradition this essay draws on, grounding moral standing in the capacity for flourishing rather than the capacity for rational self-defense, precisely so that beings who cannot debate their own standing are not thereby excluded from it. This essay’s commitment to cultivating human deliberative capacity does not require denying that other beings warrant moral consideration on different grounds; it claims only that the particular capacity this Institute studies — deliberation under conditions of increasingly capable machines — is a human one, and that studying and protecting it does not, in itself, require deciding the broader question Nussbaum’s tradition poses.

The citizen half of the formation

Consistent with the civic cast of the tradition it draws on, this formation is not complete as a private discipline. Baron’s civic humanists did not stop at cultivating their own judgment; they brought it into chancery and council. The contemporary parallel is participation in the institutions through which AI-mediated claims now travel — asking, in a professional or civic setting, who authorized this system’s use, on what evidence its output rests, and who remains accountable if it is wrong. These are the questions this Institute’s Institutional Integrity and Governed Intelligence pillars pose at the organizational level; the scholar-citizen is the person who knows to ask them at the individual one, in the meetings and decisions where such questions are otherwise least likely to be raised.

Four theses

This essay’s claims are subject to challenge under the terms of the Institute’s Disputation Protocol. Editorial disclosure, including the essay’s AI review process, appears in the closing section of this edition.

References

Baron, H. (1955). The Crisis of the Early Italian Renaissance: Civic Humanism and Republican Liberty in an Age of Classicism and Tyranny. Princeton University Press.

Bruni, L. (c. 1424). De studiis et litteris [Letter to Battista Malatesta].

Goldman, A. I. (2001). Experts: Which ones should you trust? Philosophy and Phenomenological Research, 63(1), 85–110.

Haraway, D. (1985). A cyborg manifesto: Science, technology, and socialist-feminism in the late twentieth century. Socialist Review, 80, 65–108.

Hardwig, J. (1985). Epistemic dependence. The Journal of Philosophy, 82(7), 335–349.

Kristeller, P. O. (1961). Renaissance Thought: The Classic, Scholastic, and Humanist Strains. Harper & Row.

Nguyen, C. T. (2020). Echo chambers and epistemic bubbles. Episteme, 17(2), 141–161.

Nguyen, C. T. (2022). Trust as an unquestioning attitude. Oxford Studies in Epistemology, 7, 214–244.

Nussbaum, M. C. (2006). Frontiers of Justice: Disability, Nationality, Species Membership. Belknap Press of Harvard University Press.

Vergerio, P. P. (c. 1402). De ingenuis moribus et liberalibus adulescentiae studiis.

Wolfe, C. (2010). What Is Posthumanism? University of Minnesota Press.