The Grammar of Assent
The Grammar of Assent was Newman's 1870 attempt to describe, with philosophical rigor, how real people actually reach certitude in matters that resist formal demonstration. Against the empiricist tradition that held conviction should be proportioned to evidence, Newman argued that concrete certitude is reached through the convergence of independent probabilities — none sufficient alone, together compelling — assessed by the trained judgment of the individual reasoner. The process is rational but not formally demonstrative; personal but not arbitrary; rigorous without being reducible to rules. In the AI age, the grammar of assent stands in sharp contrast to what can be called the grammar of the prompt — the emerging discipline of framing questions to a machine in ways that produce useful outputs. The two grammars are structurally different, and the confusion between them is, in Newman's terms, the characteristic intellectual pathology of the present moment.
In the AI Story
The grammar of assent has three essential features. First, it requires encounter — direct engagement with the primary material, the source, the reality the proposition represents. The lawyer must read the cases. The physician must examine the patient. The encounter may be tedious, frustrating, apparently wasteful in its inefficiency — but the encounter is where real assent is formed. Second, it requires personal judgment at every stage — the weighing of evidence, the recognition of what matters, the exercise of the illative sense within the specific domain of inquiry. Third, it requires conscience — the interior faculty that holds the reasoner accountable not merely for the logical consistency of the conclusion but for its truth.
The grammar of the prompt requires none of these things. It requires real skill in framing questions, in specifying context, in iterating on output — but it does not require encounter with primary material, does not require personal judgment about the truth of the output, does not require the exercise of conscience. The user frames the question. The machine produces the answer. The user evaluates the answer by its external qualities — coherence, plausibility, usefulness — and deploys it.
Chase Mitchell, writing in the Christian Scholar's Review in 2025, captured the danger with a striking inversion. Newman's Grammar of Assent describes the ascent of the mind from probability to certitude. Mitchell argued that AI propagates what he called a 'grammar of descent' — a downward movement in which the ready availability of sophisticated outputs erodes the motivation, the discipline, and ultimately the capacity for the ascent Newman described. The descent is incremental, comfortable, and almost entirely invisible to the person undergoing it.
The Berkeley study of AI-augmented knowledge work provides empirical texture. Workers produce more outputs across more domains. The grammar of the prompt is being mastered at scale. Whether the grammar of assent is being correspondingly cultivated — whether practitioners are doing the slower, costlier work of personally appropriating the material their tools help them produce — is the question the study does not directly answer, but that its findings make newly urgent.
Origin
Newman worked on the Grammar of Assent for twenty years before publication, drafting and abandoning versions until he found the form that satisfied him. The book was his most philosophically ambitious work, aiming to do for concrete reasoning what formal logic had done for abstract demonstration. The title was chosen carefully: 'grammar' signaled that assent has structure — describable principles governing legitimate conviction — even where formal proof is unavailable.
The book was received with puzzlement by many of Newman's contemporaries and has had a long, slow influence since. Twentieth-century philosophers of religion rediscovered it; phenomenologists found resonances with Husserl; contemporary epistemologists working on testimony and credence have returned to it repeatedly. The AI age has given it a new audience, because the questions it addresses — how does a concrete mind responsibly reach conviction? — are the questions the AI transformation makes inescapable.
Key Ideas
Certitude comes through convergence. No single argument compels; the accumulated weight of independent probabilities, felt as convergent, produces rational conviction.
The convergence is personal, not mechanical. The illative sense of the individual reasoner assesses the convergence; no algorithm can replace the assessment.
The grammar of the prompt is categorically different. It produces notional outputs without requiring the personal appropriation that real assent demands.
Confusion of the two grammars is the characteristic AI-age pathology. Practitioners mistake satisfying outputs for genuine understanding.
The solution is not to reject the prompt but to preserve the assent. Prompt engineering is a real skill; it is not, by itself, a substitute for the formation that makes judgment trustworthy.
Debates & Critiques
Whether Newman's grammar applies uniformly across all domains, or whether certain highly formalized domains (mathematics, theoretical physics) escape its jurisdiction, remains contested. Most contemporary readers accept that concrete matters — practical reasoning, historical judgment, moral conviction, and the evaluation of AI outputs — fall squarely within its scope.
Appears in the Orange Pill Cycle
Further reading
- John Henry Newman, An Essay in Aid of a Grammar of Assent (1870)
- Chase Mitchell, 'A Grammar of Descent,' Christian Scholar's Review (2025)
- M. Jamie Ferreira, Doubt and Religious Commitment (1980)
- Basil Mitchell, The Justification of Religious Belief (1973)
- Frederick Crosson, 'Newman and the Problem of Inference,' Proceedings of the ACPA (1989)