Discursive capacity is the personal ability to raise validity claims in communicatively oriented discourse and to stand behind them. It is a key concept of the discourse ethics of Karl-Otto Apel and Jürgen Habermas, and it presupposes what an AI system structurally cannot accomplish: the raising — not merely the generating — of validity claims.
Habermas: three validity claims
In The Theory of Communicative Action (1981, vol. 1, ch. III) Jürgen Habermas reconstructs the formal-pragmatic structure of every communicatively oriented speech act: it raises three validity claims at once.
| Validity claim | Reference dimension | Violable by |
|---|---|---|
| Truth | objective world (propositional) | false assertion |
| Rightness | social world (normative) | norm violation |
| Truthfulness | inner world (expressive) | hypocrisy |
Whoever rejects these claims must justify himself — for they are not arbitrary stipulations but structural anticipations of every serious speech act. Whoever denies them in general falls into a performative self-contradiction: in the act of denying, he claims what he thematically denies.
Apel’s transcendental pragmatics
Karl-Otto Apel (Transformation der Philosophie, 1973, vol. 2; German) sharpens this in transcendental-pragmatic terms: whoever argues at all has already acknowledged the communication community as a normative presupposition. To contest this presupposition is possible only under its presupposition — a performative self-contradiction.
The ideal speech situation (Habermas, Wahrheitstheorien, 1972; German) is counterfactual: the counterfactual supposition that only the force of the better argument prevails — an idea that allows real discourses to be judged without ever being fully realized.
Raising — not generating
Here lies the ontological point for the AI debate. In 2022/2023 (among others in Ein neuer Strukturwandel der Öffentlichkeit, 2022; German) Habermas expressed skepticism about AI in discourse. His argument: an AI can formally generate validity claims — it can produce declarative sentences that formally have the structure of a validity claim. But it cannot raise them.
Raising is a performative enactment by a responsible person: someone stands behind the claim, assumes its consequences, is answerable for its violation. Whoever raises a validity claim can also fall into a performative self-contradiction — because he relates performatively to something. A large language model cannot do this. It can produce sentences that contradict one another without this constituting a performative self-contradiction within the discourse — because no subject relates performatively to its statements.
It follows that AI is discourse material, not a discourse participant. Its output can be brought into a discourse — as evidence, as prompt, as datum — but it does not itself discuss.
Conditions of discursive capacity
From Habermas/Apel and the substance-ontological tradition, the conditions of discursive capacity can be named:
- Original intentionality — the participant is the source of his own directed acts.
- Truthfulness — he can bear validity claims because he commits himself to them.
- Free will — he can examine arguments and let himself be persuaded.
- Responsibility — he is accountable for his validity claims.
- Shared form of life — he stands within the web of practice in which validity claims have any sense at all (Wittgenstein).
All five conditions are met only by a person.
Discourse ethics and the dignity of the person
The discourse ethics of Apel and Habermas is given a substance-ontological grounding in the Spaemann/Hildebrand tradition: the formal-pragmatic validity claims that every discourse participant raises presuppose the recognition of the other as a person. Whoever speaks in discourse cannot at the same time treat the other as a thing without contradicting himself. Thus discourse ethics is open to connection with the Personalist Norm.
Ontological classification
- is a capacity of the person in her linguistic-rational existence
- exclusive to persons (axiom: a discourse-capable participant is a person)
- requires: original intentionality, truthfulness, free will, a shared form of life
- constitutive for: dialogue in its argumentative dimension
- distinct from: AI system as quasi-participant
Sources: Generated by querying the Personhood ontology.
Further sources:
- Habermas, Jürgen (1973): Wahrheitstheorien. In: H. Fahrenbach (ed.), Wirklichkeit und Reflexion. Walter Schulz zum 60. Geburtstag. Pfullingen: Neske, pp. 211–265. (German)
- Habermas, Jürgen, The Theory of Communicative Action, 2 vols., transl. Thomas McCarthy. Boston: Beacon Press, 1984/1987.
- Habermas, Jürgen (2022): Ein neuer Strukturwandel der Öffentlichkeit und die deliberative Politik. Berlin: Suhrkamp. (German)
- Apel, Karl-Otto (1973): Transformation der Philosophie, 2 vols. Frankfurt a. M.: Suhrkamp. (German)
- Searle, John R. (1969): Speech Acts. Cambridge: CUP.
- Spaemann, Robert, Persons: The Difference between ‘Someone’ and ‘Something’, transl. Oliver O’Donovan. Oxford: OUP, 2006.
- Wittgenstein, Ludwig, Philosophical Investigations, transl. G. E. M. Anscombe. Oxford: Blackwell, 1953 (bilingual).