Key points are not available for this paper at this time.
Abstract (English) The question of whether Large Language Models (LLMs) can be regarded as rational beings in the Kantian sense is not merely academic—it touches the foundations of our self-understanding as thinking, moral agents. This article develops a systematic catalogue of criteria from two sources: (1) Kant's transcendental-philosophical concept of reason—in particular the hallmarks of Pure Reason from the Critique of Pure Reason (1781/1787) and the concept of autonomy from the Groundwork of the Metaphysics of Morals (1785)—and (2) empirical mentalization research (Theory of Mind), which since Premack and Woodruff (1978) has investigated the capacity to attribute mental states. From this dual foundation, seven dimensions of a Kantian assessment framework for LLMs are derived: (i) a priori structuring, (ii) synthetic a priori judgments, (iii) transcendental unity, (iv) critique and metacognition, (v) antinomy recognition, (vi) Theory of Mind, and (vii) moral autonomy. For each dimension, conceptual operationalizations are sketched and evaluated on the basis of current empirical evidence (2022–2026). Important caveat: the operationalizations are programmatic and conceptual, identifying what would need to be measured rather than providing standardized protocols or validated thresholds. This is a philosophical framework that awaits empirical operationalization, not a ready-made test battery. The result is nuanced: LLMs partially fulfill some criteria—in particular emergent representational structures and functional ToM performance—but fail on the conditions central to Kant: Transcendental Apperception, the Spontaneity of Understanding, reliable metacognition, and Moral Autonomy. The article concludes with the thesis that the genuine philosophical provocation of LLM research lies not in the answer to the question of whether machines think, but in the counter-question posed to the human: What exactly do we mean when we say that we think? Zusammenfassung (Deutsch) Die Frage, ob Large Language Models (LLMs) als vernünftige Wesen im kantischen Sinne gelten können, ist nicht bloß akademisch — sie berührt die Grundlagen unseres Selbstverständnisses als denkende, moralische Akteure. Dieser Artikel entwickelt einen systematischen Kriteriumskatalog aus zwei Quellen: (1) Kants transzendentalphilosophischem Vernunftbegriff — insbesondere den Kennzeichen der reinen Vernunft aus der Kritik der reinen Vernunft (1781/1787) und dem Autonomiebegriff der Grundlegung zur Metaphysik der Sitten (1785) — und (2) der empirischen Mentalisierungsforschung (Theory of Mind), die seit Premack und Woodruff (1978) die Fähigkeit zur Zuschreibung mentaler Zustände untersucht. Aus dieser doppelten Grundlage werden sieben Dimensionen eines kantischen Tests für LLMs abgeleitet: (i) a priori Strukturierung, (ii) synthetische Urteile a priori, (iii) transzendentale Einheit, (iv) Kritik und Metakognition, (v) Antinomien-Erkennung, (vi) Theory of Mind und (vii) moralische Autonomie. Für jede Dimension werden konzeptionelle Operationalisierungen skizziert und auf der Grundlage aktueller empirischer Evidenz (2022–2026) bewertet. Das Ergebnis ist differenziert: LLMs erfüllen einige Kriterien partiell — insbesondere emergente Repräsentationsstrukturen und funktionale ToM-Leistungen — scheitern aber an den für Kant zentralen Bedingungen: der transzendentalen Apperzeption, der Spontaneität des Verstandes, der zuverlässigen Metakognition und der moralischen Autonomie. Der Artikel schließt mit der These, dass die eigentliche philosophische Provokation der LLM-Forschung nicht in der Antwort auf die Frage liegt, ob Maschinen denken, sondern in der Rückfrage an den Menschen: Was genau meinen wir, wenn wir sagen, dass wir denken? CHANGELOG Changes in 8.3 (May 2026): Bibliographic source verification update — added missing DOI, page, issue, and article-number metadata for Bender et al. 2021, Evans 2022, Fazi 2024, Gigerenzer 2024, Lampinen et al. 2024, Macmillan-Scott and Musolesi 2024, Manna and Nath 2021, Orbik 2024, Schüring and Schmid 2024, and Wu et al. 2025; expanded Cheung et al. 2025 and Mitchell 2024 with volume/issue/article metadata; completed the Sacco 2025 title and DOI; moved Wu et al. 2025 from the preprint caveat to the version-of-record list. No change to the article's core argument. Changes in 8.2 (May 2026): Bibliographic source verification update — Xu et al. corrected back to arXiv/CoRR metadata rather than an unverified ACL 2024 proceedings entry; Shin et al. corrected to Journal of Yeungnam Medical Science with DOI; Street et al. updated to the 2026 Frontiers in Human Neuroscience version of record; Qiu et al. updated to Nature Communications 17, 1238 (2026); Tak et al. corrected to 2026 arXiv metadata; Kambhampati, Kosinski, Griot, Fontana, Steyvers, Liu, Sclar, and Cheung DOI/venue details expanded. No change to the article's core argument. Changes in 8.1 (April 2026): Bibliographic correction — Pi et al. updated from arXiv-only preprint metadata to the CogSci 2025 proceedings record; no argument-level changes. Changes in 8.0 (March 2026): Bibliographic corrections — Systematic verification and correction of all references against original publications. Corrections: Mitchell/Science (author), Sparks Rationality (subtitle), Lampinen (title). Changes in 7.0 (March 2026) Two full internal review cycles Major: Added radar diagram, heatmap, distributional rationality concept, Tier 1/2/3 methodology, Reflexive Transparency section, Practical Implications, embodied AI analysis, reasoning models in D4, structural/contingent criteria, Other Minds Problem Minor: Fixed 5 BibTeX keys, updated 3 preprints to peer-reviewed, compressed existentialists and Section 10, added functionalism steelman, CoI/Funding statements DE/EN: German version fully synchronized. Kombi-PDF deprecated.
Building similarity graph...
Analyzing shared references across papers
Loading...
Lukas Geiger
Oldham Council
Building similarity graph...
Analyzing shared references across papers
Loading...
Lukas Geiger (Sun,) studied this question.
www.synapsesocial.com/papers/6a0bfdc7166b51b53d3790de — DOI: https://doi.org/10.5281/zenodo.20251975