October 04th, 2026

Updated November 2nd, 2026 – version 5.27

AI-Driven Framework for Epistemology

 

What happens if you take Kant's three faculties of mind – understanding, judgment and reason –

add Freud's unconscious and some modern brain research, and then ask an AI to help turn it all into one diagram?

To put it briefly: it starts with a colorful picture and ends with a model.

 

 

The first draft looked great but did not hold up. Emotions showed up twice, some labels were mixed up,

and the "information flow" contradicted the very claim that cognition is not a pipeline.

Claude Opus 5.5 and I went through it piece by piece – first in a single day,

and since then in many rounds of critical reviews by other AIs (ChatGPT, Gemini and models on arena.ai):

• Every element now has exactly one place: nine domains in a recursive "Cognitive Field", plus eight categories on which every domain takes a value.
• A domain has to earn its place: it must be separable by lesions or developmental disorders, explain something nothing else explains, and predict something new.
  That is why judgment is now part of understanding, and language became a domain of its own.
• Below the domains lies the depth of the field, after Freud: preconscious and unconscious, then the drives and the body that carries it all.
• The model was tested against real cases: the conscience, the soul in Kant's sense, the DMT state, autism, a fruit fly, a raven – and finally against AI itself.
• It makes six testable predictions and fixes eleven tests in advance, each with a condition under which the model would be wrong.

 

From Freud the model takes one idea in particular: conscious, preconscious and unconscious are not a separate axis

but the depth of the field – with the preconscious as a gatekeeper that decides what reaches awareness.

 

The result is an interactive page. Hover over a domain to see its connections, or use the "lens" to see how a phenomenon

or a species maps onto the model. Below the diagram there is a preprint-style article with abstract and references.

Architecture of Human Cognition – interactive model and article

Also available in Deutsch, Français and in short, easy-to-read versions: Simple English, Einfaches Deutsch, Français, Español, Bahasa Indonesia, 中文, العربية and Русский (translated with AI, not yet reviewed by native speakers).

So far the model has been checked by AIs only, not yet by human experts – critical feedback is very welcome.

 

 

Key works behind the model

The books and papers our model builds on or sets itself apart from – with one line on what we take over.

The invariance criterion is our own idea and does not derive from any of them. Full references are in the article.

The thought experiment behind the invariance criterion: about 125 million years ago, a tree falls by a lake in today's north-eastern China.

Eomaia, Sinosauropteryx and Confuciusornis hear different sounds, yet register the same place, time and sequence – what stays the same belongs to the world.

AI illustration: ChatGPT (OpenAI), created on behalf of Sebastian Bohnen, October 31st, 2026. A narrative illustration, not a scientific reconstruction.

Epistemology and ontology

• Aristotle, On the Soul (De anima) – The soul as the form of the living body – oldest neighbour of our formula “carrier and carried”.
• Locke 1690, An Essay Concerning Human Understanding – Primary and secondary qualities as a forerunner; the invariance criterion turns them into a test.
• Kant 1781/1787, Critique of Pure Reason – The faculties from sensibility to reason, the sequence of B 355, modality, and opining – believing – knowing; read realistically in our model.
• Kant 1788, Critique of Practical Reason – Reason as the faculty of principles, the higher faculty of desire and the highest good – in our model the principle formation of reason and the upper level of striving.
• Kant 1790, Critique of Judgment – Judgment, the beautiful and the sublime.
• Husserl 1913, Ideas Pertaining to a Pure Phenomenology (Ideas I) – How something is given in experience: perceived, remembered or merely imagined, and as certain, presumed or doubted – a parallel to our axes modality and certainty, not a derivation.
• Popper 1934, The Logic of Scientific Discovery – Falsification as the standard; hence tests fixed in advance, each with a refutation condition.
• Lorenz 1941, Kants Lehre vom Apriorischen im Lichte gegenwärtiger Biologie – The a priori as shaped by evolution – with Kant we share the question, with Lorenz the answer.
• Worrall 1989, Structural Realism: The Best of Both Worlds? – Structural realism – by contrast, in our model structure also lies in the subject.

Depth, drives, will

• Schopenhauer 1844, The World as Will and Representation, vol. 2 – The will as the basic force – forerunner of, and contrast to, our striving from drives and will.
• Freud 1915, Instincts and Their Vicissitudes – Drive as a frontier concept with source, pressure, aim and object – the drive layer in our model.
• Freud 1923, The Ego and the Id – The topography of conscious, preconscious and unconscious – in our model the depth of the field, with the preconscious as gatekeeper.

Functional circle, brain, emotion

• Dewey 1896, The Reflex Arc Concept in Psychology – Circle instead of reflex arc – the basic form of our circle between world and action.
• Luria 1973, The Working Brain – Functional systems instead of brain sites – which is why our domains are defined functionally.
• Gardner 1983, Frames of Mind – Multiple intelligences – a neighbour that sorts by contents, where we sort by processes.
• Baars 1988, A Cognitive Theory of Consciousness – Global workspace – access in our model; what is our own in the experience thesis is the three markings, not the broadcast.
• Johnson, Hashtroudi & Lindsay 1993, Source Monitoring – Source monitoring – empirical basis of the modality axis.
• Damasio 1994, Descartes’ Error – Emotion as part of reasonable decision-making – couplings of emotion with understanding and reason, gut feeling.
• Panksepp 1998, Affective Neuroscience – Basic affective systems in humans and animals – support for the drive layer.
• Fleeson 2001, Traits as Density Distributions of States – Traits as distributions of states – character, baseline and shift.
• Clark 2013, Whatever Next? Predictive Brains … – Predictive processing – cognition as a circle that predicts what comes next.
• Fleming & Lau 2014, How to Measure Metacognition – Making metacognition measurable – reason and the certainty axis, test 11.

Consciousness

• Nagel 1974, What Is It Like to Be a Bat? – What it is like to be another creature – the limit of other minds.
• Dennett 1991, Consciousness Explained – No Cartesian theatre – counter-position and neighbour of our answer.
• Chalmers 1995, Facing Up to the Problem of Consciousness – The hard problem – our guiding question alongside Kant.
• Metzinger 2003, Being No One – Self-model and transparency – basis of our bridging assumption.
• Kriegel 2009, Subjective Consciousness – Self-representation – nearest neighbour; we add the three markings and the transparency effect.
• Frankish 2016, Illusionism as a Theory of Consciousness – Illusionism – in his terms our position, at its core an identity thesis, is closer to weak than to strong illusionism.

 

The work raises a couple of questions. To me two stand out:

• Can altered states of mind be mapped onto just a few axes?
• Can a language model look inside itself?