Higher confidence in an AI system was associated with reduced critical thinking (β = −0.69, p < 0.001), while higher confidence in one’s own ability was associated with more of it (β = 0.26, p = 0.026).
The Science
It is not artificial intelligence that degrades human judgment. It is undesigned artificial intelligence.
The evidence
What the published record already shows.
Across the published experimental record, pairing a person with a system produced results worse than the better of the two working alone (g = −0.23, 95% CI −0.39 to −0.07). The losses were concentrated in decision-making. The gains were in producing content. Putting a person and a machine together is not, by itself, an improvement.
Vaccaro, Almaatouq & Malone, Nat Hum Behav 8:2293–2303. doi:10.1038/s41562-024-02024-1
When the system proposed the wrong category, correct readings by the five most experienced radiologists in the study fell from 82.3% to 45.5% (P = .003). These were readers with a decade at the workstation. Experience did not protect them from a confident machine.
Dratsch et al., Radiology 307(4):e222176. doi:10.1148/radiol.222176
Structured, guided prompting reduced cognitive offloading. Unguided use of the same underlying model did not. The instrument was constant; the design of the interaction was not.
Routine AI assistance was introduced. Afterwards, when the same endoscopists worked without it, their adenoma detection rate had fallen from 28.4% to 22.4% — six percentage points, absolute (95% CI −10.5 to −1.6, P = .009). Fifteen of the nineteen declined individually.
The authors are careful. The study is observational, and they offer deskilling as a hypothesis for confirmatory trials rather than as a settled finding. It remains the first clinical evidence that the capability itself can quietly recede — and it was measured in patients, not in a laboratory.
Budzyń et al., Lancet Gastroenterol Hepatol 10(10):896–903. doi:10.1016/S2468-1253(25)00133-5
The design of the interaction determines whether the mind is exercised or retired.
The centrepiece
The Recursive Human Systems Model™
A lens, not a mechanism. Six states that condition one another in sequence and then return to the beginning. Select any node to read it.
Why recursion is the point
Most models of behaviour are linear: a stimulus produces a response, and the account ends there. Institutions inherit the consequences of that assumption — they intervene once, at one point, and the loop restores its prior state within a quarter.
The Recursive Human Systems Model™ treats the sequence as closed. Consequence teaches belief; belief conditions the next thought; the next thought is not neutral. A single well-placed intervention changes the entire circuit, and a poorly placed one is absorbed by it.
The model is published here as a lens. The elicitation design, sequencing, and weighting behind URIEL are not described on this site.
The construct hierarchy
How the intellectual property nests.
-
The governing discipline: the study of how human judgment is strengthened, or quietly eroded, by repeated interaction with intelligent systems.
-
Augmented NeuroSynthesis™
The applied method by which behavioural, cognitive, and clinical evidence is brought together into a single coherent interaction with a person.
-
Recursive Human Systems Model™
The lens: thought, emotion, neurochemistry, behaviour, consequence, and belief as one closed and trainable circuit.
-
Cognitive Ecology™
The environment a mind is asked to operate inside — its load, its interruptions, and its permissions.
-
Cognitive Coherence™
The alignment between what a person understands, what they feel, and what they are able to do next.
-
Ethical Cognition™
Judgment exercised under pressure without abandoning the standards the institution says it holds.
The evidence base
Eighteen frameworks, none of them named here.
URIEL™ integrates eighteen validated behavioural, cognitive, and clinical frameworks spanning the motivational, cognitive-behavioural, dialectical, developmental, attachment-based, and neuroplasticity traditions. Each was selected for published validity in applied settings, and each was adapted for use inside a system that must operate under institutional governance rather than in a consulting room.
The frameworks are named in our institutional documentation under NDA. They are not published here, and that is deliberate.
Peer-reviewed work
From the Heart: A Neuro-Behavioural Protocol for Sustainable Change & Patient Adherence
Dr. D. Ivan Young. Accepted for presentation, Lifestyle Medicine 2026, American College of Lifestyle Medicine. Orlando, Florida, November 8–11, 2026.
Publication citation in the American Journal of Lifestyle Medicine to follow.
The researcher
The work behind URIEL sits at an unusually narrow intersection: the applied study of how artificial intelligence and human cognition interact to build resilience, self-advocacy, and adherence. Dr. D. Ivan Young’s research on self-empathy as a trainable clinical skill has been accepted for presentation by the American College of Lifestyle Medicine.
The question it examines is the neurobiology of judgment under pressure — the shift from limbic reactivity to deliberate prefrontal engagement — and whether that shift can be trained rather than merely encouraged. More than two decades of applied practice with clinicians, commanders, executives, and families preceded the frameworks that follow from it.
Claims discipline
Stated plainly, so that nothing has to be inferred.
What is observed
Third-party published research documents both the erosion of critical thinking under unguided AI use and the protective effect of structured interaction design. Those findings are cited above and are not ours.
What is proposed
That interaction design is the decisive variable, and that judgment under pressure can be trained rather than merely encouraged. This is the hypothesis URIEL is built to test in the field, with institutions, over time.
What is not claimed
URIEL is non-diagnostic. It is not a medical device, not a therapist, and not a replacement for clinical judgment. No efficacy figure is claimed for URIEL itself, and none should be inferred from the research cited on this page.
What comes next
The next step is not a larger claim. It is a study.
The assembled capability, inside a named institution, in one of the environments URIEL is built for, measured against outcomes that institution already reports. Institutions willing to host that work — or to field URIEL ahead of it, under their own governance — are the conversation we are having now.
Begin an Institutional Conversation