Epistemica · Selected Insights
Research, working papers, and decision instruments on Applied AI Epistemics, AI governance, and development finance — held together by one question: which claims survive scrutiny, and which merely look like they do.
Featured · External
AI governance frameworks are proliferating, yet the public whose lives they regulate remains structurally excluded from their design. This piece interrogates the epistemic legitimacy of current AI governance architectures and argues that without genuine democratic mandate, they function as technocratic illusions — authoritative in form, hollow in substance.
Working Papers · Applied AI Epistemics
Examines the phenomenon whereby AI-generated outputs create the appearance of genuine knowledge while concealing the absence of understanding. The "epistemic mirage" is not merely hallucination — it is a structural feature of probabilistic systems operating in meaning-laden domains where surface fluency masks epistemic vacancy.
Investigates how cognitive depletion — the gradual erosion of critical reasoning under conditions of information overload — creates conditions in which epistemic arrogance flourishes. In AI-mediated environments, users who offload reasoning to systems may systematically over-estimate the quality of their own remaining judgements.
A structural account of how AI is entering higher education — mapping the distinct role domains AI occupies (from assessment to pedagogy), the archetypes of institutional response, and the allocation logic that governs deployment. Challenges both the optimism of seamless integration and the anxiety of wholesale displacement.
Decision Instruments
An epistemic approach to South Africa’s R50.4 trillion protection gap. Read the paper in full, hear the 23-minute discussion with the paper answering alongside it, move the model’s own parameters and watch every figure recompute, or put a question to it. Each figure carries its source; each result shows its arithmetic.
Development finance institutions measure success by capital deployed. This instrument measures a different thing: how many times a unit of scarce capital can be recycled into new developmental lending before it retires. Enter your own capital, prudential and securitisation parameters, and it recomputes the net capital release, the velocity ratio, the cumulative multiplier and five prudential constraint tests — exposing the gap between what a placed-share calculation would claim and what retained-position capital charges actually permit. Every figure is computed deterministically from the inputs. No estimation, no AI.
More from Epistemica
Explore the Institut — live AI tools built on Applied AI Epistemics.
We run a South African site and an international one. They carry the same work — picking your region just points you at the right one.