AI-Native Science & Knowledge Infrastructure
I study how scientific knowledge can be created, independently verified, and continuously accumulated when AI makes candidate research abundant.
Undergraduate researcher · The University of Hong Kong Research Assistant · CAMO
A research program in four linked questions
Current empirical work
The study pairs short management-research manuscripts with controlled, consequential errors. Each error must change at least one correct response in an existing structured review form. Human and AI reviewers receive the same materials, task, and evaluation rules.
Current evidence state
This is the first empirical test of the verification architecture developed in the conceptual work below.
Conceptual foundations
DGP Multiverse Science as an AI-Native Architecture for Scientific Knowledge
Proposes a scientific architecture in which competing data-generating-process hypotheses are produced, tested, independently verified, and accumulated as a living, versioned knowledge state.
The Knowledge Abundance Paradox and the Institutional Economics of Scientific Coordination
Explains why cheaper research production can raise the downstream costs of understanding, verification, and accumulation—and specifies when a DGP-linked alternative should be tested, redesigned, or rejected.
Research direction
Institutional theory explains why change may be needed. Scientific architecture defines a candidate alternative. Empirical verification tests whether it works. Living infrastructure should follow only when those tests succeed.