This project examines the role of Niels Bohr’s notion of classical concepts in the use and interpretation of quantum theory, and explores its operational consequences. Prior work has shown that Bohr’s “classical concepts” are conceptually distinct from “classical mechanics”: whereas classical mechanics concerns dynamical laws, classical concepts concern the epistemic and linguistic preconditions required to describe experiments and communicate their results (see, for example, Faye’s reconstruction of Bohr).

In the first part of the project, I investigate how these epistemic and linguistic preconditions might be formalized. In particular, they involve (i) the capacity to describe experimental outcomes as definite occurrences in ordinary language, and (ii) the capacity to communicate such occurrences intersubjectively.

In the second part, I explore how this formalization can be applied to concrete physical scenarios, and what logical and conceptual consequences follow for the quantum–classical transition and for interpretations of quantum theory.

This project investigates quantum mereology—the relation between parts and wholes in quantum systems—using an operational approach that minimizes historically inherited metaphysical assumptions whose applicability may be limited. 

Since Descartes, the dominant explanatory mode in physics has been reductionist: a system’s behavior is taken to be nothing more than the aggregate behavior of its constituent parts, which are presumed to exist independently and to be distinctly identifiable. In classical physics, this picture works well: a heap of sand can be treated as the sum of its grains, each with independent and well-defined existence, and classical statistical mechanics fits comfortably within this framework. 

 This project revisits this underlying mereological assumption, because the practice of quantum mechanics complicates it.  An operationally rigorous part–whole relation in quantum systems must account for the probing mechanisms through which “parts” are defined and accessed. This point has been demonstrated in the extensive literature on entanglement in identical particles, for example, in Paolo Zanardi et al.’s “Quantum Tensor Product Structures Are Observable-Induced” (Physical Review Letters 92, no. 6, 2004). 

Furthermore, this mereological assumption, when applied to quantum many-body theory, amounts to this conclusion: macroscopic objects are composed of particles; particles obey quantum theory; therefore, macroscopic objects obey quantum theory. This conclusion, however, is not a logical necessity of quantum theory per se, but rather the manifestation of a metaphysical assumption concerning mereology. 

Revisions to this mereology, motivated by a desire to maintain closer alignment with experience, do not affect whether quantum theory is universal but instead affect the sense in which it is universal. 

Can current or near-future AI models be made to ask the right questions in physics research and make conceptual discoveries in physics, as opposed to their current abilities, which include executing symbolic manipulations and simulations for pre-specified problems? This is a new project (started in August 2026) in which I am exploring the above questions from different angles. The immediate goal is to clarify the nature of conceptual discovery and thus pinpoint the appropriate training data; the next goal is to build AI systems, either using a combination of existing learning models or new learning models, to implement such discovery in simplified world models. More on this project can be found in this project plan 

This is an interdisciplinary optimization problem. Two extremes come to mind in human–AI collaboration. 

Right extreme: no AI at all—just calculators and sustained human thought.

Left extreme: AI for every task AI could do. 

We need to optimize between these two extremes such that the following conditions are met:

1. Need for sustained, skilled, human thought to manipulate their working memory in synthetic ways. If humans are to retain at least some creative command, such sustained, skilled thought cannot be avoided. Creative insights rarely come from fragmented attention.

2. The individual's need for purpose, satisfaction, and social and economic relevance from their craft.

This problem is unique to the new technology we have in our hands. When computers and electronics came, or when industrial-scale machines and heat engines came, 2) was certainly affected for certain classes of professionals, but nowhere near the scale we are witnessing now, and certainly 1) was not affected (as long as we were not getting distracted by the entertainment, such as video games, that the technologies provided). The presence of computer software or calculators did not substantially negatively affect the human learning process and the ability to have sustained, uninterrupted thought trains.