Pedro A. M. Mediano, Fernando E. Rosas, Andrea I. Luppi, Henrik J. Jensen, Anil K. Seth, Adam B. Barrett, et al. (2022), Philosophical Transactions of the Royal Society A (vol. 380, issue 2227, art. 20210246; theme issue “Emergent phenomena in complex physical and socio-technical systems”) — ARXIV-2111.06518
Key takeaways
- Within the ΦID framework, a system can host causally emergent features of order k if and only if its dynamics carry k-th order synergy about their own future. That emergence capacity splits exactly into downward causation plus causal decoupling. Both can be detected with redundancy-function-free criteria that use only low-order marginals.
Seeded from the abstract and a skim, before the reading in NOTE-010. What follows is what the work says about itself.
Emergence is widely discussed, from galaxy formation to consciousness, but it lacks formalisms that make it measurable. This review restates and extends a recent theory of causal emergence built on information decomposition. In that theory, a system is emergent to the extent that information about its temporal evolution cannot be obtained from its parts taken separately. The article aims to be an accessible but rigorous introduction, discusses the approach’s merits across scenarios, and addresses common misunderstandings and interpretive issues.
Standing in the record
Filed by the reading-time triage of 2026-09-25: 1,275 seconds of active reading over 6 sessions in the papers-feed tracker. It was filed Deferred, unread. NOTE-010 is the close reading of 2026-09-25, and it placed the work: Active — a short, accurate map of the ΦID emergence formalism and the right entry point to Rosas et al. 2020, which a reader will still need for the proofs, the assumptions and the redundancy-free practical criteria. None of these is stated in the review.
When seeded, before the reading in NOTE-010, the entry gave a priority for a deeper reading: medium — A clear, short, foundational entry to the ΦID emergence formalism, with possible ML-interpretability relevance. The skim captures the definitions, so the next step is the primary Rosas et al. 2020 paper rather than this one (t = 1275 s already spent).
What the seeding asked a deeper reading to check (NOTE-010 is that reading):
- The iff theorem between emergence and synergy is the load-bearing result. It lives in the cited Rosas et al. (2020, PLoS Comp Bio) paper, and the review only restates it; check the assumptions there. Also check how much the results depend on the choice of redundancy function.
- ML link, stated honestly: PID/ΦID synergy measures are used in interpretability and representation analysis (e.g., synergy between neurons or features). This review is a compact entry point, but it contains no ML experiments.
- Coordinate dependence is a real limitation when the “parts” of a learned representation are arbitrary, for example a basis choice in neural activations.
Access when seeded: Crossref metadata and abstract for the DOI (online 2022-05-23). royalsocietypublishing.org returned 403 for the PDF. I read the full text of the arXiv preprint 2111.06518v1 (12 Nov 2021, 8 pp.). The remaining authors are Robin L. Carhart-Harris and Daniel Bor, eight in total. No keyword list appears in the preprint; the keywords above are mine.