{"id":"668f357e-f684-4d70-ba31-929460303239","arxiv_id":"1909.02297","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"ΦID decomposes the mutual information between the past and future states of a bivariate Markov process into sixteen atoms, showing that integrated information measures blend distinct dynamical modes such as storage, transfer, and synergistic causation.","lead":"Researchers introduce Integrated Information Decomposition (ΦID), a mathematical framework that splits the information a bivariate system carries from its past to its future into sixteen distinct parts, each corresponding to a different mode of information dynamics. The decomposition shows that popular measures of 'integrated information' are mixtures of qualitatively different phenomena, which may explain their inconsistent behavior.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"ΦID atoms are underdetermined by the axioms (Prop. 1), so the taxonomy may be convention-dependent unless the qualitative claims are invariant across admissible redundancy choices.","rationale":"The reader's weakest-assumption identification is exactly right: the double-redundancy function and the single-target redundancy function are not fixed by the axioms, and Proposition 1 explicitly names both inputs. This is the most load-bearing concern because the paper's headline contributions are taxonomic: atoms are assigned to phenomena such as upward causation and synergistic storage, and existing measures are dissected into these atoms. If those assignments depend on arbitrary choices, then the taxonomy is not a unique description of the system but a family of possible descriptions. The paper is honest about this limitation and even supplies a concrete implementation (Appendix C), so this is not an internal inconsistency; it is a correctness/interpretability risk. The proposed test would settle whether the qualitative demonstrations are robust across admissible choices. I therefore recommend keeping the CONDITIONAL verdict rather than upgrading to ACCEPT or downgrading to REJECT: the mathematical core is sound and the non-uniqueness is acknowledged, but the central interpretive claim needs either an invariance result or a canonical choice of redundancy functions.","tokens_in":14933,"tokens_out":8783,"duration_ms":94358,"concrete_test":"Recompute the full 16-atom ΦID for the three Fig. 3 systems (copy transfer, downward XOR, PPR) using at least two distinct single-target PID redundancy functions (e.g., I_min and the I_dep construction from Appendix C) and two admissible choices for I^{{1}{2}->{1}{2}}∩ (e.g., its Axiom-2 lower bound 0 and the value from the Appendix C double-unique construction), then compare the nonzero-atom patterns and the Table I sign pattern. If the single-nonzero-atom patterns survive across choices, the qualitative taxonomy is robust; if atoms shift or become negative under otherwise valid choices, the central claims are convention-dependent.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 1 makes the underdetermination explicit: Axioms 1 and 2 fix all double-redundancies except I^{{1}{2}->{1}{2}}∩, and the atoms additionally require a single-target PID redundancy function Red(·). Because PID has multiple competing redundancy functions (I_min, I_dep, I_ccs, etc.) with no consensus, the numerical values of all 16 atoms — and therefore every conclusion identifying a specific storage, transfer, or causation atom — depend on two unforced choices. This matters for the central claim: the statement that 'integration' is really an aggregate of heterogeneous phenomena is only meaningful relative to a particular decomposition. The Fig. 3 examples are plausibly robust, since the Appendix proofs rely only on Red≤min, nonnegative double-redundancy, and the partial-order axiom, but the paper does not show these extra properties follow from Axioms 1–2 and offers no canonical choice for the double-redundancy term. Until invariance or a principled selection rule is established, the taxonomy is a family of decompositions, not a determinate one.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces Integrated Information Decomposition (ΦID), a framework that decomposes the multivariate mutual information E = I(X1,X2;Y1,Y2) of a bivariate Markovian system into 16 atoms organized on a product lattice of double-redundancies. The construction combines Williams and Beer's partial information decomposition (PID) with a second lattice over target variables, defining double-redundancy functions and using Möbius inversion to obtain atoms. The authors show that the resulting decomposition can be interpreted as a taxonomy of six information-dynamics phenomena (storage, copy, transfer, erasure, downward causation, upward causation), and they use it to argue that standard integrated-information measures such as ΦWMS conflate qualitatively different modes of information flow. They also provide example systems (copy transfer, downward XOR, parity-preserving random) with identical ΦWMS but different non-zero ΦID atoms, and they analyze why ΦWMS can be negative and why unnormalized causal density can exceed total mutual information.","tokens_in":15240,"tokens_out":4791,"duration_ms":46474,"significance":"If the construction is accepted, ΦID is a conceptually valuable extension of PID that offers a principled way to separate storage, transfer, and higher-order causal effects in multivariate time series. The paper's structural results—the lattice decomposition, the algebraic identities for ΦWMS and uCD, and the demonstration that different systems with the same scalar integrated information can differ in their information dynamics—are of genuine interest to the information dynamics and IIT communities. The proofs in the appendices are careful under the stated assumptions, and the authors are transparent about many limitations. The main open question is whether the qualitative taxonomy is invariant under the choice of redundancy function; this is the key issue that prevents the framework from being fully determinate.","major_comments":[{"comment":"The proofs of the example systems and Table I rely on assumptions beyond Axioms 1 and 2: non-negativity of the double-redundancy function and the bound Red(X,Y;Z) ≤ min{I(X;Z), I(Y;Z)}. These are plausible and satisfied by common PID redundancy functions, but they are not stated as axioms of ΦID, and the paper does not show that they follow from Axioms 1–2. Consequently, the claims of Appendices D and E are conditional on a restricted class of ΦID instances, not on the framework as defined. The manuscript should specify that these results hold for any ΦID satisfying the stated extra conditions, and should make clear whether the authors regard these conditions as part of the definition of a 'valid' ΦID or as additional hypotheses for the examples.","section":"Appendices D and E"}],"minor_comments":[{"comment":"The claim that upward causation and synergistic storage 'have, to our knowledge, not been reported in the literature' is a strong novelty claim. The authors may wish to soften it or provide a more systematic literature search, since related ideas may exist in the synergy literature.","section":"Discussion"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What you should know: this is a solid conceptual paper that builds a two-lattice PID product to decompose multivariate mutual information into 16 atoms, and uses it to dissect existing integrated information measures. The decomposition construction is genuinely new, the lattice proof is fine, and the worked examples are careful.\n\nThe paper does several things well. It provides a clean formal scaffolding (product of redundancy lattices, Möbius inversion) and shows how ΦWMS, CD, ψ, and ΦG decompose into different subsets of atoms, which explains their inconsistencies. The three example systems (copy, XOR, PPR) with equal ΦWMS but different atoms are a nice illustration. The appendices give real proofs, including a procedure to build single-atom systems and a concrete computational route via Idep double-unique information. That is reproducible work.\n\nThe main soft spot is the one the reader flagged: the ΦID atoms are not uniquely determined. Axioms 1 and 2 fix most double-redundancies, but the bottom atom I^{{1}{2}→{1}{2}} is free, and any single-target PID redundancy function has to be chosen. Proposition 1 says this explicitly. The authors are honest about it, and they offer one specific construction (double-unique based on dependencies) for numerics, but they do not make a case that this choice is canonical or that the qualitative taxonomy is invariant across legitimate choices. That matters: calling \"integration\" an aggregate of heterogeneous phenomena is a statement about a particular decomposition family, not about the underlying system in a decomposition-independent way. The Fig. 3 examples are likely robust—the proofs use only weak assumptions (nonnegative double-redundancy, Red ≤ min)—but that robustness is not established as a general theorem.\n\nI also think the taxonomy of six phenomena is more interpretive than the formal part; the mapping of atoms to \"copy,\" \"erasure,\" \"upward/downward causation\" is natural but not forced by the math. That is okay for a conceptual paper, but it should be labeled as interpretation.\n\nWho this is for: anyone working on PID, IIT, or information dynamics in complex systems. It is a well-written, transparent framework paper with real proofs and real examples. It deserves peer review. I would suggest the referee request a discussion of the uniqueness/invariance issue, possibly a canonical double-redundancy axiom or invariance results for the qualitative claims. If the authors cannot prove invariance, they should at least map which taxonomy conclusions depend on the choice.","headline":"A genuinely useful decomposition framework that resolves some cross-measure inconsistencies, but its taxonomy is a family of decompositions until the double-redundancy term is pinned down or shown irrelevant.","tokens_in":15681,"tokens_out":2017,"would_cite":true,"duration_ms":20882,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["94A17"],"pacs":[],"model":"deepseek-v4-flash","headline":"A 16-atom decomposition of information flow shows why 'integration' is not one thing.","keywords":["integrated information","partial information decomposition","information dynamics","redundancy lattice","synergy","transfer entropy","causal density","Möbius inversion"],"falsifier":"Take the copy-transfer system defined in the paper (X1, X2, Y1 fair coin flips, Y2 = X1) and compute all 16 atoms with two different valid single-target redundancy functions, such as a minimum-redundancy function and the dependency-based $I_{\\mathrm{dep}}$ measure. If either function assigns nonzero weight to any atom other than $I^{\\{1\\}\\to\\{2\\}}_{\\partial}$, the paper's claim that its example proofs hold for all $\\Phi$IDs satisfying the axioms would fail, and the qualitative taxonomy would depend on an arbitrary choice.","tokens_in":14739,"feed_emoji":"🧩","tokens_out":5824,"duration_ms":56089,"temperature":0.7,"pith_summary":"This paper tries to replace the one-number idea of 'integrated information' with a richer account: it decomposes the total mutual information flowing from a system's past to its future into 16 disjoint atoms, each corresponding to a distinct way information can be stored, copied, transferred, erased, or caused upward or downward. It argues that what previous measures call integration is really an aggregate of heterogeneous phenomena, so two systems with the same integrated-information score can be dynamically unlike. It then uses the decomposition to explain known pathologies of existing measures and to show how purpose-built measures could be tailored to specific information-dynamics phenomena.","feed_headline":"Information flow splits into 16 atoms, not one 'integration' number","feed_subtitle":"Systems with identical integrated-information scores can still store, copy, transfer, and erase information in different ways.","key_machinery":"The central object is the double-redundancy lattice $A \\times A$, built as the product of the PID redundancy lattice for the source variables and for the target variables. Its 16 nodes $\\alpha \\to \\beta$ carry values of a double-redundancy function $I^{\\alpha \\to \\beta}_{\\cap}$, and Möbius inversion over this lattice produces the 16 $\\Phi$ID atoms $I^{\\alpha \\to \\beta}_{\\partial}$. Two axioms (compatibility with PID and Shannon terms, and monotonicity under the partial order) plus one free choice of the bottom atom $I^{\\{1\\}\\{2\\}\\to\\{1\\}\\{2\\}}_{\\partial}$ and a single-target redundancy function determine all atoms; the paper computes them numerically using a dependency-based unique-information measure.","core_discovery":"The central discovery is that the multivariate mutual information $E = I(X_1,X_2;Y_1,Y_2)$ of a two-variable Markovian process can be broken into 16 atoms arranged on a product of two PID redundancy lattices—one for the past and one for the future—via Möbius inversion of a double-redundancy function. These atoms correspond to qualitatively different dynamical phenomena: storage, copy, transfer, erasure, downward causation, and upward causation. The paper demonstrates that the widely used measure $\\Phi_{\\mathrm{WMS}}$ assigns the same value (1 bit) to three simple systems—copy transfer, downward XOR, and parity-preserving random—while their $\\Phi$ID decompositions are completely different, with only $I^{\\{1\\}\\to\\{2\\}}_{\\partial}$, $I^{\\{12\\}\\to\\{1\\}}_{\\partial}$, or $I^{\\{12\\}\\to\\{12\\}}_{\\partial}$ nonzero respectively. It also shows that the four existing measures $\\Phi_{\\mathrm{WMS}}$, CD, $\\psi$, and $\\Phi_G$ respond to different subsets of atoms, implying they are not approximations of one concept but capture different phenomena.","pith_inferences":["If the taxonomy is adopted, empirical studies could classify dynamical regimes by their atom profiles rather than by a scalar, for example comparing conscious versus unconscious brain states; the paper itself does not report such applications.","Because the axioms leave the bottom atom unspecified, quantitative $\\Phi$ID results may depend on the chosen redundancy function; comparing atom values across studies will require standardizing that choice or reporting robustness across choices.","The framework is developed for bivariate Markovian systems, so extending the 16-atom picture to more variables or non-Markovian processes would require additional theory; the paper notes the Markovian restriction and leaves it for future work.","The same decomposition could be used to construct new targeted integration measures for specific applications, choosing weights over atoms to match the phenomenon of interest; the paper suggests this possibility but does not build such measures."],"forward_implications":["Systems with identical $\\Phi_{\\mathrm{WMS}}$ can have qualitatively different information dynamics, so scalar integrated-information scores should not be read as measuring a single kind of integration.","The taxonomy gives six disjoint phenomena—storage, copy, transfer, erasure, downward causation, and upward causation—that can be quantified separately from time-series data.","Existing measures' inconsistencies are explained: each measure weights a different combination of atoms, so disagreements are in-principle, not just practical.","The decomposition pinpoints why $\\Phi_{\\mathrm{WMS}}$ can be negative (a negative double-redundancy term) and suggests the corrected measure $\\Phi_{\\mathrm{WMS,c}}$.","It diagnoses why unnormalised causal density can exceed total mutual information: it double-counts the downward-causation atom $I^{\\{12\\}\\to\\{1\\}\\{2\\}}_{\\partial}$; a system with $y_1=y_2=x_1\\oplus x_2$ gives uCD = 2 bits vs. $E = 1$ bit."],"supporting_citations":[{"why":"Provides the partial information decomposition and redundancy lattice that $\\Phi$ID generalises to multiple targets.","marker":"[14]"},{"why":"Supplies the dependency-based unique-information measure used to compute numerical $\\Phi$ID atoms.","marker":"[28]"},{"why":"Documents the inconsistent behaviour of integrated-information measures that $\\Phi$ID explains.","marker":"[10]"},{"why":"Defines the $\\Phi_{\\mathrm{WMS}}$ measure that the paper decomposes and corrects.","marker":"[25]"},{"why":"Introduces $\\Phi_G$ and the observation that unnormalised causal density can exceed total mutual information, which $\\Phi$ID diagnoses.","marker":"[27]"},{"why":"Proposes the storage/transfer/modification taxonomy that the six-category $\\Phi$ID taxonomy builds on.","marker":"[13]"},{"why":"Shows transfer entropy is not a pure measure of transfer, motivating a finer decomposition.","marker":"[12]"}],"fun_headline_variants":["Integrated information? It splits into 16 distinct dynamics atoms","ΦID breaks integration into 16 atoms: storage, copy, transfer, erasure","One Φ score can't capture the 16 modes of information dynamics","Information dynamics taxonomy: 16 atoms replace single integration number"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The axioms of $\\Phi$ID do not determine the double-redundancy atom $I^{\\{1\\}\\{2\\}\\to\\{1\\}\\{2\\}}_{\\partial}$, and together with the choice of a single-target redundancy function this free choice fixes the numerical values of all 16 atoms.","fun_headline_variants_meta":{"raw":{"variants":["Integrated information? It splits into 16 distinct dynamics atoms","ΦID breaks integration into 16 atoms: storage, copy, transfer, erasure","One Φ score can't capture the 16 modes of information dynamics","Information dynamics taxonomy: 16 atoms replace single integration number"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00107,"raw_usage":{"total_tokens":4505,"prompt_tokens":988,"completion_tokens":3517,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":604,"completion_tokens_details":{"reasoning_tokens":3442}},"tokens_in":604,"tokens_out":3517,"duration_ms":23298,"temperature":1.0,"reasoning_tokens":3442,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T04:54:39.498861+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the copy-transfer system defined in the paper (X1, X2, Y1 fair coin flips, Y2 = X1) and compute all 16 atoms with two different valid single-target redundancy functions, such as a minimum-redundancy function and the dependency-based $I_{\\mathrm{dep}}$ measure. If either function assigns nonzero weight to any atom other than $I^{\\{1\\}\\to\\{2\\}}_{\\partial}$, the paper's claim that its example proofs hold for all $\\Phi$IDs satisfying the axioms would fail, and the qualitative taxonomy would depend on an arbitrary choice.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the dependency-based unique-information measure used to compute numerical $\\Phi$ID atoms."},{"cited_title":"Dosi and A","cited_arxiv_id":null,"evidence_quote":"Documents the inconsistent behaviour of integrated-information measures that $\\Phi$ID explains."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the $\\Phi_{\\mathrm{WMS}}$ measure that the paper decomposes and corrects."},{"cited_title":"We focus on the bivariate case for clarity, although our results hold for any N","cited_arxiv_id":null,"evidence_quote":"Introduces $\\Phi_G$ and the observation that unnormalised causal density can exceed total mutual information, which $\\Phi$ID diagnoses."},{"cited_title":"Balduzzi and G","cited_arxiv_id":null,"evidence_quote":"Proposes the storage/transfer/modification taxonomy that the six-category $\\Phi$ID taxonomy builds on."},{"cited_title":"Tononi, O","cited_arxiv_id":null,"evidence_quote":"Shows transfer entropy is not a pure measure of transfer, motivating a finer decomposition."}],"review_version":1}