{"id":"f70f4a1e-7278-44d6-842f-453493288279","arxiv_id":"2412.10626","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":2.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Meaning is identified with a system's cause-effect structure (Φ-structure), so communication of meaning requires structural similarity between sender and receiver, a stronger condition than symbol transmission.","lead":"This paper contrasts Shannon information theory, which deals with symbols and transmission, with integrated information theory, which identifies meaning with the intrinsic cause-effect structure of conscious systems. It concludes that communicating meaning requires sender and receiver to support similar internal structures, not just exchange symbols.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claim that communicating meaning requires similar Φ-structures is underdefined: no formal similarity measure is given, and footnote 15's 'translation' scenario suggests similarity is an added extrinsic criterion, not a consequence of IIT.","rationale":"The reader identified the missing formal similarity measure as the weakest assumption, and my stress-test converges on the same point. The paper is clearly a position piece: it synthesizes IIT and contrasts it with Shannon information theory, and it explicitly flags the comparison problem in footnote 14. The central inference, however, is not supported by the formal apparatus presented: IIT defines intrinsic meaning as a specific Φ-structure, but it does not define a relation of similarity between two Φ-structures. Without such a relation, the statement that communication of meaning requires similar Φ-structures is not a theorem of IIT; it is a proposed definition of what it would mean for meaning to be communicated. The footnote 15 discussion of translation makes this especially clear, because it allows a mapping between radically different structures to serve as successful communication, which would not be captured by any simple notion of similarity. This does not require rejecting the paper's broader conceptual framework, but it does mean the central claim is conditional on future formalization. The reader's CONDITIONAL verdict is therefore appropriate, and I would not change it.","tokens_in":17702,"tokens_out":4998,"duration_ms":53133,"concrete_test":"Implement exact Φ-structure unfolding for two small (4–6 unit) complexes in the IIT 4.0 formalism, where one complex is a relabeling of the other (same abstract causal graph, same state). Compare the unfolded structures under two candidate metrics: (i) graph-edit distance on the distinction/relation graph, and (ii) an Earth-mover distance on the multiset of φ-weighted cause–effect distributions. If the two metrics disagree about whether a third structure is more similar to one or the other, or if either metric fails to identify the relabeled pair as identical, then the paper's central claim is metric-dependent and undefined without a stipulated choice of similarity measure.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract states that 'for the communication of meaning, the cause–effect structures of sender and receiver must be similar,' but 'similar' is never formally defined. In the section 'Communicating integrated information as meaning,' the authors say that the similarity of two Φ-folds could measure the amount of meaning communicated, and footnote 14 concedes that this is not easy because it requires unfolding large structures and developing optimal comparison methods. This is not a minor technicality: the central conclusion is a relation of communication defined by similarity, and without a metric or equivalence relation on Φ-structures, the claim cannot be evaluated as true or false. Moreover, IIT's fundamental explanatory identity states that a Φ-structure accounts for an experience 'with no additional ingredients.' Similarity between two Φ-structures is an additional, extrinsic relation between two intrinsic structures. Footnote 15 explicitly says that meaning is defined in absolute terms, not by similarity to other Φ-structures, and then allows a transformation or 'translation' that maps radically different structures to one another. This suggests that successful communication may hinge on an arbitrary mapping rather than on intrinsic similarity. Thus the 'it follows' in the abstract is not derived from IIT's axioms and postulates; it is a new, extra-theoretical definition. A concrete metric could rescue the claim, but none is supplied.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper contrasts Shannon information theory with integrated information theory (IIT), arguing that Shannon information concerns messages, codes, and extrinsic transmission, whereas IIT treats meaning as an intrinsically defined cause–effect structure (Φ-structure) unfolded from a conscious complex. The authors introduce the notion of a Φ-fold as a sub-structure corresponding to a content of experience, and claim that communication of meaning occurs when a source Φ-fold triggers a similar Φ-fold in a target complex. The paper reviews IIT's axioms and postulates, compares the two frameworks in a table, discusses implications for neuroscience and artificial systems, and includes two supplementary sections on information theory and IIT.","tokens_in":17905,"tokens_out":4489,"duration_ms":38134,"significance":"If the central claim held, the paper would provide a principled distinction between information as message transmission and meaning as consciousness-dependent structure, and would ground a research program for measuring meaning transfer. The paper is clearly written and gives a useful side-by-side comparison of Shannon information and IIT, and it is transparent about the major open problem, namely the absence of a concrete similarity measure for Φ-structures (footnote 14). However, the communication-of-meaning claim is not derived from IIT's axioms; it is an additional definitional step. The paper would be significant as a conceptual proposal, but as it stands the key notion of similarity remains unformalized, so the central conclusion is not yet a testable result. The paper also credits prior and forthcoming work appropriately, and it explicitly acknowledges the computational infeasibility of unfolding large Φ-structures, which is a fair limitation.","major_comments":[{"comment":"The paper's central claim, stated in the abstract and repeated in this section, is that 'for the communication of meaning, the cause–effect structures of sender and receiver must be similar.' However, no formal definition of 'similar' is provided. The text says 'the similarity of the two Φ-folds could be used to measure the amount of meaning communicated,' and footnote 14 concedes that developing ways to compare them optimally is 'not easy.' This is not a minor technicality: without a metric or an equivalence relation on Φ-structures, the claim cannot be evaluated as true or false, and the amount of meaning communicated is undefined. The paper should either supply a candidate similarity measure (or a set of necessary conditions) or explicitly present the communication claim as a conjecture rather than a consequence of IIT.","section":"Communicating integrated information as meaning (Fig. 6 and following text)"},{"comment":"Footnote 15 creates an internal tension with the main claim. It states that 'in IIT's analysis, the meaning of an experience is defined in absolute terms by the distinctions and relations that compose a specific Φ-structure, rather than by its similarity to other Φ-structures,' and it allows for a 'translation' mapping between radically different Φ-structures. If meaning is absolute, then similarity between Φ-structures is an extrinsic relation imposed by the analyst; it is not part of IIT's fundamental explanatory identity ('with no additional ingredients'). The 'it follows' in the abstract therefore does not follow from the axioms and postulates. The paper should clarify how the similarity relation is grounded in IIT or acknowledge that it is a new, extra-theoretical definition.","section":"Communicating integrated information as meaning, footnote 15"},{"comment":"The identification of meaning with the Φ-structure is asserted as IIT's fundamental explanatory identity, not independently demonstrated. The paper states that the Φ-structure 'fully characterizes, with no additional ingredients, how an experience feels—its quality—which defines its intrinsic meaning.' For a paper whose central contribution is a consequence of this identity, this is acceptable as a starting point, but the paper should more clearly separate (i) the IIT-internal definition of meaning from (ii) the new claim about communication of meaning. In particular, the transition from 'meaning is the Φ-structure' to 'communicating meaning requires similar Φ-structures' involves an additional premise about how meaning transfers between systems, which is not part of IIT's postulates. The paper should state this premise explicitly.","section":"Integrated information theory and Information processing vs. information structures"}],"minor_comments":[{"comment":"The notion of a 'Φ-fold' is used throughout the main text but only informally characterized in the supplementary material; a precise definition (e.g., a sub-structure formed by a subset of distinctions and relations) would improve clarity.","section":"Supplementary material II"},{"comment":"The capitalization in the final row is inconsistent: 'Meaning Must be provided extrinsically' should be 'Meaning must be provided extrinsically.'","section":"Table I"},{"comment":"The reference list includes several items marked 'forthcoming' (refs. 13, 14, 52) and an arXiv manuscript (ref. 11) without publication details; please provide updated versions or URLs where possible.","section":"References"},{"comment":"The displayed equation for Shannon entropy appears to be missing the summation sign in the provided text; please ensure all equations are typeset correctly.","section":"Supplementary material I"},{"comment":"Panel (c) of Figure 6 refers to 'perceptual richness' without defining the term in the caption; a brief definition or a pointer to Supplementary material II would help.","section":"Figure 6"},{"comment":"The statement 'information processing can only ever decrease the information transmitted over a channel' is too broad as written; the data-processing inequality applies to specific Markov chains and should be qualified accordingly.","section":"Information processing vs. information structures"}],"recommendation":"major_revision","confidential_remarks":"This is a conceptual essay rather than a technical result. The main contribution is expository, and the new claim about communication is under-specified. The paper would be strengthened by either a formal similarity measure or an explicit downgrade of the claim to a hypothesis. I see no grounds for rejection, but the current version overstates the deductive strength of the conclusion. The paper also relies heavily on prior and forthcoming work by the same group; this is appropriate for a framework paper but should be made clear to readers."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear Colleague,\n\nThe paper is a clearly written position piece that maps IIT's notion of meaning as Φ-structure onto the problem of communication. What's new is the explicit claim that communicating meaning requires the sender's and receiver's Φ-folds to be similar, and the contrast with Shannon information is well drawn. It does a good job of explaining why Shannon information is silent about meaning and why, within IIT, meaning is intrinsic and absolute.\n\nThe soft spot is exactly where the reader and stress-test put their fingers. The phrase \"it follows\" in the abstract is not a consequence of IIT's axioms. IIT defines an experience's meaning intrinsically, by the Φ-structure, with no reference to other structures. Similarity between two Φ-structures is an additional, extrinsic relation. Footnote 15 concedes that meaning is defined in absolute terms, and even allows a \"translation\" mapping between radically different structures. That makes the communication claim a proposed research agenda, not a derived result. And as footnote 14 admits, there is no formal measure of structural similarity on offer. Without such a measure, \"amount of meaning communicated\" is not an operational quantity.\n\nThat said, the paper is honest about these gaps, and the conceptual framing is coherent. It will be useful to people who work on IIT, consciousness, and the philosophy of communication. It does not need to be dismissed; it needs to be read as a programmatic statement, not a theorem.\n\nIf I were editing, I would send it to peer review as a perspective article, but I would ask the authors to either (a) provide a working definition of Φ-fold similarity, even a toy example, or (b) explicitly frame the communication claim as a hypothesis requiring future formal work. The paper is more of an invitation than a destination, but it's a thoughtful invitation.\n\nBest.","headline":"A clear IIT-based position paper on meaning and communication, undermined only by the undefined notion of structural similarity that carries the central claim.","tokens_in":18473,"tokens_out":1957,"would_cite":false,"duration_ms":18692,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper argues that the meaning of an experience is the cause–effect structure a conscious system unfolds from its own activity, so meaning is communicated only when sender and receiver unfold similar structures.","keywords":["integrated information theory","Φ-structure","intrinsic meaning","Shannon information","communication of meaning","consciousness","cause–effect structure"],"falsifier":"A direct test would compare two conscious subjects with deliberately different causal architectures: if they can be shown to share the same specific conscious content while their unfolded Φ-folds differ substantially under the IIT formalism, the identity between meaning and Φ-structure fails. Concretely, one could compute Φ-fold similarity for subjects with different connectomes and measure whether their success in communicating a specific percept tracks Φ-fold similarity better than it tracks Shannon mutual information of the transmitted symbols.","tokens_in":17450,"feed_emoji":"🧠","tokens_out":6723,"duration_ms":52713,"temperature":0.7,"pith_summary":"Where does the meaning of a message live? This paper argues that Shannon information theory, which measures messages in bits and brackets meaning by design, cannot locate it: meaning is not a property of symbols, codes, or computations. The paper's central claim is that meaning is identical to the cause–effect structure—called the Φ-structure—that integrated information theory (IIT) unfolds from a conscious system in a state. If that claim is right, then Alice's experience of a house on fire is the structure of causal distinctions and relations her brain specifies at that moment, and Bob understands her only when her message triggers a structurally similar Φ-structure in his brain. The upshot is that communication of meaning is intrinsic, approximate, and never fully shared.","feed_headline":"Meaning isn't in the message—it's in the mind's causal shape","feed_subtitle":"Meaning, the paper argues, is the brain's cause–effect structure—communication needs matching structures.","key_machinery":"The central object is the Φ-structure (cause–effect structure): the set of causal distinctions specified by subsets of units within a complex, together with the relations that bind them, determined from the system's transition probability matrix and its current state. It carries the argument by replacing the picture of meaning as a code with a picture of meaning as a structure: meaning is not read off a symbol but unfolded from a substrate, and communication of meaning becomes a matter of structural similarity between the Φ-folds of sender and receiver.","core_discovery":"The paper contrasts two axiomatic uses of probability: Shannon's extrinsic account of messages and IIT's intrinsic account of meaning. It claims that the Φ-structure unfolded from a complex—a maximally irreducible substrate—in its current state fully characterizes, with no additional ingredients, how an experience feels and what it intrinsically means. Within the Φ-structure, each content of experience is a Φ-fold, a sub-structure of distinctions and relations; the feeling of seeing the house on fire is that Φ-fold. Communication of meaning therefore requires that a source complex's Φ-fold trigger a similar Φ-fold in a target complex. Symbol transmission is neither sufficient (a perfect channel can fail to trigger the intended fold) nor necessary (without similar folds, no meaning is carried), and the amount of meaning communicated could in principle be measured by the similarity of source and target Φ-folds.","pith_inferences":["If meaning is identical to Φ-structure, then the claim that digital computers have no intrinsic meaning is an empirical bet about their wiring, not a consequence of the identity itself; a silicon substrate that could support a high-Φ complex would, by the paper's own logic, have meaning.","The framework suggests a testable research program: approximate Φ-fold similarity with computable measures and check whether subjective reports of 'getting it' track structural similarity better than mutual information of neural codes.","It also implies a strong form of untranslatability: since meanings are absolute structures rather than positions in a semantic space, translation can only maximize overlap between Φ-folds, and some meanings may be untranslatable."],"forward_implications":["Standard digital computers, whose wiring cannot support high-Φ complexes, can transmit messages perfectly while communicating no intrinsic meaning; their utterances mean something only when a conscious being interprets them.","Because the wiring of each brain differs through development and learning, shared meanings are never identical: communication of meaning is always approximate, and some meanings will be idiosyncratic.","Shannon information and integrated information are complementary rather than competing: one quantifies symbol transmission, the other quantifies meaning, and a full account of communication needs both.","The amount of meaning successfully communicated can in principle be measured by the similarity of the source and target Φ-folds, not by bit rate, channel capacity, or perceptual richness."],"supporting_citations":[{"why":"Supplies the message/probability definition of information that the paper contrasts with meaning.","marker":"[4]"},{"why":"Provides the IIT 4.0 formalism and postulates used to unfold Φ-structures from a substrate.","marker":"[5]"},{"why":"Defines intrinsic meaning, Φ-folds, perceptual richness, and matching, which the communication-of-meaning argument relies on.","marker":"[11]"},{"why":"Gives the substrate account of why some brain regions support consciousness while others do not.","marker":"[12]"},{"why":"Argues that digital computers cannot support high-Φ complexes, grounding the paper's claim that AI has no intrinsic meaning.","marker":"[19]"},{"why":"Shows how a phenomenal content such as spatial experience is unfolded as a structure, exemplifying the Φ-structure account.","marker":"[15]"}],"fun_headline_variants":["Meaning is structure, not signal","Consciousness: meaning as integrated cause-effect structure","IIT: meaning in causal shape, not in bits","Meaning isn't sent; it's matched by structure","From messages to cause-effect structures: meaning in IIT"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that Φ-structures from different substrates can be compared by a well-defined similarity measure; the paper itself notes that this is not easy, since it requires unfolding very large structures and finding optimal ways to compare them. Without such a measure, the claim that meaning is communicated when source and target Φ-folds are similar is not yet well-defined.","fun_headline_variants_meta":{"raw":{"variants":["Meaning is structure, not signal","Consciousness: meaning as integrated cause-effect structure","IIT: meaning in causal shape, not in bits","Meaning isn't sent; it's matched by structure","From messages to cause-effect structures: meaning in IIT"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000187,"raw_usage":{"total_tokens":1290,"prompt_tokens":865,"completion_tokens":425,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":481,"completion_tokens_details":{"reasoning_tokens":361}},"tokens_in":481,"tokens_out":425,"duration_ms":4156,"temperature":1.0,"reasoning_tokens":361,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T15:45:14.366198+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct test would compare two conscious subjects with deliberately different causal architectures: if they can be shown to share the same specific conscious content while their unfolded Φ-folds differ substantially under the IIT formalism, the identity between meaning and Φ-structure fails. Concretely, one could compute Φ-fold similarity for subjects with different connectomes and measure whether their success in communicating a specific percept tracks Φ-fold similarity better than it tracks Shannon mutual information of the transmitted symbols.","supporting_citations":[{"cited_title":"Nat Rev Neurosci, 2016","cited_arxiv_id":null,"evidence_quote":"Gives the substrate account of why some brain regions support consciousness while others do not."},{"cited_title":"Juel, and G","cited_arxiv_id":null,"evidence_quote":"Defines intrinsic meaning, Φ-folds, perceptual richness, and matching, which the communication-of-meaning argument relies on."},{"cited_title":"Bell System Technical Journal, 1948","cited_arxiv_id":null,"evidence_quote":"Supplies the message/probability definition of information that the paper contrasts with meaning."},{"cited_title":"PLoS Comput Biol, 2023","cited_arxiv_id":null,"evidence_quote":"Provides the IIT 4.0 formalism and postulates used to unfold Φ-structures from a substrate."},{"cited_title":"Towards Conscious AI Systems Symposium - Proceedings of the AAAI 2019, 2019","cited_arxiv_id":null,"evidence_quote":"Argues that digital computers cannot support high-Φ complexes, grounding the paper's claim that AI has no intrinsic meaning."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows how a phenomenal content such as spatial experience is unfolded as a structure, exemplifying the Φ-structure account."}],"review_version":1}