Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision		Previous revision
bib:albantakis2023computing [2026/01/28 10:47] – [3.7 Plain-language translation (no IIT terms) with the full math pipeline] kymkibib:albantakis2023computing [2026/05/21 06:29] (current) – [Steelman] kymki
Line 17: Line 17:
  
 ===== BibTeX ===== ===== BibTeX =====
-<code bibtex>+<code javascript>
 @article{albantakis2023computing, @article{albantakis2023computing,
   title={Computing the integrated information of a quantum mechanism},   title={Computing the integrated information of a quantum mechanism},
Line 29: Line 29:
 } }
 </code> </code>
 +====== Albantakis et al. (2023) — QM re-formulation and failure modes ======
  
-===== One-paragraph summary (plain language) ===== +Im reading this with minimal background in IIT and most of its concepts are unknown to me. The reading is based on trying to extract what QIT the authors lean on to understand if what they propose makes "physical sense".
-The paper proposes a method (QIIT/QIIT-like) to compute “integrated information” for quantum systems by defining cause/effect “repertoires” using density matrices and comparing them to a baseline “chance” state. It adapts classical IIT machinery (mechanism/purview partitions, counterfactual noise injection, integration via partitioning) into a quantum formalism. +
- +
-===== Review ===== +
- +
-==== What it claims to do (in operational terms) ==== +
-  * Defines quantum cause/effect repertoires for a chosen “mechanism” and “purview”. +
-  * Defines an “intrinsic difference” measure (QID) between a constrained repertoire and a baseline repertoire. +
-  * Defines “integration” by how much this measure drops under partitions (analogous to classical IIT). +
- +
-==== What is genuinely new (vs renaming) ==== +
-  * A concrete proposal for handling entanglement when factorizing repertoires (nontrivial compared to classical product-factor approaches). +
-  * A specific divergence-like measure (QID) intended to encode “intrinsic” rather than channel-designer information. +
- +
-==== What looks like repackaging / relabeling ==== +
-  * “Mechanism” ≈ chosen subsystem. +
-  * “Purview” ≈ another chosen subsystem. +
-  * “Repertoire” ≈ a (conditional / counterfactual) reduced density matrix. +
-  * The conceptual novelty is not in the quantum objects, but in the *interpretation* (intrinsic/self-specifying) and the *intervention rule*. +
- +
-==== Core conceptual friction points (physics-first critique) ==== +
-  - **Counterfactual noise injection:** The method “disconnects” everything outside the mechanism by replacing it with maximally mixed noise. This is not a physical open-system approximation; it is a *chosen intervention rule*. Any “intrinsic” claims depend on this convention. +
-  - **Factorization dependence:** Results depend on the choice of subsystem decomposition (“units”). In quantum theory, factorization is not always unique or physically privileged; if the computed structure changes under refactorization, it’s hard to call it intrinsic. +
-  - **Unitary bias / measurement gap:** The clean formalism largely lives in unitary evolution; measurement/non-unitary updates create ambiguity for “cause” directionality and can become interpretation-dependent. +
-  - **Mixed-state ambiguity:** A density matrix can represent ignorance (epistemic) or a reduced state from entanglement (ontic-but-subsystem). The framework’s language often slides between these readings. +
- +
-==== Where the prose risks misleading the reader ==== +
-  * Phrases like “the system knows” or “specifies information about itself” read like ontology, but the actual operations are: choose a partition, apply an intervention/noise rule, compute a state, compute a divergence, pick a maximizing element. +
-  * The mathematical pipeline can be valid as a *defined metric*, but the paper’s language can make it sound like a derived physical necessity. +
- +
-==== Strongest charitable reading ==== +
-The framework is a proposed *measure of “how concentrated and partition-resistant” a mechanism’s counterfactually-defined influence is* under a particular intervention scheme. It is a formal extension of IIT-style attribution to density matrices. +
- +
-==== Strongest skeptical reading ==== +
-It is a rebranding of subsystem/channel calculations with a heavy interpretive layer. “Intrinsic” properties are not shown to be invariant under factorization, interpretation of measurement, or physically constrained interventions—so the ontological talk outruns what the formalism guarantees. +
- +
-===== Notes / excerpts ===== +
-  * (Add your own quotes here as you read.) +
-  * (Add page/section pointers you want to revisit.) +
- +
-===== Open questions to test the framework ===== +
-  - If you compute the quantity under two physically equivalent descriptions (different tensor factorizations / dilations), do you get the same “intrinsic” structure? +
-  - If you replace “maximally mixed noise” with a physically motivated environment state (thermal, constrained by energy), how stable are the results? +
-  - Does QID overemphasize top-eigenvalue behavior in ways that wash out phase-sensitive/coherence structure you’d expect to matter? +
- +
-====== Albantakis et al. (2023) — QM re-formulation and failure modes ======+
  
-This note rewrites the paper’s core definitions in standard quantum information language, and isolates where the framework is (i) pure relabeling, (ii) a specific intervention convention, and (iii) where it becomes physically ambiguous.+This note rewrites the paper’s core definitions in standard quantum information language, and tries to explain conclusions from that standpoint. The case may be that I completely miss subtleties due to not being well read in IIT.
 ===== 1. Translation into standard QM language ===== ===== 1. Translation into standard QM language =====
  
Line 273: Line 229:
 ==== 3.5 The score $\varphi(m,Z,\theta)$ ==== ==== 3.5 The score $\varphi(m,Z,\theta)$ ====
  
-They define an “integrated information for this purview under this partition” as a divergence between:+Let (Q) be factorized and let 
 +$$ 
 +T:\mathcal B(\mathcal H_Q)\to\mathcal B(\mathcal H_Q))  
 +$$ 
 +be the global channel. For a chosen **input subsystem** (M) prepared in state $\rho_M$, define the **intervention-induced effective channel** 
 +$$
  
-  - the full repertoire $\pi_e(Z\mid m)$, and +\mathcal E_{M\to Z}(X):=\operatorname{tr}*{Z^0}!\bigl(T(X\otimes \tau*{M^0})\bigr), 
-  - the cut repertoire $\pi_e^{\theta}(Z\mid m)$,+\qquad \tau_{M^0}:=\frac{I_{M^0}}{d_{M^0}},
  
-using their QID: 
 $$ $$
-\varphi(m,Z,\theta) +and the corresponding **output marginal** on the chosen **target subsystem** (Z):
-\;\equiv\; +
-\mathrm{QID}\Bigl(\pi_e(Z\mid m)\;\Vert\;\pi_e^{\theta}(Z\mid m)\Bigr).+
 $$ $$
  
-Crucial detail: with their maximally mixed baseline choice, QID behaves like a **max-eigenvalue-weighted** distinguishability, and the “intrinsic effect” is the eigenvector of $\pi_e(Z\mid m)$ with the largest eigenvalue. In practice this means the comparison is dominated by the top-eigenvalue eigenspace rather than the full density matrix structure.+\rho_Z:=\mathcal E_{M\to Z}(\rho_M).
  
-(Their Eq. (31) writes this out in components and evaluates it at the maximizing eigenstate.)+$$
  
-----+For a partition $\theta\in\Theta(M,Z)$, construct the paper’s **cut / partitioned output** $\sigma_Z^{(\theta)}$ by (i) injecting maximally mixed noise across the cut and (ii) taking a tensor product across the partition blocks (optionally after the paper’s ($P^*$) “entanglement cluster” factorization step for $|Z|>1)$. Denote the resulting state by 
 +$$ 
 + 
 +\sigma_Z^{(\theta)}:=\pi^{\theta}_e(Z\mid m)\quad\text{(paper’s notation)}. 
 + 
 +$$ 
 + 
 +Now diagonalize 
 +$$ 
 + 
 +\rho_Z=\sum_i p_i|i\rangle\langle i|,\qquad 
 +\sigma_Z^{(\theta)}=\sum_j q_j|j\rangle\langle j|, 
 +\qquad 
 +P_{ij}:=|\langle i|j\rangle|^2. 
 + 
 +$$ 
 +**Intrinsic (selected) effect state.** The paper first selects a particular eigenstate $|i^*\rangle$ of $\rho_Z$ via an “intrinsic information vs baseline” optimization. With their maximally mixed baseline on $Z$, this selection reduces to choosing the **largest-eigenvalue eigenvector** of $\rho_Z$ (or the corresponding eigenspace if degenerate). Call that chosen eigenstate 
 +$$ 
 + 
 +z'_e(m,Z):=|i^*\rangle. 
 + 
 +$$ 
 +**Integrated effect information for a partition.** The paper then evaluates its QID expression **at that chosen eigenstate** (not by maximizing over (i) again): 
 +$$ 
 + 
 +\phi_e(m,Z,\theta) 
 +:=\phi(m,z'_e,\theta) 
 += 
 + 
 +p_{i^*}\left(\log p_{i^*}-\sum_j P_{i^*j}\log q_j\right). 
 + 
 +$$ 
 +Equivalently, define the geometric mean 
 +$$ 
 + 
 +\bar q_{i^*}:=\exp!\left(\sum_j P_{i^*j}\log q_j\right), 
 + 
 +$$ 
 +so that 
 +$$ 
 + 
 +\phi_e(m,Z,\theta)=p_{i^*}\log!\frac{p_{i^*}}{\bar q_{i^*}}. 
 + 
 +$$
  
 ==== 3.6 Optimization: MIP and “best” purview ==== ==== 3.6 Optimization: MIP and “best” purview ====
  
-For each subset $Z$, they define a “minimum information partition” (MIP) as the partition that minimizes $\varphi$:+For fixed $M,\rho_M$ and candidate target $Z$, the paper chooses the minimum information partition (MIP) using a normalized criterion:
 $$ $$
-\theta^{\star}(m,Z) +\theta'(m,Z) 
-\;:=\; += 
-\arg\min_{\theta}\;\varphi(m,Z,\theta).+\arg\min_{\theta\in\Theta(M,Z)} 
 +\frac{\phi_e(m,Z,\theta)}{\max_{T'*S}\phi_e(m,Z,\theta)}.
 $$ $$
 +Here $\max*{T'_S}$ ranges over alternative systems/channels of the same dimensions; this normalization is part of the selection of $\theta$.
  
-Then they define the integrated information for the mechanism (for effects) by maximizing over purviews:+Then the (reported) integrated effect information for that $Z$ is
 $$ $$
-\varphi_e(m) +\phi_e(m,Z):=\phi_e(m,Z,\theta'(m,Z)).
-\;:=\; +
-\max_{Z\subseteq Q}\;\min_{\theta}\;\varphi(m,Z,\theta).+
 $$ $$
  
-(They do an analogous construction for causes and combine them in the full IIT-style definition.)+Finally, the best target subsystem (or in paper terms: “maximally irreducible effect purview” - not sure why it is irreducible) is chosen by 
 +$$ 
 +Z^**e(m)=\arg\max*{Z\subseteq Q}\phi_e(m,Z), 
 +\qquad 
 +\phi_e(m)=\max_{Z\subseteq Q}\phi_e(m,Z). 
 +$$
  
----- 
  
 ==== 3.7 Mathematics Framework ==== ==== 3.7 Mathematics Framework ====
Line 551: Line 556:
  
 ==== Steelman ==== ==== Steelman ====
-As a *defined* counterfactual causal attribution scheme, the framework:+As a defined counterfactual causal attribution scheme, the framework:
   * avoids common-cause “spurious correlation” effects by construction (their COPY-XOR/CNOT motivation),    * avoids common-cause “spurious correlation” effects by construction (their COPY-XOR/CNOT motivation), 
   * tries to preserve entanglement-generated correlations by clustering entanglement before taking products,   * tries to preserve entanglement-generated correlations by clustering entanglement before taking products,