IIT Expert

Precomputed result

The six-unit copy cycle of the IIT 4.0 paper is one complex with only first-order distinctions

Substrate Six binary units A–F with states in {−1, 1}, each a logistic of its weighted inputs (Eq 60) with k = 4, connected in a directed cycle A→B→C→D→E→F→A with weight 1.0 and no other connections. Each unit copies the state of its predecessor with some indeterminism. The paper's text gives these weights exactly; PyPhi ships the substrate as pyphi.examples.iit4_2023_fig6c_substrate.
State (1, −1, −1, −1, −1, −1) — Abcdef, unit A ON
φs 0.15710886599709775
Minimum partition A directed partition that severs two of the six cycle connections, D→E and F→A. The same under both formulations.
Formalism iit-4.0-2026
PyPhi 2.0.0rc4
Script albantakis-2023-fig6c-copy-cycle.py

Albantakis et al. (2023) use the copy cycle to show that a large complex can specify a poor Φ-structure. Without cross-connections, only single units specify distinctions, so the structure has few relations and low Φ.

quantity Fig 6C 2023 2026
complex Abcdef Abcdef Abcdef
φs 1.74 1.7467 0.1571
distinctions 6 of 63 6 6
Σφd 5.7377 5.7378 5.7378
relations 4 4 4
Σφr 1.9125 1.9126 1.9126
Φ 7.65 7.6504 7.6504

Each unit specifies its predecessor as cause and its successor as effect, with φd = 0.9563: A → F | B, b → a | c, and so on around the cycle, as in the figure. No mechanism of two or more units is irreducible.

The Φ-structure is the same under both formulations. φs is not. The paper states that the copy cycle "constitutes a 6-unit complex with a maximal φs = 1.74 ibits" and that highly deterministic directed cycles "can easily be extended to constitute large complexes". Under the current formulation the cycle is still the only complex, but its φs falls to 0.16. The current formulation sets φs = min{φc, φe, ii}, where ii includes the intrinsic differentiation of the system's cause and effect (Mayner et al. 2026, §2.3–2.4). With k = 4 and weight 1.0, each unit copies its predecessor with probability 1/(1 + e⁻⁴) = 0.982 (Eq 60), so the specified six-unit effect state has probability 0.982⁶ = 0.8968. Its intrinsic differentiation, −log₂ 0.8968 = 0.1571, is the minimum; the cause side gives the same value. The script prints these terms. φc and φe at the minimum partition are both 1.7467, the 2023 value. The paper's remark that lengthening the cycle "would not change the amount of φs" is a claim about the 2023 formulation and has not been recomputed here.

The script uses the default configuration of each formulation. To reproduce, run the script with uv, which installs PyPhi 2.0.0rc4 for the run:

uv run --with "pyphi==2.0.0rc4" python canon/albantakis-2023-fig6c-copy-cycle.py

Sources