Chu's exact quantum opinion model diverges numerically from Friedkin–Johnsen, but only on synthetic six-node toy networks

Weiqi Chu's 2026 quantum opinion-dynamics preprint (arXiv:2607.01452) simulates the full, non-approximated Lindblad master equation on four synthetic six-agent networks — complete, chain, ring, and Barbell graphs — and compares the trajectories against the classical Friedkin–Johnsen (FJ) reduction run on the same graphs. The two models "follow similar trends" but "do not agree exactly." At steady state, the exact quantum model produces consistently smaller opinion variance than FJ on all four networks — the paper attributes this to inter-agent correlations generated by the model's two-body "jump operators," which act as "an additional consensus-promoting mechanism" beyond what the product-state approximation predicts.

This is the paper's one clean piece of evidence that the quantum layer is not merely FJ with extra parameters: it produces a specific, directional, numerically demonstrated divergence. But the demonstration is confined to n = 6 toy graphs — three orders of magnitude smaller than the one real-world network the paper tests (see claim-chu-facebook100-run-uses-fj-equivalent-approximation). Whether the same divergence survives at real-world scale is untested, since simulating the exact (non-approximated) dynamics is computationally hard — the state space scales as 2^n.