d4-journals-mss-2016-0028

Theorem 4.1 classifies outcomes of a finite signed network but does not state a computational result. The proposed population-level classification is a plausible follow-on, yet it cannot supply the missing computational anchor.

The objection that survived

The opponent observes that the outcome categories depend on signed-graph structure, while the proposed type masses do no work in that classification.

What the mirror covers

The proposed extension covers only Theorem 4.1’s convergence and polarization trichotomy; it leaves Theorems 3.1 and 4.2, Proposition 3.1, and the discussion of wisdom aside.

Open questions for a prover

The case FOR (proponent)

The most credible positive foothold is a follow-on about large societies with a small number of recurring social roles. A type would specify an agent’s faction and their normalized influence on every type; \(\mu_t\) would be that type’s share of society. For example, many people in a political community might share one of a few faction and interaction profiles. This is a plausible high-multiplicity regime, though the assumption that people of a type share the same interaction profile is strong.

A precise candidate is Mass-Signed-Opinion Classification. Its input is a finite type set \(T\), rational masses \(\mu_t>0\) summing to one, a rational row-stochastic interaction matrix \(P\) reversible with respect to \(\mu\), and symmetric signs \(\eta_{ts}\in\{-1,+1\}\) on its positive entries. Assume the interaction graph is connected and aperiodic. The type-level opinion evolves by
\[ x_t^{(r+1)}=\sum_{s\in T}P_{ts}\eta_{ts}x_s^{(r)}. \]
The question is which outcome the dynamics induce for all initial type-opinion profiles: nonzero polarization, nonconvergence for some initial profile, or convergence to the neutral zero consensus. The expected answer is tractability: test the signed type graph for opposition bipartiteness and reverse opposition bipartiteness. This is a type-mass version of the paper’s structured signed-network setting, not a claim that the paper itself gives a complexity bound.

The closest named result is Theorem 4.1, proved in this paper. For a strongly connected, aperiodic simple symmetric network with no self-loops and soft opposition \(\mathfrak D(x)=-x\), it says: the network induces a polarization that is not always zero iff it is opposition bipartite; it is divergent iff it is reverse opposition bipartite; and it induces the neutral zero consensus iff it is neither. That is a natural analytical result to extend to the type-mass model.

But it is not an admissible computational anchor under the ChoCo criteria: it classifies long-run dynamics, without asserting a complexity result for a computational problem. The paper has no named result asserting polynomial time, hardness, or another computational classification. Its “continuous model” makes the opinion spectrum continuous while the population remains the finite set \([n]\). So the candidate above is an unanchored research question, not a valid per-anchor mirror. The weakness is decisive for the requested case: the paper offers a plausible population-level extension, but no computational result for that extension to mirror.

The case AGAINST (opponent, writing after the proponent)

The strongest objection is that this paper has no named computational result to mirror. Theorem 4.1 classifies the long-run behavior of a fixed finite network; Theorems 3.1 and 4.2 likewise give structural or analytic characterizations. None asks for an algorithm, establishes a complexity bound, or studies an optimization problem. The paper’s “continuous model” makes opinions continuous while keeping the society finite. That is not a continuous-population treatment, and it does not fill the computational gap ChoCo targets.

Mass-Signed-Opinion Classification is a coherent new problem, but it is supplied by the proponent rather than anchored in the paper. In that model, types are network roles and the type-level dynamics can be sensible. But the proposed output—polarization, divergence, or zero consensus—is precisely the classification Theorem 4.1 already gives for a signed network. The type masses do no work in that classification under the stated formulation: the outcome conditions concern the signed graph. Making masses affect the interaction kernel would strengthen the population model, but it would still turn an analytic theorem into a newly posed computational task, rather than mirror a computational result in the paper.

The stronger alternatives have the same gap. A type-level version of Theorem 4.2 could ask for influence by role, but the paper gives an eigenvector formula, not a complexity result. Optimization over changes to the network could create a substantial computational project, but that adds an intervention objective the paper never studies. So no anchor raised here survives as a ChoCo mirror. The honest limit of this negative case is that a worthwhile follow-on could be built around these models; the paper itself does not supply the computational result that would make it a mirror.

The adversarial triple: the proponent anchors on up to three named results; the opponent sees that case and must defeat every anchor; the judge decides which case convinced it. These are the pipeline’s own outputs, generated by tools/triple_run.py — no human edited them. The paper’s own text is not reproduced here beyond the quoted statement above.