| paper | d4-journals-geb-2013-0046 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | medium |
| authors would recognise it | yes |
Corollary 6
statement extracted from the paper’s text layer
Given rational type masses \((\mu_t)_{t\in T}\) summing to \(1\), rational piecewise-linear nondecreasing update rules \(F_t:[0,1]\to[0,1]\) with \(F_t(0)=0\) and \(F_t(1)=1\), and a rational initial yes fraction \(q_0\), consider each population size \(N\) for which all \(N\mu_t\) and \(Nq_0\) are integers. Start with \(Nq_0\) yes opinions; when the current yes count is \(k\), each type-\(t\) agent independently says yes at the next step with probability \(F_t(k/N)\). Let \(\alpha_N\) be the probability of eventually reaching unanimous yes. Does \(\lim_{N\to\infty}\alpha_N\) exist along these population sizes, and, if so, what is its exact value?
The society is a distribution \(\mu\) over finitely many update-rule types \(t\), where type \(t\) has mass \(\mu_t\) and responds to a yes fraction \(q\) with probability \(F_t(q)\). The state of the mass model is the yes fraction \(q\); its finite-population approximations use \(N\mu_t\) agents of each type and independent updates. The quantity of interest is the limit of the finite approximations’ probabilities of eventual unanimous yes.
The order of limits leaves \(\lim_{N\to\infty}\alpha_N\) sensitive to rare finite-population fluctuations over an unbounded absorption horizon, so it may differ from absorption in deterministic mass dynamics; the proponent flags this distinction but does not resolve it.
fatal: False
The mirror covers Corollary 6’s absorption probabilities under large-population scaling. It leaves the finite-population terminal-class characterizations, including Theorem 1, the speed-of-convergence formulas, and the quantifier consensus results as standalone results.
My strongest case is a large-population version of the paper’s absorption-probability computation. The computational anchor is Corollary 6, proved in this paper. It says that in an anonymous model, starting from a transient state of size \(s\), the probability of absorption into each terminal class is given by the corresponding entry of
\[
(I-Q^a)^{-1}\widetilde D^a,
\]
where \(Q^a\) is the transient part of the reduced anonymous transition matrix and \(\widetilde D^a\) gives transition probabilities from transient states into terminal classes. The paper’s anonymous reduction uses a matrix with only \(n+1\) size states. With an explicit rational input, that formula gives an exact computation by rational linear algebra; the polynomial-time interpretation is mine, not a complexity classification stated by the authors.
A natural continuous society has finitely many update-rule types \(t\), each with mass \(\mu_t\), where \(\sum_t\mu_t=1\). A type is its complete rule for responding to the current fraction of “yes” opinions: a rational, piecewise-linear, nondecreasing quantifier \(F_t:[0,1]\to[0,1]\), with \(F_t(0)=0\) and \(F_t(1)=1\). For example, \(F_t(q)\) might specify how likely a member of that type is to say “yes” when fraction \(q\) currently do so. A plausible regime is a very large online community discussing one issue, with many users but relatively few shared response rules. The paper’s conclusion about extended OWA operators, which allow the rule to vary with population size, gives this scenario a direct connection to its model.
My lead mirror is the following limit problem. The input is \((\mu_t,F_t)_{t\in T}\), an initial “yes” fraction \(q_0\), and their rational encodings. For each population size \(N\) divisible by the input denominators, instantiate \(N\mu_t\) agents of type \(t\), initially with \(Nq_0\) “yes” opinions. If the current count is \(k\), every type-\(t\) agent independently says “yes” at the next step with probability \(F_t(k/N)\). Let \(\alpha_N\) be the probability that this finite process eventually reaches unanimous “yes.” The continuous problem is: does \(\lim_{N\to\infty}\alpha_N\) exist, and, if so, what is its value? A solution reports the exact limit or proves nonconvergence. Natural complexity parameters include the number of types, the number of linear pieces, and the input bit length.
This is a plausible mirror of the paper’s absorption calculation: the population is the distribution over update-rule types, while the measured outcome is the large-population limit of a quantity the authors already compute. The paper’s Theorem 1, proved here, supplies related structural context: it characterizes when an OWA model has no terminal classes beyond the two consensus states. That holds exactly when there is a threshold \(\bar k\) such that, for every \(k\ge\bar k\), \(k+1\) distinct agents can each say “yes” with positive probability when \(k\) agents currently say “yes,” and for every \(k<\bar k\), \(n-k+1\) distinct agents can each say “no” with positive probability when \(n-k\) currently say “yes.” The theorem concerns finite-process convergence, rather than the complexity of computing it.
I would expect the limit problem to be continuum-specifically hard, though that is a research hypothesis, not a result of the paper. For any fixed number of steps, the population’s yes fraction follows the deterministic map \(q\mapsto\sum_t\mu_tF_t(q)\). But eventual absorption takes place over an unbounded time horizon: rare finite-population fluctuations can affect which absorbing outcome is reached, even as those fluctuations disappear at each fixed time in the continuum limit. A central follow-up is whether that limit can instead be computed tractably from the deterministic map and the fluctuations around it. Another is whether the limit exists for every rational instance.
The soft spot is the order of limits. This proposal takes finite stochastic processes to absorption first, then sends population size to infinity. The result need not equal the outcome of simply replacing the finite society with deterministic mass dynamics. That distinction is motivated by the paper’s own focus on absorption probabilities, but the authors would have to accept the limit question as a useful extension. Also, the paper does not state a standard complexity-class result: Corollary 6 is an exact computation formula, not a theorem explicitly classifying a problem as being in P. If the required anchor must itself state such a complexity classification, this paper has no qualifying anchor.
Corollary 6 is the strongest anchor, and I can’t honestly dismiss it. It gives absorption probabilities from an explicit finite matrix calculation. Recasting its input as a distribution over update-rule types, then asking for the large-population limit, is a credible continuous-population question. The paper’s use of anonymity and its discussion of extended OWA operators make that regime especially plausible.
The strongest objection is that the proposed limit preserves a feature the continuous society itself loses. For a fixed number of steps, the fractions of “yes” opinions converge to the deterministic update \(q\mapsto\sum_t\mu_tF_t(q)\). The proposed \(\lim_{N\to\infty}\alpha_N\), however, takes absorption over an unbounded horizon in each finite society first. It can therefore depend on rare fluctuations that the deterministic population dynamics does not represent. That makes it a legitimate limit problem, but its object is the asymptotic behavior of finite stochastic approximations, rather than absorption in the continuous society. Whether the authors would regard that as their problem is uncertain.
The better continuous mirror would define the dynamics directly on masses and ask about consensus or other terminal outcomes. That fits the population axis, but it turns the paper’s main question into analysis of a finite-dimensional deterministic system. It does not inherit Corollary 6’s stochastic absorption calculation as a computational problem. Theorem 1 does not supply another computational anchor: it characterizes when only consensus terminal classes exist.
This is a weak negative case. The stochastic-limit proposal survives the strongest objections: update rules can be sensible types, many agents can share them, and the absorption probability is a real quantity the paper computes. My case against it rests on doubt about whether a limit governed by finite-population fluctuations is the right continuous mirror for this programme—not on a fundamental obstacle to formulating or studying one.
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.