- Rating
- 2
- Confidence
- 3
- Informativeness
- 3
- Presentation
- 2
- Soundness
- 1
Read review
Paper summary
This paper studies how to reduce opinion polarization and disagreement by modifying social-network connections when individuals’ innate opinions are unobservable and only noisy aggregate feedback is available. The authors show that, under the Friedkin–Johnsen model, intervention outcomes can be represented exactly as a linear relationship between known network-structural features and unknown opinion parameters. This eliminates the need to estimate the opinion subspace before learning. Based on this structure, the paper proposes Lift-TS for stationary settings and the Learned Dynamics Tracker (LDT) for settings with continuously drifting opinions. LDT combines parameter estimates and uncertainty information from multiple temporal windows in a low-dimensional spectral space and uses an offline-trained recurrent model to correct the online estimates. Experiments show that LDT achieves lower dynamic regret across different horizons, candidate-pool sizes, and network topologies.
Critical strengths
- The core structural observation is simple and important, eliminating the per-instance subspace-estimation stage required by prior methods.
- LDT’s low-dimensional parameter-tracking design is well aligned with the problem structure and is computationally efficient.
- The theory, algorithms, and experiments are generally well connected.
- The experiments include strong baselines, ablations, statistical tests, runtime analysis, and cross-network transfer evaluations.
- The authors clearly disclose that the real-network experiments use real topologies but synthetic opinion dynamics.
Critical weaknesses
- Theorem 2 evaluates posterior sampling on histories generated by the pretrained policy, but directly applies a regret bound established for histories generated by posterior sampling itself. A distribution-shift argument is missing.
- The drift-error bound in Theorem 3 may omit a dependence on the conditioning of the design matrix, potentially affecting the claimed dynamic-regret rate.
- The theoretically optimal window varies with the horizon and variation budget, whereas the deployed LDT is projected around a fixed reference window. The asymptotic guarantee for the deployed algorithm is therefore not fully established.
- Cross-graph transfer appears to require shared dynamics in spectral coordinates in addition to low-rank homophily, but this stronger assumption is not formalized.
- LDT still requires online feedback, parameter updates, and forced exploration, so the claim that pretraining “replaces online search” is overstated.
Questions for authors
- How can the history-distribution mismatch in Theorem 2 be addressed rigorously?
- Can the authors prove that the sliding-window drift error does not contain an additional design-conditioning or effective-sample-size factor?
- How does the deployed LDT with a fixed reference window achieve the claimed dynamic-regret rate?
- Does cross-graph transfer require an explicit assumption that parameter dynamics are shared in spectral coordinates?
- When innate opinions are unobservable, how can the gap-dependent pretraining error be estimated before deployment to determine whether amortization is appropriate?
5 comments
Response to Reviewer QCqN (2/3)
W2 & Q2. No additional conditioning factor in the drift error
Yes: for the window estimate of Theorem 3(ii), which the revised proof states is computed on the scheduled design draws, the drift error contains no additional conditioning or effective-sample-size factor, so the rate is unchanged; the conditioning only affects a constant independent of and . The lemma below is added as Lemma 4 in Appendix B, and the proof of Theorem 3 now states this estimate and carries its constant.
Reason. The scheduled design draws of Lemma 5 come from a -optimal design, so with high probability the draws in a window give ; the effective sample size is already the denominator of the variance term (Equation (21)), and it is what sets the order.
Proof for Q2. For the ridge estimate on a window with Gram matrix , the drift bias is . With ,
By (ii), the rate of Theorem 3(ii) is unchanged, with an extra factor in the constant. For estimates that use every round of the window (as in LDT), Lemma 5 gives after a short burn-in under , and the rate is again at , the schedule of Theorem 4′ (W3 & Q3).
Measurement. Replaying the deployed LDT on five episodes of each of the 20 streams and computing and exactly at every decision round, the realized drift bias always stays below the factor allowed by (i), and its median is more than two orders of magnitude below it:
| Window | Realized , median | Factor allowed by (i), median | |
|---|---|---|---|
| 200 | 60/150/full | 0.017/0.030/0.036 | 5.2/7.1/7.2 |
| 800 | 60/150/full | 0.023/0.021/0.040 | 5.3/8.3/13.9 |
W3 & Q3. Rate guarantees for the deployed LDT
The deployed LDT has two guarantees. First, with a fixed reference window, Proposition 2 bounds its regret along every realized trajectory, and the leading term of this bound grows as on the deployed streams (Table 25), which is sublinear. Second, letting the reference window grow with the round index, with matching exploration and safeguard schedules and the trained filter unchanged, gives the claimed a-priori rate (Theorem 4′), and this configuration is never significantly worse than the original one.
Reason. The theoretically optimal window grows with the horizon because a window of fixed length keeps a fixed number of observations (and, under a decaying exploration schedule, an ever smaller number of design draws), so its a-priori error bound does not decrease over time; this holds for any fixed-window method and is not specific to LDT. Theorem 4′ therefore lets the reference window of LDT grow with the round index, without knowing the horizon.
Theorem 4′. Let , and . Under the budgeted path condition of Theorem 3, for filter weights fixed before deployment, every and every , with probability at least ,
Every constant is explicit and independent of . The proof splits the regret into scheduled rounds, safeguard radius, anchor noise and drift bias; the anchor window holds design draws, so (Lemma 5), the drift bias is controlled by (i) of W2 & Q2, and each of the four parts sums to . Theorem 4′ is now stated in §4.3, with its proof in Appendix B.
Performance. With chosen on the validation streams, the configuration of Theorem 4′ is also significantly better than the strongest baseline at every horizon (differences in regret):
| Theorem 4′ minus | |||||
|---|---|---|---|---|---|
| submitted LDT | -1.1 [-1.6,-0.7] | +1.1 [-0.6,+2.6] | +2.8 [-1.4,+7.1] | +6.3 [-2.9,+15.8] | +7.0 [-9.7,+23.8] |
| SW-Ridge | -4.6 [-7.4,-1.9] | -12.0 [-19.5,-4.6] | -28.4 [-42.9,-13.9] | -66.0 [-96.2,-36.5] | -131.4 [-193.9,-70.4] |
Response to Reviewer QCqN (3/3)
W4 & Q4. Cross-graph transfer needs no shared-dynamics assumption
No. None of the guarantees for cross-graph transfer relies on an assumption of shared dynamics; such an assumption only determines how large the improvement from learning is, and the revision states it explicitly.
Reason. Proposition 2, Corollary 1 and Theorem 4′ hold for every filter weight, every drift trajectory and every graph (Theorem 4′ under Assumption 1 and the variation budget), so the performance after transfer to a new graph is never worse than these guarantees. Only the size of the improvement of LDT over its anchor (Corollary 1) depends on whether the deployment drift resembles the pretraining drift, and the revision states this as Assumption 2 in §5.2, used only for that gain statement:
Assumption 2 (shared spectral drift law). In the canonicalized spectral coordinates of Appendix E, the episode trajectories of a deployment are drawn from a law in the support of the pretraining distribution.
Experiments. Under drift laws outside the pretraining family, LDT keeps its advantage: on a persistent oscillator with no stable point (Table 2), it leads SW-Ridge significantly at , by 84.5 [37.8,133.9]; on a smooth random wander over the whole sphere, it has the lowest mean regret at every horizon. Across graphs, LDT is significantly below SW-Ridge on six of the seven real networks of Table 33 and lower in mean on the seventh, including networks 25 and 40 times larger than any pretraining graph.
W5. More precise wording
We have changed this phrase to "replaces per instance estimation and tuning" in the abstract, introduction, related work and conclusion. During deployment, LDT does not update its model parameters.
The online work of LDT consists of updating the window statistics and one forward step of the frozen recurrent network (0.11 ms per round, independent of ), with scheduled exploration on at most 2% of rounds (Table 10); LDT does use the observed feedback, as any online learner must.
By online search we mean the per-instance work of the two-stage approach on each new graph and episode, namely an exploration stage to estimate a representation followed by tuning and fitting of the learner; pretraining replaces exactly this work.
Q5. Estimation before deployment without observing opinions
It can be estimated without observing the true opinions: the pretraining term is an expectation over the prior distribution of opinions, so it depends only on the deployment graph, the candidate pool, the noise level and the prior, and can be computed before deployment from opinions simulated from the prior. On five networks, this estimate agrees with the value at deployment.
Explanation. The prior is the distribution the platform specifies in its model (the same family used in pretraining), not the true opinions; the computation follows the procedure of Table 36, and the true opinions are never read.
Experiment. The pre-deployment estimate uses 20 simulated opinion draws only, and the deployment value is computed on 20 further, disjoint draws (Appendix F):
| Network () | , pre-deployment estimate | , at deployment | Realized regret at deployment |
|---|---|---|---|
| Erdős-Rényi (32) | 50.0 [27.6,78.0] | 47.5 [31.3,65.9] | 47.4 [31.2,65.8] |
| Florentine (15) | 57.3 [32.2,86.1] | 64.3 [38.7,93.0] | 64.0 [38.5,92.6] |
| Davis (32) | 31.8 [19.8,51.7] | 39.9 [27.8,55.4] | 39.7 [27.7,55.2] |
| Karate (34) | 24.5 [19.0,30.6] | 25.0 [19.9,30.5] | 25.0 [19.9,30.6] |
| Les Misérables (77) | 216.7 [151.1,285.5] | 257.5 [150.4,388.1] | 256.0 [148.7,386.4] |
On every network the pre-deployment estimate falls inside the interval of the deployment value and ranks the networks correctly, so whether amortization is appropriate can be judged before deployment. The same principle gives a pre-deployment rule for choosing between LDT and Lift-TS; on all 29 transfer settings of the paper, the learner it selects is never significantly worse than the better of the two.
Response to Reviewer QCqN (1/3)
We sincerely thank the reviewer for the careful and constructive questions, which helped us make the theory more precise.
Our revised PDF colors every change: green for this review, orange for changes that also answer other reviews. In the revision, Theorem 2 is stated on the pretrained policy's own histories, with its proof in Appendix B; the drift-bias constant is explicit in Lemma 4 and in the proof of Theorem 3; the anytime certificate, Theorem 4′, is stated in §4.3; Assumption 2 is stated in §5.2; the phrase "replaces online search" is corrected; and Appendix F adds the pre-deployment estimate of Rpre.
W1 & Q1. Theorem 2 holds under the policy's own history distribution
No additional distribution-shift argument is needed. Theorem 2 holds under the pretrained policy's own history distribution, with the same form and one additional measurable term DT; the rate is unchanged whenever DT=O(logK), and in our measurements DT is below logK. The revised Theorem 2 states this form, and Appendix B gives the full proof.
Rigorous proof for Q1. The core inequality of Theorem 5 (the information-ratio bound, Russo and Van Roy, 2016, Proposition 5) holds round by round under any posterior, so it applies directly on the histories of the pretrained policy π; only the final summation of the per-round information involves the history distribution, and we carry it out along the rollout of π itself. With pt exact posterior sampling evaluated on the histories of π, this gives
E[RegretT(π)]≤max{σ,Δmax/2}2deffT(logK+DT)+Rpre,DT=∑tEπ[∑a(pt(a)−πt(a))+It(A⋆;Yt,a)],
with expectations under the history distribution of π; DT measures the information posterior sampling would have collected and π did not. The proof has four steps:
Direct comparison. Against the regret of posterior sampling on its own histories, a hybrid argument gives E[RegretT(π)]≤Ep[RegretT]+21ΔmaxT(T+1)εˉ, where Ep denotes that rollout and εˉ is the imitation error on the deployed policy's own histories.
Measurement. On the 40 instances of Table 36, DT=2.71 nats, below logK=4.61, so the constant of the bound grows by at most a factor 1.26. The posterior-sampling term is small both on the policy's histories (0.39) and on posterior sampling's own histories (0.26), and both are two orders of magnitude below Rpre=28.06, so the history mismatch has a negligible effect in practice.