It is 2006, and a consortium of cities has posed a grimly practical challenge: given a municipal water network of tens of thousands of junctions and the simulated spread of a contaminant from any of them, place a hundred sensors to catch outbreaks fast. The objective — expected detection benefit — again rewards diversity over redundancy: sensors that cover different failure scenarios beat sensors that all watch the same mains. Exploit that diminishing-returns structure for guarantees and for speed, because the naive greedy evaluation is itself too slow at this scale. Then report honestly what the hundred sensors cannot see. Get it wrong and the placements cluster where the model was certain, while the contamination enters through the neighborhood the simulation undersampled — discovered by emergency-room arrivals instead of instruments.
Jelinek's statistical speech recognition, built on hidden Markov models trained from large corpora, established that empirical, data-driven statistical modeling could outperform hand-built rules — a methodological stance broadly applicable to modeling contamination spread from data rather than first principles. That is the extent of the connection. He has no direct engagement with combinatorial optimization, submodular set functions, or civil-infrastructure network design; his entire career was built around language and acoustic modeling for speech systems. Dropped into this water-network problem, his statistical instincts might inform a data-driven contamination model, but he brings nothing to the placement algorithm itself, its approximation guarantees, or the lazy-evaluation trick needed to make greedy search fast enough for the network's scale.
Gelman's hierarchical Bayesian modeling is well suited to representing uncertainty about where contamination might enter an under-instrumented network and to honestly propagating that uncertainty into a statement about what the hundred sensors cannot see, which is the reporting half of this problem. His applied-statistics instinct for calibration and honest uncertainty is a real asset. He has little direct connection to submodular combinatorial optimization or to the lazy-evaluation algorithmic tricks needed to make placement tractable across tens of thousands of junctions; his contribution would sit downstream of the placement algorithm, quantifying and communicating its residual risk, rather than upstream in designing the placement itself. He would strengthen the honesty of the answer more than its algorithmic core.
Battle #18 · 8/9/2026, 5:06:47 PM · this result is deterministic: the same two personas on this problem always resolve the same way.