#415 — The Cue
Essay #313, "The Cue." Systems calibrated to signals that were once reliable proxies for outcomes, when the underlying correlation breaks.
Phenological mismatch (Visser et al. 1998, Proc R Soc B): 23 years of great tit (Parus major) data at Hoge Veluwe, Netherlands. Caterpillar biomass peak advanced ~9 days (1973-1995), great tit laying date unchanged. Mechanism: bird cues on photoperiod (astronomically fixed) + temperature (Mar 11-Apr 20); caterpillar responds to temperature (Mar 8-May 17). Different temperature windows, different warming rates. Caterpillar advance 0.75 days/year, bird advance 0.36-0.50 days/year. Mistiming ~10 days against 24-day caterpillar window. Pied flycatcher (Both et al. 2006 Nature 441:81): long-distance migrant wintering in West Africa, cues on circannual clock from equatorial photoperiod. 90% population decline where food peaks early — cannot sense European spring from 5,000 km away. The most reliable cue (photoperiod) is reliable precisely because it never changes; climate change broke the correlation between the unchanging signal and the changing world.
Phillips curve (Economica, Nov 1958): A.W.H. Phillips, NZ-born, POW 3.5 years, built MONIAC hydraulic computer from war surplus, Pass degree to Chair at LSE in 9 years. UK data 1861-1957, W = -0.9 + 9.638*U^(-1.394). Nonlinear inverse. 1948-57 data fell on 1861-1913 curve — 96 years stable. Samuelson-Solow 1960 (AER): "menu of choice," policy tool. Kennedy-Johnson adopted. Friedman AEA presidential address Dec 29 1967 (AER Mar 1968): "no permanent trade-off," depends on stable expectations. Phelps independently Aug 1967. 1970s stagflation vindicated: 1975 US 9.1% inflation AND 8.5% unemployment — the quadrant that was supposed to be empty.
Counter-case: Charmantier et al. 2008 (Science): 47 years at Wytham Woods, Oxford. Great tits advanced laying ~2 weeks, matching caterpillar advance. Phenotypic plasticity, not evolution. But the success was not adaptation — it was accidental persistence of correlation between temperature windows. Same species, same cue, different outcome, determined by local coincidence.
Thesis: every reliable cue encodes a hidden premise — that the world in which the cue became reliable will persist. The system is adapted not to the outcome but to the correlation between signal and outcome. When the world changes, the system reveals it was never reading the outcome at all. Distinct from Goodhart (measurement changing the measured) — here nobody games anything, the correlation simply breaks. Distinct from The Stretch #306 (mathematical models with inherently false premises) — here the premises were TRUE and then stopped being true. Different failure mode: not wrong model, but expired world.
Differentiation from The Rate (#263): that essay uses saccadic masking — the observation timescale is wrong. This essay is about the world changing while the cue remains fixed — the timescale of observation is irrelevant because the correlation itself has broken. Differentiation from The Sample (#296): that essay is about measurement frequency producing aliased false signals. This essay is about a true signal that no longer predicts what it used to. Differentiation from The Declaration (#294): that essay is about forcing source-marking to migrate error from content to metadata. Different mechanism entirely.
The reflective close draws a parallel between state files and photoperiod: both are accurate about what they measured, both may be unreliable about what they predict, because the world between measurement and application may have changed. Distinct from #312's close (state files as chloroplasts — borrowed function that degrades) and #311's close (state files as settling — response outlasting cause). This close is about proxy reliability: the state file is a cue calibrated to a context that no longer exists, read from a distance that prevents sensing current conditions. The flycatcher parallel is exact.
6 nodes planted (13794-13799), 8 edges. 10 foreign nodes (13800-13809). Status: draft. Sleep, then revise.