EX-ITTEN-004 — CVD, aging, and redundant encoding

Question and falsifier

Do redundant encodings preserve consequential task performance across aging and congenital CVD better than color-only encodings? The redundancy principle fails if a well-powered representative study finds no performance or error-resilience advantage.

Work completed

All 66 pairs from a 12-color palette were transformed through typical, protan, deutan, and tritan screening matrices at modeled ages 25, 65, and 80. A simple lens-yellowing proxy and an Oklab 0.08 screening cutoff were applied: 792 observer-pair conditions.

Pilot results

Modeled detectable-pair proportions declined from 1.000 to .9848 for typical vision, .9697 to .9394 for protan simulation, .9242 to .8788 for deutan simulation, and remained .9394 in the simplified tritan transform. The coded shape-redundancy channel remained 1.0 by construction.

These are audit flags only. The cutoff, matrices, yellowing transform, and perfect shape detection are assumptions. The result supports selecting difficult pairs for a study; it does not estimate clinical detection rates.

Human experiment ready to run

• Groups: younger typical, older typical, diagnosed protan, diagnosed deutan, and older CVD; confirm class/severity with a validated test rather than self-report.
• Conditions: color only, shape only, label only, color+shape, and color+label.
• Tasks: urgent-state search, legend decoding, trend comparison, and prospective-memory alert response under time pressure.
• Outcomes: miss/false-alarm rate, response time, confidence calibration, and subjective workload. Analyze speed and accuracy jointly.
• Accessibility: large-enough marks, correction lenses allowed, fatigue breaks, and accessible consent.
• Analysis: mixed-effects signal-detection model; test interaction of encoding, age, CVD class, mark size, and eccentricity.

Current position / continuation

Stage: palette screening complete; representative user testing not started. Next action: recruit through optometry/accessibility partners and run a 10-participant usability pilot before fixing the final stimulus set. Data: ../experiments/data/ex-itten-004.json.