Six simulated ECLSS subsystems (O₂, CO₂, cabin pressure, temperature, power, water) each drive six indicator panels on a 6×6 console, drawn here as a flat 2D grid (drag it to tilt the view — the panels were laid out on a flat plane in the source 3D scene too, so this is the same physical layout, just rendered without a camera). Alarms fire at random (biased 60% toward the same subsystem as the previous one, since real cascading faults cluster) and must be clicked before a 9 s timeout.
Predicted acknowledgment time combines three well-established human-factors models (verified numerically — internally consistent with the constants below):
Fitts's Law (pointing): MT = 0.15 + 0.20·log2(D/W + 1) s
Hick-Hyman Law (choice): DT = 0.20 + 0.15·log2(N + 1) s
Serial search (mono only): ST = 0.05·(36/2) s
Predicted RT = DT + ST + MT
D is the distance (in console grid units) from the previously acknowledged panel to the new alarm, W is the panel's rendered width (the size slider), and N is how many alarms are simultaneously active (more choices = slower decisions, per Hick-Hyman). The bar chart under the console breaks the latest prediction into its DT/ST/MT parts.
- Color-coded alerts — active panels glow at full brightness in their subsystem's hue: a pre-attentive "pop-out" that visual-search theory (Treisman & Gelade) predicts is found in roughly constant time regardless of how many panels exist.
- Monochrome — active panels only get slightly brighter with no hue difference, forcing a serial scan of the console; the model adds ~0.05 s per panel scanned (half of 36, on average).
- Grouped layout — same-subsystem panels sit in one row, so a cascading fault keeps D small. Scattered randomizes the subsystem-to-cell mapping, inflating D for the same cascade.
The "Cognitive load" readout is a heuristic 0–100 composite of alarm density, miss rate and whether color coding is off — illustrating why real spacecraft HMI standards (e.g. NASA-STD-3001) mandate color- and priority-coded alerting.