The spotlight that could not explain itself
The rehearsal room was dim, and the new lighting desk kept snapping the spotlight onto one actor, leaving another in shadow. The director leaned into the booth. "Why that person, right now?" The desk felt like a medical computer, picking who gets the light.
The operator shrugged and pointed at the smooth-looking stage. Most nights it worked, sure. But one wrong cue can wreck a scene, and a wrong medical suggestion can hurt someone. When families and rule-checkers ask why, "it usually works" is not an answer.
So the crew brought out a thick stage manager’s binder, not a new gadget. It sorted the ways people try to explain these computer choices in medicine. It kept coming back to three questions the booth always asks, why explain, how to make one, and when to show it.
The binder also fixed a word mess. One note can be easy to read but still be wrong about what the desk actually did. A useful explanation needs clarity, so a person can follow it fast, and it needs faithfulness, so it matches the desk’s real reasons.
It grouped the main explanation styles into five families. Some squash lots of signals into a short list. Some point to the details that mattered most. Some show where the system focused over time, like a restless beam. Some turn behaviour into simple cue rules. Some use a simpler stand-in to copy the desk.
Then came the awkward bit. People often stop after printing an explanation and never check if it helped anyone make a safer call. Too many highlights can turn into noise, so people ignore them. A stand-in can also drift, sounding neat while no longer matching the real desk. Explanations still sit beside basics like careful checks and clear rules.
By the next run-through, the director wanted a setup the crew could understand from the start, not a mystery box with excuses taped on later. If they had to use the powerful desk, they asked for notes that were clear and faithful, then still rehearsed and cross-checked. The spotlight looked the same, but the trust felt different.