When work changes hands, what gets lost?

01The problem

Handover is one of the most consistent points of failure in collaborative work — a nurse ending a shift, an operator changing stations, a teammate stepping away mid-project. Whatever isn't said or written down at that moment tends to be the thing the next person has to rediscover the hard way.

Handover has been studied closely in medicine, where standardized forms and mnemonics have measurably reduced errors at shift-change. But those interventions are built around fixed checklists — they help people remember what to say, without addressing what information is actually available to include, or how it should be prioritized for the person receiving it. That's the space this project sits in: using system-side data alongside a person's own account of what happened, so a handover report can be more complete than either source would be alone.

02The approach

The core idea is to treat two sources as complementary rather than redundant: what a system observed directly, and what a person experienced and chose to report. Neither is complete on its own — system logs miss judgment and context, human reports miss things people forget to mention or didn't think were worth writing down. The system reconciles both into a single handover report.

The system is designed to degrade gracefully — it can still produce a useful report from telemetry alone if no human input is available, and improves as more human context is added.

03The testbed

To study handover somewhere controlled and repeatable, we built a simplified simulation loosely inspired by real human-factors research on nurses' cognitive work: a "hospital floor" task where participants juggle several concurrent goals under time pressure, then get pulled away and asked to hand off to the next person — themselves or someone else — with only a short window to write down what matters.

// simulation screenshot — swap in your own capture

That constraint — a real person, a real clock, an imperfect account of a partially-finished task — is what makes the handover problem concrete enough to actually measure.

04Status

We're currently finalizing analysis and preparing the manuscript for submission. This page will be updated with a preprint link once one is available — check back, or reach out directly if you'd like more detail in the meantime.

05Team