Taking Human Performance Seriously in Software
summary
Argues that outages are cognitive problems not just code problems, using line of representation to show operators act on keyhole views of systems and cognitive task analysis and timeline mapping to trace real work during incidents. Presents results from the industry-academic SNAFU Catchers consortium and the Stellar report, shows red-team/blue-team trace maps and examples of cooperative advocacy, and claims shallow metrics like time-to-resolve hide what people actually experienced. Recommends treating incidents as unplanned investments, dropping root-cause hunting, and measuring human performance and questions people actually ask.
AI-written summary · about
chapters (8)
- 0:00Introduction and roadmap
- 1:57Research background
- 6:57Above and below the line
- 13:00Cognitive work and mental models
- 17:04Studying incidents
- 24:52Incident influences
- 27:45Implications and takeaways
- 32:50Call to action
related talks
43mSoftware Operations and Human Cognition
29mHacking Human Systems
31mHard Problems We Handle in Incidents but Aren't Recognized
31mClosing Keynote
38mTechniques and Tools for a Coherent Discussion about Performance in Complex...
03mSituation normal: All fouled up
41mAre We Getting Better Yet? Progress Toward Safer Operations
21mOptimizing IT Performance in Complex Systems