Description
For each error type, check who has access, knowledge, attention and the right perspective to detect it, then trace how it is recovered and who learns from it.
Hutchins argues that where systems rely on on-the-job learning, error is inevitable, so design should support not only prevention and recovery but learning. Detecting an error requires four conditions: access to the behaviour or its trace, knowledge or expectation of correct performance, attention to it, and a perspective that brings the relevant expectations to bear. Task design can force errors to show, as a third line of position on a chart makes a bad fix visible as a large triangle. Different roles have different perspectives: a plotter sees bearings as directions where the recorder sees digits. An evaluator role has access and knowledge without doing the work, at a cost in processing capacity. Corrections teach the person who erred, the detector and witnesses, so a low, reliably detected error rate can be functional for a team.
Attributes
- Discipline
- Human Factors & Error
- Loss type
- Latent
- Data
- Qualitative
- Effort
- Weeks
Steps
- List the error types observed in the team's work.
- For each, identify who has access to the error or its trace.
- Identify who has the knowledge to recognise it as an error.
- Check whether the task design forces attention to it, for example through redundant cross-checks.
- Identify which role's perspective makes the error visible.
- Consider adding an evaluator role and weigh it against lost processing capacity.
- Trace how errors are recovered, with or without diagnosis and instruction, and who learns: the person who erred, the detector and witnesses.
Sources
- Cognition in the Wild (Hutchins, 1995)
Also known as
- Access-knowledge-attention-perspective check
- Designing systems that learn from error
- Evaluator role analysis
Categories
- Accessibility
- Human Factors & Error