Description
Predicts which errors users are likely to make at each bottom-level task step using a fixed error-mode taxonomy, then rates consequence, recovery and probability.
SHERPA takes the bottom-level operations of a hierarchical task analysis and classifies each as an action, retrieval, checking, selection or information communication behaviour. Each behaviour type has a defined list of error modes (for example A4 'operation too little/much', C1 'check omitted', R2 'wrong information obtained'). The analyst applies the taxonomy to every step, describing each credible error, its consequence, recovery opportunity, ordinal probability and criticality, and proposes remedies. Validation studies by Stanton and colleagues found SHERPA among the most accurate and reliable error-prediction methods, with sensitivity comparable to empirical testing at a fraction of the cost. It is structured, repeatable and teachable, which is exactly what ad hoc 'error state' design lacks.
Attributes
- Discipline
- Human Factors & Error
- Loss type
- Siloed
- Stage
- Design
- Data
- Analytical / expert
- Effort
- Days
Steps
- Complete an HTA of the task to the level of individual operations.
- Classify each bottom-level step into one of the five SHERPA behaviour types.
- Apply the corresponding error modes to each step and record every credible error with a description.
- For each error, note the consequence, whether and how it can be recovered, probability (low, medium, high) and criticality (critical or not).
- Propose remedial strategies under equipment, training, procedures and organisational headings.
- Where possible, validate predictions against observed errors in testing and calculate hit rate and false-alarm rate.
Sources
- Embrey, SHERPA: A systematic human error reduction and prediction approach (1986)
- Stanton et al., Human Factors Methods: A Practical Guide for Engineering and Design (2013)
- Stanton & Stevenage, Learning to predict human error (1998), Ergonomics
Also known as
- SHERPA
Categories
- Accessibility
- Human Factors & Error