Skill-Rule-Knowledge Analysis

Description

Classifies the cognitive control level a task demands (skill, rule or knowledge based) to predict error types and design displays that support each level.

Rasmussen's SRK framework distinguishes three levels of cognitive control. Skill-based behaviour is automatic and driven by perceptual signals; rule-based behaviour applies stored if-then rules triggered by recognised signs; knowledge-based behaviour is effortful reasoning about symbols and models in novel situations. Each level has characteristic errors and needs different interface support: skill-based tasks need direct manipulation and consistent spatial mappings; rule-based tasks need clear cues that match rules; knowledge-based tasks need representations of the system's functional structure. As a method, researchers analyse tasks and incidents to determine which level is in use at each step, often through observation and cognitive interviews, and then check that the interface supports that level. It underpins ecological interface design and Reason's GEMS error model.

Attributes

Loss type
Extinct
Effort
Weeks

Steps

  1. Decompose the task into steps using observation and interviews with experienced and novice users.
  2. Classify each step as skill, rule or knowledge based by the cues used and the degree of deliberation.
  3. Map known or observed errors to SRK levels (e.g. slips at skill level, misapplied rules, knowledge-based mistakes).
  4. Assess whether the interface provides signals, signs or symbols appropriate to each level.
  5. Redesign to support the right level, for example functional displays for knowledge-based fault diagnosis.
  6. Test performance under both routine and novel scenarios to confirm support at each level.

Sources

  • Handbook of Human Factors and Ergonomics
  • Information Processing and Human-Machine Interaction (Jens Rasmussen)

Also known as

  • Skill-Rule-Knowledge Model
  • SRK framework
  • Rasmussen's SRK taxonomy

Categories

  • Accessibility
  • Human Factors & Error