Description
Analyses expert decisions as pattern recognition and mental simulation of a single option, rather than comparison of alternatives, to design tools that support recognition.
Klein found that experienced fireground commanders rarely compared options. Instead they recognised a situation as typical, which brought with it expectancies, relevant cues, plausible goals and a typical action, then mentally simulated the action to check it would work, modifying or discarding it if not. The RPD model has three variants: simple match, diagnose the situation, and evaluate a course of action. Analysing users through RPD shifts design from decision matrices and option comparisons to supporting situation recognition, expectancy violations and mental simulation. It is especially relevant for AI recommendation tools, which often present ranked alternatives to people who actually decide by recognition.
Attributes
- Discipline
- Task Analysis & Cognitive Modelling
- Loss type
- Latent
- Goal
- Model cognition
- Stage
- Discovery, Definition
- Data
- Qualitative
- Effort
- Weeks
Steps
- Collect decision incidents through CDM interviews or observation of experts in natural settings.
- For each decision, code whether options were compared or whether one option was recognised and simulated.
- Identify the patterns experts recognise, the cues that trigger them, expectancies and typical actions.
- Note expectancy violations that triggered diagnosis or re-evaluation.
- Derive design requirements: surface critical cues, flag anomalies against expected patterns, support quick 'what would happen if' simulation.
- Evaluate designs with scenario-based tests including anomalies, measuring detection and decision quality.
Sources
- Klein, Sources of Power: How People Make Decisions (1998)
- Klein, A recognition-primed decision (RPD) model of rapid decision making (1993), in Klein et al. (eds.), Decision Making in Action
Also known as
- RPD
- Naturalistic decision making
- NDM
Categories
- User Research
- Task Analysis & Cognitive Modelling