CREAM (Cognitive Reliability and Error Analysis Method)

Description

Assesses how working conditions set a person's control mode (scrambled, opportunistic, tactical, strategic) and uses this to predict or retrospectively explain cognitive failures.

CREAM is a second-generation HRA method that rejects context-free error probabilities. It uses nine Common Performance Conditions, including adequacy of the human-machine interface, available time, time of day, training and crew collaboration, rated as improving, not significant or reducing performance. The count of improving versus reducing conditions places the operator in one of four contextual control modes, each with an error probability interval. CREAM also offers a retrospective analysis with classification groups linking observed error modes (phenotypes) to causes (genotypes) through cause-consequence tables. For UX, it formalises the idea that the same interface yields very different reliability under different contexts, and it gives a structured way to analyse failures in terms of observation, interpretation, planning and execution.

Attributes

Loss type
Unmigrated
Effort
Days

Steps

  1. Describe the task and its context; perform a task analysis down to cognitive activities (observe, interpret, plan, execute).
  2. Rate the nine Common Performance Conditions using field observation, interviews and log evidence.
  3. Count improving and reducing conditions and determine the expected control mode from the CREAM diagram.
  4. For predictive use, identify likely cognitive function failures per step and adjust estimates by control mode.
  5. For retrospective use, trace observed error modes back through antecedent-consequent tables to plausible root causes.
  6. Recommend design changes that shift conditions from reducing to improving, especially interface adequacy and time available.

Sources

  • Hollnagel, Cognitive Reliability and Error Analysis Method: CREAM (1998)

Also known as

  • CREAM

Categories

  • Accessibility
  • Human Factors & Error