IDHEAS-ECA¶
IDHEAS-ECA (Integrated Human Event Analysis System for Event and Condition Assessment) is an HRA method documented in NUREG-2199. It analyses human performance through five macrocognitive functions, providing a more detailed cognitive model than spar-h's PSF-based approach.
Five Macrocognitive Functions¶
IDHEAS-ECA decomposes human performance into five cognitive functions:
- Detection: Perceiving that a situation requires attention (noticing alarms, recognising parameter deviations)
- Understanding: Comprehending what the detected information means (diagnosing the plant state)
- Decision-making: Selecting the appropriate response from available options
- Action execution: Physically carrying out the chosen response correctly
- Teamwork: Coordinating with other crew members, communicating, and maintaining shared awareness
Natural Human-AI Extension¶
The teamwork function already models interaction between crew members, making IDHEAS-ECA the most natural framework for extending to human-AI interaction. The AI advisory system can be modelled as a team member that the operator interacts with during each macrocognitive function.
AI-Related Cognitive Failure Modes¶
AI introduces new cognitive failure modes (CFMs) mapped to specific macrocognitive functions:
automation-bias → Decision-making: Operator follows incorrect AI recommendation without independent evaluation. The AI recommendation biases the decision-making process.
Alert fatigue → Detection: Repeated low-value AI outputs cause the operator to stop attending to AI advisory channels. A genuine AI alert is missed because the operator has learned to ignore the channel.
Mode confusion → Understanding: Operator misunderstands the AI system's current operating mode, context, or confidence level. The operator interprets AI output based on incorrect assumptions about what the AI "knows."
Reconciliation failure → Teamwork: Operator fails to resolve discrepancy between their own assessment and the AI advisory. Rather than investigating the disagreement, they either blindly accept or dismiss the AI input.
CFM/PIF Distinction¶
IDHEAS-ECA maintains a clearer distinction between cognitive failure modes (what goes wrong cognitively) and performance influencing factors (what conditions make the failure more likely) than spar-h. This cleaner separation makes it easier to systematically map AI influences to the correct level of the analysis — AI system reliability is a PIF that affects the probability of automation bias (a CFM) in the decision-making function.