H2F BITESIZE #76

I bring you a weekly bite-sized chunk of the science behind helicopter human factors and CRM in practice, simplifying the complex and distilling a helicopter related study into a summary of less than 500 words.

TITLE:

Enhancing decision-making in single-pilot operations through proactive and explainable human-autonomy teaming.

WHAT?

An experimental study investigating whether AI assistants can improve pilot decision-making, situational awareness and workload management during complex, time-critical single-pilot operations.

WHERE?

Bordeaux-INP, France, using an Airbus A320-200 static simulator.

WHEN?

Published October 2026 in Journal of Cognitive Engineering and Decision Making.

WHY?

Replacing a second pilot with automation raises questions about how crews will maintain situational awareness and make effective decisions under pressure. The researchers wanted to establish whether AI that actively participates in monitoring and decision-making offers greater benefits than an assistant that simply answers questions.

HOW?

22 airline pilots completed a demanding simulated approach involving critically low fuel, a go-around, changing winds and an electrical failure. Together, these created a situation where the original landing runway became unsafe, requiring pilots to recognise the problem and select an alternative.

Four conditions were compared:

  1. Unaided: No AI assistance.
  2. On-demand AI: Responded only when questioned.
  3. Proactive AI: Independently identified emerging problems and prompted reassessment.
  4. Proactive explainable AI: Also explained its recommendations and the operational factors behind them.

The AI communicated verbally, supplemented by augmented-reality information. Researchers assessed runway selection, workload, situational awareness, scenario comprehension and pilot–AI interactions. Results were compared with an earlier multi-pilot baseline.

FINDINGS:

  • Decision-making improved dramatically. Only 20% of unaided pilots and 40% using on-demand AI selected the safest runway, compared with 100% using either proactive AI system.
  • AI explanations improved understanding. Scenario comprehension increased from 20% unaided to 40% with on-demand AI, 43% with proactive AI and 75% with explainable AI.
  • Proactive AI generated significantly more exchanges between human and AI concerning decision-making and situational awareness.
  • Workload nevertheless remained comparable. Subjective workload was reported high across all conditions, with no statistically significant differences. Additional AI explanations did not measurably increase or decrease workload.

SO WHAT?

The study suggests that one potential contribution of AI may be recognising when pilots need assistance, rather than simply responding when asked.

Proactive monitoring and timely prompts by AI helped pilots reconsider an increasingly unsafe plan. Explanations offered additional benefits by helping them understand why circumstances required reassessment.

This has important implications for human–AI teaming dynamics, CRM and future cockpit design. Effective AI assistants may need to function more like proactive colleagues rather than being sophisticated information-retrieval tools.

It is worth noting that the experimental AI was simulated by researchers. Pilots were told it was completely reliable, and therefore did not independently verify its calculations. The findings therefore demonstrate potential benefits under idealised conditions, while highlighting the unresolved challenge of automation bias and appropriately calibrated trust.

REFERENCE: 

Dormoy, C., Hentati, T., & Letouzé, T. (2026). Enhancing decision-making in single-pilot operations through proactive and explainable human-autonomy teaming. Journal of Cognitive Engineering and Decision Making. Advance online publication.

Leave a Reply