Real-time and multi-factor determination of mental workload for the aircraft certification process

Certifying cockpit systems is a long and costly process, with much emphasis on the negotiations between industry and authority's representatives. Also, a recent trend has resulted in a split between conventional and human factor scenarios approaches. As such, Human-machine Interface certification is performed rather late in the process and still relies on well-known and accepted methodologies, such as NASA-TLX tool for assessing workload. Conversely, ground truth for mental workload evaluation in aeronautics remains subjective values obtained from very limited expert pilot panels. On the other hand, other methodologies such as dual tasks and physiologic measures have been developed in labs for decades and used in simulated or real flights, but with a somewhat limited perceived added value by both industry and States experts. This article aims at discussing this conundrum through a recent example. It will present the rationale for the development of a physiologic approach based on realistic scenarios in aircraft simulators, and machine learning tools to implement a real-time complementary approach for mental workload measurement, which could help the certification process. The perceived need for pilot real-time behavior modeling and some elements of trust and acceptability will be discussed.