“…On the other hand, physiological indicators ( [4], [5]) seem difficult to exploit and how well they relate to traffic complexity is unclear. Finally, widely used subjective ratings ( [6], [7]) provide high quality data (as they obviously relate to the kind of complexity investigated), but are often seen as subject to biases (such as the recency effect denounced in [5], and the possibility of raters errors in the case of "over-the shoulder workload ratings" [8]). In all of these cases, data are very expensive to collect, as they require the active participation of controllers.…”