Ricerc@Sapienza

Human performance can significantly affect the success and safety of works. For example, the crew piloting an airplane, surgeons in the surgical room, and controllers managing air traffic. The objective assessment of human performance would therefore allow for better tailor and intervene on the management and ongoing work of the operator to finally maximize their efficiency and mitigate the risk of committing errors. In general, human error is consistently identified as one of the main causes of accidents. In this regards, Human Factors (HFs) research investigates the causes of operators' most common errors with the aim of preventing degradation of performance that could bring the operator out of the safety limits. Innovative approaches based on the use of operator's neurophysiological mental states evaluation have been proposed in addition to conventional methods (questionnaires, self-reports, expert evaluation). Being based on real time data directly linked to mind-body reactions, thus intrinsically objective if compared with self-assessed and subjective evaluation, this approach turned out to effectively and rapidly detect any impairment of performance. In fact, it was demonstrated how user's performance heavily depends on interactions between different HFs (e.g. Stress, Attention). The Human Performance Envelope (HPE) was proposed to completely define single operator's performance along with their HFs interaction. However, HPE is still a psycho-behavioural model that lacks of a reliable measure. To overcome this limitation, NEUROID aims to generate a measurable HPE model by combining machine learning (ML) algorithms and multivariate autoregressive (MVAR) models for the analysis of the operator's neurophysiological signals. More specifically, the HFs time-series will be modelled as a network where the nodes are the HFs, and the edges their directed causal relationships. The characterization of this network will allow to model and quantify the operator's HPE.