It just so happens that we are not there yet. Automated driving technology is still evolving and the point at which it may be able to fully replace the driver in every situation may not be so close. In the meantime, what we have are semi-automated vehicles that are able handle certain functions and situations themselves, but still need human support in others. The best way to mediate this relation has become an important matter of discussion in the scientific and engineering community. It was also the main focus of the COOPERATIVE initiative, that during one and a half years has brought together researchers from CCG and IFE, in an effort to identify relevant Human Factors challenges related with Connected, Cooperative and Automated Driving (CCAD). The initiative has materialized in several knowledge sharing and discussion events, also involving moments of outreach to the scientific community and the general public, as well as a scientific publication.

Analyzing the literature regarding human-vehicle interaction in the context of CCAD, we have reached some conclusions. Quite important was the understanding that at the core of many Human Factors issues is the paradigm of function allocation, which places the focus of the interaction in the control transitions between human and automation. Concretely, it assumes that a role division should exist between automation and human, with the former taking care of the “low-level” operational part of the driving task (lane keeping, steering) and the latter assuming a vigilant role and expected to intervene in situations that the automation is unable to handle (Metge et al., 2021). Importantly, this function allocation is supposed to be dynamic. Automation may take more or less responsibility depending on situational circumstances that may allow it greater or lower level of autonomy and authority (e.g. highways with clear markings and without vulnerable road users allow for a greater degree of automation).

Transitions thus become the crux of the question. Humans are not good at maintaining vigilance and it may be that they are required to takeover control when they are “out-of-the-loop”. They may be inattentive or drowsy, or simply unaware of the surrounding situation, and consequently, strongly diminished in their ability to control the vehicle and make proper decisions at the moment of takeover (Morales-Alvarez et al., 2020). Also, in a dynamic function allocation paradigm, they may often be not fully aware of how autonomously can the system operate and consequently, of which degree of driving responsibility they should assume. The initiative has thus focused on finding alternative visions to (dynamic) function allocation paradigm, that may allow taking advantage of automation technology, while avoiding some of its pitfalls. We believe that the idea of human-machine collaboration (inspired by the literature on human-robot collaboration – e.g. Bauer et al., 2008; Bicho et al., 2012; Fong et al., 2003) may provide a path to do so. Contrarily to a notion of role distribution, the main idea behind the humanmachine collaboration is that instead of a definition of different tasks to each collaborator, both may be engaged simultaneously in the same task, interacting continuously between themselves and merging contributions until a final objective is completed.