Information Technology Reference
In-Depth Information
collision risk), thus justifying an action from the automaton. In all these cases, the
device “criticizes” the behavior of drivers and consequently assists them to improve
safety.
Here we are starting to address one of the key questions of automobile
copiloting: on the one hand the machine must “support” human activity by
facilitating the common objective of the human-machine system as a whole being
reached. Good cooperation in this case implies the sharing of a
common frame of
reference
[HOC 01], [PAC 02], [TER 90] between agents (adopted in [HOC 01]),
i.e. a certain agreement in the analysis of the situation and on what each of the
agents is or must carry out. This sharing is greatly facilitated if each of the agents
has a basic model of the way the other functions (which can imply training from the
driver's point of view and the integration of a driver model into the machine from a
technological point of view).
Nonetheless, a “good” copilot must also know how “relay” to the failing human
driver when he commits an obvious error or when himself unable to correctly
control the situation. Ultimately, copiloting aims to get the best out of the respective
abilities of the human and the machine. The human has abilities of analysis that are
often unequalled by the automaton, and are not soon going to be equaled, but he can
also make a mistake or reach his own limitations, thus justifying the relayed
assistance. It is therefore not about rejecting the substitutive philosophy that is
inherent to the automation of driving; in fact quite the opposite. It looks to
distinguish the situations for which this option fully justifies itself, from those where
other solutions would be better. Such is the real stake of “intelligent copiloting”: to
choose, in the range of possible human-machine interactions (from the diffusion of
information to taking control of the car) the action that will best meet the needs of
this
driver and the specific requirements of
this
situation [BEL 03]. From the point
of view of the ergonomist, the (total or partial) automation of driving does not
constitute an end in itself. It is only a means, one mode of assistance among others,
which has its advantages but also its limitations and problematic effects that it is
advisable to control.
Evidently, incorporating automobile copiloting into the logic of human-machine
cooperation is not without consequence from a technological point of view, and this
implies the active participation of social science researchers alongside engineers. It
is advisable to provide the machine with the ability to analyze the situation as well
as the activity of the driver (diagnose how critical a situation is, for example, and
judge how adequate the human behavior is [BEL 06]). It is also advisable to give it
a “knowledge of how to cooperate” [MIL 98], [MIL 03] in view of guaranteeing
good coordination of the copilot's activity, to anticipate the negative interference
risks between the human activity and the automaton and, as the case may be, to
resolve these conflicts if they occur. This is essential if we envisage going beyond