Modern industrial production environments, such as satellite payload assembly, face growing complexity and frequent disruptions. In the face of these uncertainties, traditional decision support tools often suffer from techno-centrism. By operating opaquely, these systems give rise to the automation paradox: they handle nominal situations autonomously but degrade the human supervisor’s situation awareness, exposing them to major cognitive and emotional overload when they must suddenly take back manual control in an emergency. As part of the transdisciplinary ANR project HIS3 (ANR-22-CE10-0012), this thesis aims to break away from this dynamic by designing an interactive, human-centred decision support tool. The fundamental objective is to reconcile the power of algorithmic optimisation with the cognitive abilities of planners, in order to put humans back at the heart of the control loop while preserving their agency. To address this challenge, we adopted an approach combining human-computer interaction (HCI), operations research (OR) and cognitive ergonomics, based on an analysis of supervisors’ activity in a French satellite payload assembly company. From the HCI perspective, we designed, developed, and tested numerous interaction and visualisation methods to provide users with suitable interactions and comprehensive visualisations while keeping their cognitive load low. From the OR perspective, we modelled the scheduling problem related to satellite payload assembly. Our contribution lies in the formalisation and integration of operator workload and rest-time constraints into a preemptive job-shop problem, which we identified as a gap in the literature. We developed an iterative constraint programming solution method for this problem. Our results show that lazy constraint generation can reduce computation time and improve solution quality for heavily constrained problems. We developed a complete prototype of a decision support tool. A qualitative evaluation of this tool with industrial experts revealed the emergence of a “delegation paradox”: the machine’s assistance does not seem to eliminate the overall workload but rather to shift it from manually creating the schedule to supervising and verifying it. To objectively measure these subjectively identified mechanisms, we designed a large-scale experimental protocol based on an ecological serious game. This setup will allow us to rigorously quantify the impact of the level of automation and the complexity of disruptions on decision-makers’ stress, mental workload, and situation awareness.