Automation promises great results for the controller. Technical advances are expected to improve the performance of the human-machine system, and controllers assumed to be flexible enough to adapt to the new working environment. But partial automation can actually increase the risk of error attributable to human operators. This is because they are expected to monitor systems that are harder to understand and to exercise their skills only intermittently. In the US, the FAA’s Automated En-route ATC programme is designed to detect potential conflicts up to 20 minutes before they occur. It generates a resolution immediately, and provides simulation capabilities.As a result, the controller deals only with aircraft that the system says need attention, and while he/she has to rely on the system to detect problems and provide resolutions. The controller resolves them as soon as they are detected rather than wait to assess them visually on the situation display, so the focus of attention shifts from tactical monitoring, problem analysis and coordination toward strategic planning tasks.Many French controllers query this philosophy. Applied to French airspace, it could, mean a controller has to transmit a revised clearance to an aircraft in his sector to avoid a predicted conflict in another sector with an aircraft currently in a third. And if he disagrees with the clearance, he will have to propose an alternative without having access to information on what is happening in the other two sectors. If the controller’s role is limited to transmitting machine-generated clearances, he will soon lose his ability to make control decisions. CENA is developing an alternative at its Toulouse training centre. The En-Route Air Traffic Organizer (ERATO) is a software program to cope with what CENA terms the capacity barrier. It takes all the information available from sensors and flight plans, plus intention data obtained by data link from aircraft to create a working airspace model. The scenario it presents can identify and prioritize potential conflicts, and offer tool to resolve them. To do this, ERATO enlists the support of artificial intelligence techniques. It can describe the mental activity of controller, and assess deterioration under pressure of time, fatigue and stress. The central question is to identify the conditions under which the controller’s knowledge ceases to be accessible. It also sets out to identify and predict sources of error. 3.With newer systems, controllers （ ） ．