Assistant professor in Computer Science and Artificial Intelligence

Olivier Bartheye is an assistant professor in computer science at the French Air Force and Space Academy at Salon-de-Provence since 2019. His research concerns explainable artificial intelligence, symbolic artificial intelligence, autonomous agents, machine decision, action planning, representations of sense-making and very recently quantum computing. Formerly, he was an assistant professor in computer science at the French Ground Force Academy in Brittany, in Guer-Coëtquidan from 2001 to 2019. After his PhD about action planning, carried out from 1989 to 1994, at the Central Commission of Nuclear Energy, he started in 1995 his industrial carrier as a software engineer until 2001.He was coordinator of three international research and innovation projects, including two in cybersecurity. He has also written a prototype action plan and is continuing his research into automated decisions to determine what exactly can be delegated to a machine regarding the decision-making process.

Current area of research : In the military field, artificial intelligence techniques must be used according to a special category of use-cases: AI for operational critical systems which are very different from biology use-cases or medicine use-cases. The specific business rules for operational critical systems need to emphasize the notion of trust on the result performed by an AI component to perform a decision. The trust process concerns the contractualization of human-machine interactions.

Trust is ensured by explainable AI or sensemaking AI, or equivalently, understandable results resulting from the calculation of any type of machine learning process.Thus, predictive models can feed action models and conversely, action can feed prediction since a necessary action plan introduces new situations to learn after the execution of the action plan.To connect the operational part and the machine learning part, it is necessary to determine the best candidate machine learning model which would ultimately enable a sequence of actions to be initiated, whether automatic or not.Faced with the wave of digitalization in society, we have the capacity, thanks to this duality, to answer questions concerning the automation of IT processes and the balancing of roles between humans and machines on critical or sensitive tasks.

By increasing the degree of autonomy of machines as virtual agents, we can compensate for human failures, but this autonomy must be controlled and validated to be included in autonomous devices.The validation effort located precisely on the human-machine interface requires explaining the role of the artificial intelligence components and at the same time explaining the result of the automatic learning process.This is a very broad problem of data management and distribution of roles specific to autonomous system components.These subjects concern the notions of risk analysis, definition of requirements, contractualization in the distribution of roles, rigorous specification and formalized scenarios to clearly delimit these roles.In addition, you must be able to associate the notion of triggering action with the analysis of the situation. In general, feedback loops are complex process sequences due to the high variability of possible scenarios.