The need for ethics more than moral. The development of artificial autonomous agents raises today ethical issues. Indeed, users sometimes have ethical expectations other than optimality or legality issues, for which artificial agents are mostly designed. This is why there is an interest in equipping intelligent and autonomous machines with ethical behaviours. However why questions of ethics and not of morality? If the two words etymologically designate the same thing, the use in France differentiates them. According to Gilles Deleuze "morality is presented as a set of binding rules of a special type, which consists in judging actions and intentions by relating them to transcendent values ​​(it is good, it is bad, etc.). ; Ethics is a set of optional rules that evaluates what we do and say according to the way of existence it implies". In this context, the ANR ETHICAA project aims to answer the following two questions. What elements make it possible to appreciate a situation involving ethical issues and autonomous agents? What are the formal tools for reasoning and implementing ethical principles in autonomous agents?

An ethical reasoning framework. The ETHICAA project proposes an ethical reasoning framework implemented avec logic programming languages under a Java layer, allowing artificial agents to evaluate their environment, to integrate ethical principles and, from the implementation of these principles, to compute either action plans, or evaluations of the behaviour of the other agents. This framework is based on two explicit distinctions as well as a set of features. The first distinction is between morality and ethics, namely between reasoning about good and evil, and reasoning about right and wrong. The second distinction is between individual and collective ethics, allowing an agent to take into account the plurality of moral values and ethical principles of other agents. Finally, our framework provides a set of functionalities: perceiving situations of dilemma, attributing causality and responsibilities, judging, deciding and acting according to ethical principles, collaborating and trusting other agents, and finally being able to formally verify the ethical behaviour of an agent.

Major results of the project. By implementing ethical decision-making through symbolic reasoning over by-design or machine learning approaches, the ETHICAA project framework allows to propose reasoned and justified decisions to an operator or user. Beyond these technical results, the project ETHICAA - through its members - plays a role in several ethical committees (CERNA, COERLE, COMETS, IEEE) and has co-organized since 2015 three days Ethics & Artificial Intelligence at the French AFIA Artificial Intelligence Platforms (PFIA), contributing to the emergence of this research theme at the French national level.

Scientific productions since the beginning of the project. The work carried out within the ETHICAA project has been published many times: 2 international journals and 16 international papers, and 8 national journals and 9 national papers. Beyond this scientific production, the project has focused on vulgarisation - as ethical reflection must go beyond the research community and find echo in civil society - with twenty articles or general public communications on the subject, as well as participations in summer schools, doctoral training courses, round tables and multidisciplinary colloquia.