ETHICs and Autonomous Agents – ETHICAA

The ethical conflicts are characterized by the fact that there is no « good » way to solve them, and when a decision must be made it should be an informed decision based on an assessment of the arguments and values at stake. Broadly speaking, three important components have to be considered to define a conflict management framework dealing with ethical conflicts in agent systems: (1) definition of an ethical reasoning framework including the representation of several ethical principles and situation assessment; (2) definition of ethical conflicts detection methods that can tackle situations where agents reason individually or collectively; (3) definition of multiple ethical decision-making models to manage ethical conflicts in smartly combining different ethical principles into a multi-point-of-view ethical decision-making framework.

As ethics is an individual notion shaped on culture, context and personal experiences, ethical decision-making only makes sense in a given application. The ETHICAA project proposes to instanciate its framework on two applicative domains: robotics and privacy management. These two domains have been chosen for their impact on society in terms of trust, acceptance and ethics. In the robotic domain, the ETHICAA framework will be used to manage ethical conflicts arising in the cooperation between one artificial agent and one human operator in the context of the joint operation of an UAV (Unmanned Air Vehicle). In the case of privacy management, ethical conflicts may arise between multiple artificial agents and human users, and we will consider a social network where ethical autonomous agents control and manage the privacy policies of the accounts owned by humans.

We proposed various practical case studies of ethical conflicts and highlighted what their main system and decision features are. We provided also a state-of-the-art on moral philosophy and human (and moral) values. From this work, we studied several models of autonomous agents that embed notions of values, moral rules and ethical principles. Each of those model aims at being integrated in an autonomous agent's design and life cycle: behavioral verification, supervision, reasonment, judgment and explanation.

The results targeted by the ETHICAA project are (i) formal representation of ethical principles, values and conflicts, as well as (ii) automatic reasoning models to handle these ethical representations, (iii) all being validated on real world scenarios. Consequently the outcome of ETHICAA is composed of a framework and recommendations to design ethical systems of artificial agents, i.e. able to execute ethical regulation modes depending of the context and ethical conflicts involving other agents or humans.

As a pluridisciplinary project, the ETHICAA project produced 15 international communications and 11 national communication, in Artificial Intelligence conferences (AAMAS, LPAR) as well as Human Sciences conferences (ISPSO) and pluridisciplinary conferences (CEPE). The ETHICAA projet has also lead to around twenty articles and lectures for the general public.

Machines and software acting on behalf of humans (i.e. agents) get more autonomy and are increasingly less under the control of human operators or users. Therefore especially when machines interact with humans, we need to ensure that they do not harm them or threaten their autonomy. Consequently the question of an ethical regulation and control of such autonomous agents is raised and has been discussed in the literature, stressing the need of ethical standards.

The objective of the ETHICAA project is to define what should be an autonomous entity that could manage ethical conflicts, considering both the individual agent and the multi-agent levels. The ethical conflicts that will be considered are those occurring 1) inside one agent, 2) between one agent and the ethical principles and values of the system it belongs to, 3) between one agent and a human operator or user, 4) between several agents including or not human beings. Ethical conflicts are characterized by the fact that there is no « good » way to solve them. Nevertheless when a decision must be made it should be an informed decision based on an assessment of the arguments and values at stake. The models, methods and tools developed within the project will contribute to define the ETHICAA framework for the definition and management of agent systems featuring ethical behaviours.

ETHICAA proposes to base the achievement of this objective on two applicative domains: robotics and privacy management. These two domains have been chosen for their impact on society in terms of trust, acceptance and ethics. In the robotic domain, the ETHICAA framework will be used to manage ethical conflicts arising in the cooperation between one artificial agent and one human operator in the context of the joint operation of a UAV (Unmanned Air Vehicle). In the case of privacy management, ethical conflicts may arise between multiple artificial agents and human users inside a social network where ethical autonomous agents control and manage the privacy policies of the accounts owned by humans.

The results targeted by the ETHICAA project are (i) formal representation of ethical principles, values and conflicts, as well as (ii) automatic reasoning models to handle these ethical representations, (iii) all being validated on real world scenarios. Consequently the outcome of ETHICAA is composed of a framework and recommendations to design ethical systems of artificial agents, i.e. able to execute ethical regulation modes depending of the context and ethical conflicts involving other agents or humans.