Workshop on Ethics in the Design of Intelligent Agents (EDIA16)

The development of Artificial Intelligence is experiencing a fruitful period of incredible progress and innovation. After decades of notable successes and disappointing failures, AI is now poised to emerge in the public sphere and completely transform human society, altering how we work, how we interact with each other and our environments, and how we perceive the world. Designers have already begun implementing AI that enables machines to learn new skills and make decisions in increasingly complex situations. The result is that these machines - also called intelligent agents - decide, act and interact in shared and dynamic environments under domain constraints, where they may interact with other agents and human beings to share tasks or execute tasks on behalf of others. Search engines, self-driving cars, electronic markets, smart homes, military technology, software for big data analysis, and care robots are just a few examples.

As intelligent agents gain increased autonomy in their functioning, human supervision by operators or users decreases. As the scope of the agents’ activities broadens, it is imperative to ensure that such socio-technical systems will not make irrelevant, counter-productive, or even dangerous decisions. Even if regulation and control mechanisms are designed to ensure sound and consistent behaviors at the agent, multi-agent, and human-agent level, ethical issues are likely to remain quite complex, implicating a wide variety of human values, moral questions, and ethical principles. The issue is all the more important as intelligent agents encounter new situations, evolve in open environments, interact with other agents based on different design principles, act on behalf of human beings and share common resources. To address these concerns, design approaches should envision and account for important human values, such as safety, privacy, accountability and sustainability, and designers will have to make value trade-offs and plan for moral conflicts. For instance, we may want to design self-driving cars to exhibit human-like driving behaviors, rather than precisely following road rules, so that their actions are more predictable for other road users. This may require balancing deontic rule-following, utility maximization, and risk assessment in the agent's logic to achieve the ultimate goal of road safety.

Questions to be asked here are: How should we encode moral behavior into intelligent agents? Which ethical systems should we use to design intelligent, decision-making machines? Should end-users have ultimate control over the moral character of their devices? Should an intelligent agent be permitted to take over control from a human operator? If so, under what circumstances? Should an intelligent agent trust or cooperate with another agent embedded with other ethical principles or moral values? To what extent should society hold AI researchers and designers responsible for their creations and choices?

This workshop focuses on two questions: (1) what kind of formal organizations, norms, policy models, and logical frameworks can be proposed to deal with the control of agents' autonomous behaviors in a moral way?; and (2) what does it mean to be responsible designers of intelligent agents?

The workshop welcomes contributions from researchers in Artificial Intelligence, Multi-Agent Systems, Machine Learning, Case-based reasoning, Value-based argumentations, AI and Law, Ontologies, Human Computer Interaction, Ethics, Philosophy, and related fields.

The topics of interest include (but are not limited to):