Ethical Considerations for the Military Use of Artificial Intelligence in Visual Reconnaissance

Artificial Intelligence (AI) and related technologies in machine and deep learning are making their way into all of our lives: intelligent dialog systems and smartphones powered by AI simplify everyday life, self-driving vehicles with AI-based signal processing and control increase the level of automation, and AI is considered as one of the building blocks of the fourth industrial revolution. However, while AI in general becomes more powerful and widely used, the impact on humans and their well-being must be considered. It can thus no longer be sufficient to just technically enable AI to solve a given task. Its use has to follow guidelines for human values and ethical principles. This implication not only holds for the general use of AI, but it is especially relevant for military applications. In this white paper, we analyse and discuss how the responsible use of AI in military scenarios can be achieved with compliance to ethical and legal guidelines. In Section 2, we first review existing literature that broadly discusses ethical guidelines for responsible AI in the civil and the military domain. Based on these findings, we derive a consolidated set of ethical principles that reflects our opinion towards an ethically compliant integration of AI-powered assistance functions in military applications. Related work especially in the military domain currently focuses on rather theoretical considerations of this topic [4, 23]. In Sections 3, 4 and 5, however, we describe three specific use cases, where AI-powered assistance functions are utilized in military applications. Each specific scenario is described along with a technical point of view that shows the feasibility of the AI-based functions with current state-of-the-art technology. Furthermore, our derived ethical principles are mapped to specific observations and recommendations of the AI-powered system as well as interactions with the user, like human-in-the-loop and human-on-the-loop principles. In Section 6, we conclude that a careful design and development of AIpowered assistance systems together with suitable integrated methods of eXplainable Artificial Intelligence (XAI) can support the human user substantially and it can be well-aligned with ethical principles. Especially in military applications, the human-in-the-loop shall still be the final and sole decision-maker.