AI Compass

Drawing on well-established methods for human–system integration, such as KOMPASS, which aims at complementary function allocation in automated production systems, a new design framework, AI (artificial intelligence) Compass, is proposed that translates basic principles of socio-technical system design and human–system integration into the context of AI-based technologies. Besides reaffirming the fundamental design principle of aligning control and accountability for handling variances and disturbances at their source in light of growing AI capabilities, the framework postulates establishing networks of accountability and empowering AI developers and users to manage task interdependencies as two additional design principles jointly. These design principles account for AI’s expanding agency, the broader range of stakeholders involved in the AI supply chain, and the increasing interdependencies between AI development and use. Building on and expanding existing design methods, processes, and criteria are outlined, which follow the three design principles and provide basic guidelines for practitioners involved in AI development and use.