Agents of Value

Agents, be they human or technological, act on behalf of others in the world. Autonomous software agents may also act as representations (as in simulations) of human concerns. To do so both effectively and ethically, these agents must in some meaningful way understand and account for what those for whom they act consider important in life. At stake is no less than the risk of designing technologies that undermine important human values. Yet, with our limited view, it is not at all obvious how to design agent technologies so that they are more likely to support the actions, relationships, institutions, and experiences that human beings care deeply about. In this invited speech I will explore the question of how to design agent technology to be sensitive to human values and concerns. I will draw from over two decades of design work and theory development in Value Sensitive Design. Along the way, I will touch on how agent technologies instantiate human values, the locus of value tensions in multi-agent systems, and key value tensions. From there, I will turn to methods from value sensitive design-direct and indirect stakeholder analyses, value scenarios, Envisioning Cards-which can be used to improve sensitivity to human values in agent technology research and design work. I will end with a few research challenges for which cutting edge agent technology design has provocative implications for human futures. Throughout, I will take an interactional stance: that agent technologies shape human experience and our very being; and through our experiences and being, we continually re-imagine those very technologies.