The Rise of Autonomous Experimentation: Technical, Social, and Ethical Implications of AI

In the field of computer science, large-scale experimentation on users is not new: there have been many efforts in both the public and private sectors to analyze users and to create experimental conditions to provoke changes in their behavior. However, new autonomous and semi-autonomous systems for experimentation, driven by techniques from AI and machine learning, raise important questions for the field. Many of these questions are about the social and ethical implications of these systems. We see these questions as urgent, because they pertain to infrastructural changes in the computing environment that large populations depend upon, and because their impacts extend far into the lived experience of everyday life. Yet, despite this urgency, there has been very little by way of public debate, or even debate within the scientific community, about how these experiments are designed, how they affect individuals and communities, and what the long-term ramifications may be. We aim to define and characterize these systems, and consider the scientific and social questions that they raise.