The ability of artificial intelligence to generate data-driven insights and recommendations has potential benefits at an unprecedented scale, particularly for underserved and marginalized communities limited in their access to human expertise and resources. However, together with this promise, AI brings new risks and presents a real danger to deepen existing inequalities, reinforce injustices, and create deeper disconnects within societies.
We argue that many of the current limitations and risks of AI stem from a growing disconnect between the technology-centric approach to the creation of AI technologies and its inextricable embedding into complex socio-cultural contexts. Intelligent interactive systems, systems that support humans in their decision-making processes, are socio-technical systems; their ultimate impact depends on the combined behavior of machine and human intelligence, and their design requires both technological sophistication and deep understanding of human cognition, behaviors, norms, and values.
We propose a “human first” approach: the creation of AI systems, where the human perspective and needs drive the technological innovations, throughout all stages of the systems’ design (data collection, learning models, inference strategies, interaction paradigms, validation, deployment, evaluation, and maintenance). Delivering on this vision will require integrating disciplines that have traditionally been divorced from one another as well as overcoming challenges too formidable to address within the boundaries of individual projects or existing disciplines.
Beth Mynatt, Executive Director, Institute for People and Technology and Distinguished Professor, College of Computing