CASMI Prime Seminar with Jim Guszcza: "Envisioning a field of human-machine hybrid intelligence architecture"
May 6, 2022
AI promises to improve human decision-making and fuel economic growth, but today’s state of the art suffers from serious shortcomings. Reports abound of AI technologies that compromise human safety or wellbeing, treat people unfairly, amplify societal biases, undermine human autonomy through manipulation, and amplify the spread of misinformation and polarizing content. On a purely practical level, Gartner has estimated that 85% of big data projects never make it to production.
This talk explores the premise that many of today’s AI shortcomings arise from a basic mismatch. On the one hand, business leaders and end-users typically require effective systems of human-machine partnership. On the other hand, the teams that build AI technologies tend to be oriented towards optimizing algorithms in lab settings. Currently missing, and needed, is an applied field for developing human-machine hybrid intelligence systems. Such a field would be multidisciplinary, integrating concepts and methods from computer science and data science with those from the humanities and social sciences. For example, concepts such as confirmation bias, algorithm aversion, and choice architecture would be no less essential to the language of hybrid intelligence development than concepts such as label bias, data drift, and cross-validation.
Back to top