Teaching Engineers Diplomacy, and Other Lessons for Machine Learning

Abstract

Advancements in capabilities for machine learning and artificial intelligence (MLAI) has led to growing questions and challenges for effective sociotechnical MLAI applications. It is important to consider applications outside of consumer product development, particularly in the realms of local, regional, national, and international policy development. There are intriguing opportunities to apply MLAI techniques to improving the efficiency of standard government (“bureaucratic”) activities, including policy and diplomatic operations. Growing sociotechnical concerns address the limitations of applying machine learning tools to digital content, due in part to the unintentional (or even explicit) embedding of bias and prejudice into MLAI algorithms that become more difficult for others (particularly those without computer science training) to detect, correct, or reverse. The author’s experience provides context to consider issues of MLAI involvement in policy creation in a more subtle operational context. Many MLAI applications are built on the analysis of an existing corpus of outcome products; this reasoning might be applied to the analysis of international policy and standards documents. This paper addresses two challenges to that approach. One is the difference between engineers and policymakers on the nature of debate, evidence, and conflict resolution. A second difference addresses, from the author’s experience, emphasis on informal and interim processes, rather than final products, in the development of mutually agreed outcomes in national and international policymaking efforts.