Optimize

Abstract

AbstractThe fourth step of the prediction supply chain enacts priorities. Algorithms make choices efficiently using explicit or implicit concepts of precedence. Each prediction is optimized based on the importance of data, populations, comparisons, or values. Applying mathematical standards of consistency and perfection in human affairs can go awry. Because algorithmic optimization prioritizes specific populations and situations, unforeseen cases could spectacularly fail. An algorithm can overoptimize, meaning it focuses so much on one goal that it ignores other equally important goals not specifically articulated. Data science implemented over large systems at scale eventually must consider the secondary effects of optimized outcomes. Ethical data science is able to state the reasoning that drives choices and is aware of alternatives and trade-offs. This chapter encourages data science reasoning.