Proposed Computational Classification System of Human Cognitive Biases

Abstract

Despite our aspirations to do so, we humans don't always make optimal or rational decisions. Researchers from psychology, behavioral economics, anthropology, decision sciences, and other related fields have described many human cognitive biases which help to explain such decisions. Most of the time, these cognitive biases are relatively harmless and relatively costless. However, sometimes they do result in significant costs to individuals, companies, governments and societies in the form of wasted or misdirected money, time, effort, and sometimes even in the form of lives lost. The antidote to such decisions has long been recognized to lie in algorithmic decision making. Until relatively recently, though, requirements and complexity of such algorithms have limited their deployment in real-world situations. However, we now enjoy a convergence of computing power, decrease in computing costs, and computational and predictive methods born of data science, artificial intelligence (AI), and machine learning (ML), such that we can begin to mitigate some of the most negative effects of some of these cognitive biases. This paper proposes a method for classifying these human cognitive biases for purposes of mitigation by means of computing methods, describes some of these biases that are most ripe for mitigation through computing, and proposes future research directions that build upon this work.