In our review of human cognitive biases, stability biases featured prominently. It turns out that humans are not alone: algorithms also often exhibit stability biases. In this chapter, we will review the most important ones and explore what kind of context attributes promote them.
We first will examine the fundamental cause of the problem, namely system instability—without it, there would be no harmful stability bias. After that, we will discuss the conceptual dilemma that it is close to impossible to teach things to an algorithm that don’t exist in the ...