Mitigating Cognitive Biases in Machine Learning Algorithms for Decision Making

Abstract

Cognitive biases are an ingrained part of the human decision-making process. Nearly all machine learning algorithms that mimic human decision-making use human judgments as training data, which propagates these biases. In this paper, we conduct an empirical study in which 150 applicants are rated for suitability for three separate job openings. We develop an algorithm that learns from human judgments and consequently develops biases based on these human-generated inputs. Next, we explore and apply techniques to mitigate these algorithmic biases, using a combination of pre-processing, in-processing, and post-processing algorithms. The results from our study show that biases can be mitigated using these approaches but involve a tradeoff between complexity and effectiveness.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 22 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.