Abstract | ||
---|---|---|
Algorithmic bias due to underestimation refers to situations where an algorithm under-predicts desirable outcomes for a protected minority. In this paper we show how this can be addressed in a case-based reasoning (CBR) context by a metric learning strategy that explicitly considers bias/fairness. Since one of the advantages CBR has over alternative machine learning approaches is interpretability, it is interesting to see how much this metric learning distorts the case-retrieval process. We find that bias is addressed with a minimum impact on case-based predictions - little more than the predictions that need to be changed are changed. However, the effect on explanation is more significant as the case-retrieval order is impacted. |
Year | DOI | Venue |
---|---|---|
2022 | 10.1007/978-3-031-14923-8_4 | International Conference on Case-Based Reasoning |
DocType | Citations | PageRank |
Conference | 0 | 0.34 |
References | Authors | |
0 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Blanzeisky William | 1 | 0 | 0.34 |
Barry Smyth | 2 | 5711 | 414.55 |
Pádraig Cunningham | 3 | 3086 | 218.37 |