Gaining insight through Case-Based Explanation

Traditional explanation strategies in machine learning have been dominated by rule and decision tree based approaches. Case-based explanations represent
an alternative approach which has inherent advantages in terms of transparency and
user acceptability. Case-based explanations are based on a strategy of presenting
similar past examples in support of and as justification for recommendations made.
The traditional approach to such explanations, of simply supplying the nearest
neighbour as an explanation, has been found to have shortcomings. Cases should be
selected based on their utility in forming useful explanations. However, the relevance
of the explanation case may not be clear to the end user as it is retrieved using
domain knowledge which they themselves may not have. In this paper the focus is
on a knowledge-light approach to case-based explanations that works by selecting
cases based on explanation utility and offering insights into the effects of featurevalue differences. In this paper we examine to two such knowledge-light frameworks
for case-based explanation. We look at explanation oriented retrieval (EOR) a
strategy which explicitly models explanation utility and also at the knowledge-light
explanation framework (KLEF) that uses local logistic regression to support casebased explanation.