On Understanding the Influence of Controllable Factors with a Feature Attribution Algorithm: a Medical Case Study
This work addresses a specific issue in explainable AI for medical applications, offering an incremental improvement by partitioning features into controllable and uncontrollable categories.
The paper tackled the problem of feature attribution algorithms treating all features homogeneously, which can cause misinterpretations, by proposing CAFA to compute the importance of controllable features separately; results showed that CAFA excludes uncontrollable feature influences in explanations while using the full dataset for prediction.
Feature attribution XAI algorithms enable their users to gain insight into the underlying patterns of large datasets through their feature importance calculation. Existing feature attribution algorithms treat all features in a dataset homogeneously, which may lead to misinterpretation of consequences of changing feature values. In this work, we consider partitioning features into controllable and uncontrollable parts and propose the Controllable fActor Feature Attribution (CAFA) approach to compute the relative importance of controllable features. We carried out experiments applying CAFA to two existing datasets and our own COVID-19 non-pharmaceutical control measures dataset. Experimental results show that with CAFA, we are able to exclude influences from uncontrollable features in our explanation while keeping the full dataset for prediction.