Human cognitive input for the CEM

In this section, we will use our human cognitive abilities to pick two key features out of tens of features inside a minute and solve a problem.

In Chapter 9, The Counterfactual Explanations Method, we used WIT to visualize the counterfactuals of data points. The data points were images of people who were smiling or not smiling. The goal was to predict the category in which a person was situated.

We explored Counterfactual_explanations.ipynb. You can go back and go through this if necessary. We found that some pictures were confusing. For example, we examined the following images:

Figure 12.3: WIT interface displaying counterfactual data points

It is difficult to see whether the person on the left is smiling.

This leads us to find a cognitive explanation.

Rule-based perspectives

Rule bases can be effective when machine learning or deep learning models reach their explainable AI limits.

In this section, we will...