I was going through a pre-print on arXiv named "Quantifying interpretability and trust in machine learning systems". There, I found that a comparison of two interpretability approaches - LIME and COVAR was done based on their ITR(Information Transfer Rate) defined as
and continued in
I am having difficulty in understanding how this information is defined for humans and for the machine learning algorithm and hence the mutual information in the formula provided. I'm not even sure if I've asked correctly but it's not clear to me how are these 2 algorithms being compared using the formula mentioned above.