In a previous blog post, I discussed the three main properties which make a statistical model an (intrinsically) “interpretable” model. In this blog post, I want to talk about the other side of the interpretability-explainability divide, and discuss the methods used for post-hoc explanation of generic blackbox models. To begin, we start by recalling the…

~Some Brief Opinions after the NeurIPS ‘25 Interpretability Workshop~ I again wanted to share some thoughts on the field of interpretability, this time under the pretense of responding to the “Mechanistic Interpretability” workshop at NeurIPS 2025.  Mostly, I want to discuss the rising tide of mechanistic interpretability and ask the question: Mechanistic? As the title…

I love the field of interpretability, but one issue faced by everyone who tries dipping their toes into interpretability is: “What is Interpretability?”.  There never seems to be a universally agreed-upon definition for interpretability.  Much like philosophy, this often leads to disagreements over the definitions, fights over the contexts, and arguments over the objectives.  Interpretability…

~Some Brief Thoughts after the ICML ‘25 Interpretability Workshop~ I have told many people multiple times that I will write a blog post and I keep not ending up with my words written down on a website. I have recently been advised that if I just write down my first thoughts without any attempts to…