Stop Explaining Black Box Models And Use Interpretable Models Instead

Stop Explaining Black Box Models And Use Interpretable Models Instead This perspective clarifies the chasm between explaining black boxes and using inherently interpretable models, outlines several key reasons why explainable black boxes should be. We will discuss reasons why so many people appear to advocate for black box models with separate explanation models, rather than inherently interpretable models – even for high stakes decisions.

Cynthia Rudin Stop Explaining Black Box Machine Learning Models High Recent work on the explainability of black boxes—rather than the interpretability of models—contains and perpetuates critical misconceptions that have generally gone unnoticed, but that can have a lasting negative impact on the widespread use of ml models in society. A preliminary version of this manuscript appeared at a workshop, entitled “please stop explaining black box machine learning models for high stakes decisions” [13]. The chasm between explaining black boxes and adopting inherently interpretable models is clarified and it is demonstrated how interpretable hybrid models could potentially supplant black box ones in different domains. This manuscript clarifies the chasm between explaining black boxes and using inherently interpretable models, outlines several key reasons why explainable black boxes should be avoided in high stakes decisions, identifies challenges to interpretable machine learning, and provides several example applications where interpretable models could.

Stop Explaining Black Box Machine Learning Models For High Stakes The chasm between explaining black boxes and adopting inherently interpretable models is clarified and it is demonstrated how interpretable hybrid models could potentially supplant black box ones in different domains. This manuscript clarifies the chasm between explaining black boxes and using inherently interpretable models, outlines several key reasons why explainable black boxes should be avoided in high stakes decisions, identifies challenges to interpretable machine learning, and provides several example applications where interpretable models could. Instead of trying to develop tools that explain the machine learning models, we should all be working towards developing models that are inherently explainable. The two main takeaways from this paper: first, it underscores the difference between explainability and interpretability and presents why the former may be problematic. second, it provides some great pointers for creating truly interpretable models. It can be argued that instead efforts should be directed at building inherently interpretable models in the first place, in particular where they are applied in applications that directly affect human lives, such as in healthcare and criminal justice. There has been a recent rise of interest in developing methods for ‘explainable ai’, where models are created to explain how a first ‘black box’ machine learning model arrives at a specific decision.