Researchers at Massachusetts Institute of Technology (MIT) have published a paper titled “The Need for Interpretable Features: Motivation and Taxonomy” that aims to bridge the explanation gap between machine learning models and non-technical staff.
The paper discusses how to explain and interpret machine learning models in a way that is inclusive to all types of users, by building explainability into the model from the start. Key take aways from the paper:
- Improved trust: with the implementation of the taxonomy, end users will be better inclined to trust the decisions made by ML models.
- Enhanced understanding: the paper covers best practices for converting complex ML models into easy-to-understand formats, making it easier for non-technical users to understand what factors the AI used to come to a certain observation or decision.
- Faster time-to-tarket: the next step for the MIT researchers is developing a system for faster feature to format transformations, which should improve the time-to-market for ML models.
For more insights, study the article.