Interpretability of machine learning models and representations: an introduction

By Adrien Bibal, Benoît Frénay. 2016

119 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Legal requirements on explainability in machine learning (2020, methodology): “On the contrary, black-box models are models with a complex mathematical expression that, moreover, do not possess a representation that can ease their understanding (Bibal and Frénay 2016).”
From What to How: An Initial Review of Publicly Available AI Ethics Tools, Methods and Research to Translate Principles into Practices (2019, background): “Social scientists (Matzner 2014) and political philosophers (from Rousseau and Kant, to Rawls and Habermas) (Binns 2018b), are used to dealing with the kind of plurality and subjectivity informing the entire ethical ML field (Bibal and Frénay 2016).”
Uncertainty Estimation in Bayesian Neural Networks And Links to Interpretability (2018, methodology): “Such restrictions may reduce accuracy due to lesser model complexity, but accuracy is also important to avoid generating trivial explanations which are easy to understand yet yield no connection to the data [54, 5].”
From What to How. An Overview of AI Ethics Tools, Methods and Research to Translate Principles into Practices (2019, background): “This issue stems from the fact that the entire field is imbued with subjectivity (Bibal & Frénay, 2016) and the relative success of ethical-alignment—from ideation through to operation at scale—is also context-dependent.”
Explainable Artificial Intelligence: a Systematic Review (2020, methodology): “The same dual categorisation system was suggested in [64], but they named the two classes heuristic-based and user-based metrics.”
Defining Explainable AI for Requirements Analysis (2018, unknown): “For example, [30] refer to this as “Fidelity” and [3] as “Accuracy”, while the “Opaque” category [7] corresponds to Post-Hoc Rationalisation in our vernacular.”
Machine Learning Interpretability: A Survey on Methods and Metrics (2019, methodology): “Although there is no consensus on how to exactly measure interpretability [70,151], there is still room to define what are the goals for which interpretability aims.”
Evaluating Explanations by Cognitive Value (2018, unknown): “There are several other impressive works across various ﬁelds catered towards helping the AI engineers [11, 13, 15, 16, 29–32, 40, 38].”
Interpretability of machine learning‐based prediction models in healthcare (2020, background): “Several terms can be found describing the interpretability and related concepts, such as the multiplicity of good models or model locality (Breiman, 2001; Hall & Gill, 2018), comprehensibility or understandability (Piltaver, et al., 2016), and mental fit or explanatory ML (Bibal & Frenay, 2016).”
Inverse Classification for Comparison-based Interpretability in Machine Learning (2017, background): “Existing categorizations of inter-pretability (Bibal 2016; Kim and Doshi-Velez 2017; Doshi-Velez and Kim 2017; Biran and Cotton 2017) usually distinguish approaches mainly on the characteristics of these explanations: given a classiﬁer to be interpreted, in-model interpretability relies on…”