Comprehensible classification models: a position paper

By Alex Alves Freitas. 2014

656 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Information Systems, Safety, Robustness & Interpretability

How researchers see this paper

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

In Artificial Intelligence

In Information Systems

In Safety, Robustness & Interpretability

Citations in the corpus

Interpretable Machine Learning: Fundamental Principles and 10 Grand Challenges (2021, background): “Elomaa (1994) and Freitas (2014) make the point that “Humans by nature are mentally opposed to too simplistic representations of complex relations.””
Explanation in Human-AI Systems: A Literature Meta-Review, Synopsis of Key Ideas and Publications, and Bibliography for Explainable AI (2019, background): “systems. In addition, several researchers have identified specific taxonomic measurement frameworks or recommendations for assessment. (Doshi-Velez and Kim, 2017; J. Doyle, Radzicki, and Trees, 2008; Freitas, 2014; Groce et al., 2014; Kulesza et al., 2011; Kulesza, Stumpf, Burnett, and Kwan, 2012; T”
Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead (2018, background): “Interpretability is a domain-specific notion [4, 5, 6, 7], so there cannot be an all-purpose definition.”
Please Stop Explaining Black Box Models for High Stakes Decisions (2018, background): “Interpretability is a domain-speciﬁc notion [Freitas, 2014, Kodratoff, 1994, Huysmans et al., 2011, Rüping, 2006], so there cannot be an all-purpose deﬁnition.”
SmoothGrad: removing noise by adding noise (2017, background): “For reviews of inter-pretable models, see (Freitas, 2014; Doshi-Velez, 2017).”
On the Safety of Machine Learning: Cyber-Physical Systems, Decision Sciences, and Data Products (2016, background): “Two related ways to introduce inherently safe design is by insisting on models that can be interpreted by people and by excluding features that are not causally-related to the outcome [27]– [30].”
Model-Agnostic Interpretability of Machine Learning (2016, background): “This trade-off between model flexibility and interpretability (Freitas, 2014) implies one cannot use a model whose behavior is very complex, yet expect humans to fully comprehend it globally.”
Interpretable Two-level Boolean Rule Learning for Classification (2015, background): “The two-level Boolean rules in this work are examples of sparse decision rule lists, one of the major classes of interpretable models [4].”
The Bayesian Case Model: A Generative Approach for Case-Based Reasoning and Prototype Classification (2014, methodology): “The cognitive load of the user also increases with the complexity of the similarity measure used for comparing cases [14].”
Learning Certifiably Optimal Rule Lists (2017, background): “Interpretable Models: There is a growing interest in interpretable (transparent, compre-hensible) models because of their societal importance (see R¨uping, 2006; Bratko, 1997; Dawes, 1979; Vellido et al., 2012; Giraud-Carrier, 1998; Holte, 1993; Shmueli, 2010; Huys-mans et al., 2011; Freitas, 2014).”
Mathematical optimization in classification and regression trees (2021, background): “They are conceptually simple, show excellent learning performance, are computationally cheap, and routines and packages to train them are available in popular languages such as Python and R, and are also appealing in terms of interpretability (Freitas 2014; Hu et al.”
Neural Prototype Trees for Interpretable Fine-grained Image Recognition (2020, background): “In contrast, decision trees are easy to understand and interpret [14, 19], because they transparently arrange decision rules in a hierarchical structure.”
Learning customized and optimized lists of rules with mathematical programming (2018, background): “The need for classifiers that can be adjusted and customized is also clear from the literature (see for instance, [21]).”
Boolean Decision Rules via Column Generation (2018, background): “While the latter two classes are also considered interpretable, the metrics for measuring their complexity are different and not directly comparable [27].”
An optimization approach to learning falling rule lists (2017, background): “Attempts have also been made to underline the importance of the interpretability or the comprehensibility of models [9, 13, 14, 18].”