One Explanation Does Not Fit All: A Toolkit and Taxonomy of AI Explainability Techniques

By Vijay Arya, R. Bellamy, Pin-Yu Chen, Amit Dhurandhar, Michael Hind, Samuel C. Hoffman, Stephanie Houde, Q. Liao, Ronny Luss, Aleksandra Mojsilovic and 10 more. 2019

478 citations recorded by the source corpus.

As artificial intelligence and machine learning algorithms make further inroads into society, calls are increasing from multiple stakeholders for these algorithms to explain their outputs. At the same time, these stakeholders, whether they be affected citizens, government regulators, domain experts, or system developers, present different requirements for explanations. Toward addressing these needs, we introduce AI Explainability 360 (this http URL), an open-source software toolkit featuring eight diverse and state-of-the-art explainability methods and two evaluation metrics. Equally important, we provide a taxonomy to help entities requiring explanations to navigate the space of explanation methods, not only those in the toolkit but also in the broader literature on explainability. For data scientists and other users of the toolkit, we have implemented an extensible software architecture that organizes methods according to their place in the AI modeling pipeline. We also discuss enhancements to bring research innovations closer to consumers of explanations, ranging from simplified, more accessible versions of algorithms, to tutorials and an interactive web demo to introduce AI explainability to different audiences and application domains. Together, our toolkit and taxonomy can help identify gaps where more explainability methods are needed and provide a platform to incorporate them as they are developed.

Subfields: Artificial Intelligence, Information Systems and Management

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 and Management

Citations in the corpus

Questioning the AI: Informing Design Practices for Explainable AI User Experiences (2020, background): “…in social science on how humans give and receive explanations, Miller identiﬁed a list of human-friendly characteristics of explanation that are not given sufﬁcient attention in the algorithmic work of XAI, including contrastiveness (to a counterfactual case), selectivity, social process,…”
Evaluating the Quality of Machine Learning Explanations: A Survey on Methods and Metrics (2021, methodology): “Table 1 shows examples of associations between explanation types from Reference [43] and explanation technique categories from Reference [10].”
From Anecdotal Evidence to Quantitative Evaluation Methods: A Systematic Review on Evaluating Explainable AI (2022, methodology): “Since we could not find a complete and recent overview of explanation types, we reviewed recent XAI surveys [16, 59, 171, 287] and adapted and extended the categories identified by Guidotti et al.”
The Metacognitive Demands and Opportunities of Generative AI (2023, unknown): “Prospective self-ratings of confidence, SRL microanalysis”
Expanding Explainability: Towards Social Transparency in AI systems (2021, background): “Recent work also emphasizes that explainability is an audience-dependant instead of a model-inherent property [7, 8, 26, 70, 73].”
SurvNAM: The machine learning survival model explanation (2021, background): “The importance of the explanation problem leads to developing many methods and models which try to explain and to interpret predictions of the machine learning models [2, 3, 4, 5, 6, 7, 8, 9, 10].”
Explainable Artificial Intelligence (XAI) 2.0: A Manifesto of Open Challenges and Interdisciplinary Research Directions (2023, background): “Some researchers use terms like explainability and interpretability syn-onymously [23,26], while others draw significant distinctions between them [64].”
Explainable Deep Learning Models in Medical Image Analysis (2020, background): “Several taxonomies have been proposed in the literature to classify different explainability methods[9, 10].”
Trustworthy Artificial Intelligence: A Review (2022, background): “[6] organized these approaches in a decision tree based on what is explained, the level of the explanations, and how the explanations are generated.”
Explanation-Based Human Debugging of NLP Models: A Survey (2021, background): “2020), also referred to as transparent (Adadi and Berrada, 2018), inherently interpretable (Rudin, 2019), or directly interpretable (Arya et al., 2019).”
Benchmarking and survey of explanation methods for black box models (2021, methodology): “CaptumAI is a 21 AIX360: https://github.com/Trusted-AI/AIX360, CaptumAI: https://captum.ai/, InterpretML: https:// github.com/interpretml/interpret, Alibi https://github.com/SeldonIO/alibi, FAT-Forensics: https://github. com/fat-forensics/fat-forensics, What-If Tool: https://github.com/pair-code/wha”
Sequential Explanations with Mental Model-Based Policies (2020, methodology): “The following two explainers were selected due to their popularity within the ﬁeld of XAI and their demonstrated validity: prototypes (Kim et al., 2016; Gurumoorthy et al., 2017) and deep Taylor decomposition saliency maps (Montavon et al., 2017), implemented using the AIX360 (Arya et al., 2019) and”
A Survey of the State of Explainable AI for Natural Language Processing (2020, methodology): “Finally, we adapt some nomenclature from Arya et al. (2019) which presents a software toolkit that can help users lend explainability to their models and ML pipelines.”