Proceedings of the 2nd International Workshop on Explainable and Interpretable Machine Learning (XI-ML) - PREFACE

Abstract

With the current scientific discourse on explainable AI (XAI), algorithmic transparency, interpretability, accountability and finally, explainability of algorithmic models and decisions, the XI-ML workshop on explainable and interpretable machine learning tackles these themes from the modeling and learning perspective. In particular, it targets interpretable methods and models being able to explain themselves and their output, respectively. The workshop aims to provide an interdisciplinary forum to investigate fundamental issues in explainable and interpretable machine learning as well as to discuss recent advances, trends, and challenges.

1. Introduction

https://martin.atzmueller.net/ (M. Atzmueller); http://kliegr.eu/ (T. Kliegr) © 2021 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).

For the workshop, there were six accepted papers out of eight submissions in total. Below, we structure these according to the general topics of explanation & methods as well as explanation & applications with their specific methodological foci in the context of explainable and interpretable machine learning, as well as individual applications, respectively.

• Explanation & Methods: – Evolutionary Counterfactual Visual Explanation by Jacqueline Höllig, Stefen Thoma and Cedric Kulbach. The paper proposes a genetic algorithm for the important problem of explaining image classification models using counterfactuals. – Self-explaining variational Gaussian Processes for transparency and modelling of prior knowledge by Sarem Seitz. This paper presents an approach for combining Gaussian Processes with recent algorithms proposed in XAI to study feature importance. – Study on Criteria for Explainable AI for Laypeople by Thorsten Zylowski. This paper presented a survey aimed at identifying criteria important for the development of XAI solutions for the public. The results of this survey show that explanations for AI decisions are widely sought for. The authors suggest that global explanations are instrumental in building initial user trust, while local explanations probably contribute to maintaining trust. • Explanation & Application: – Imitation Learning of Logical Program Policies for Multi-Agent Reinforcement Learning by Manuel Eberhardinger, Johannes Maucher and Setareh Maghsudi. This paper deals with making decisions in Multi-agent systems interpretable by using using the Logical Program Policies, a recently proposed approach originally developed for few-shot Bayesian imitation learning [ 11 ]. – Mining Interesting Outlier Subgraphs In Attributed Graphs by Ahmad Mel and Tijl De Bie. The paper presents an approach for identifying interesting outlier subgraphs in attributed graphs. The algorithm searches for k interesting anomalous subgraphs by scanning the space of candidates using a branch-and-bound beam search procedure. The topic is well-motivated, the approach described algorithmically and also exemplified. – Multi-Perspective Anomaly Detection on Bipartite Multi-Layer Social Interaction Networks by Asep Maulana and Martin Atzmueller. This paper focuses on anomaly detection in complex networks, specifically bipartite multi-layer networks for modeling multiple relations. It presents an interactive approach for characterizing anomalies from diferent explanatory perspectives.

Acknowledgments

We would like to thank the KI’22 organization team for their support throughout the event. Editors

References

[1] M. Atzmueller , F. Puppe , H.-P. Buscher , Exploiting Background Knowledge for KnowledgeIntensive Subgroup Discovery , in: Proc. IJCAI , Edinburgh, Scotland, 2005 , pp. 647 - 652 .

[2] M. Atzmueller , J. Baumeister , F. Puppe , Quality measures and semi-automatic mining of diagnostic rule bases , in: Proc. INAP , number 3392 in LNAI, Springer, 2005 , pp. 65 - 78 .

[3] J. Fürnkranz , T. Kliegr , A brief overview of rule learning , in: Proc. RuleML: Rule Technologies: Foundations , Tools, and Applications., Springer, 2015 , pp. 54 - 69 .

[4] L. Ai , S. H. Muggleton , C. Hocquette , M. Gromowski , U. Schmid, Beneficial and harmful explanatory machine learning , Machine Learning 110 ( 2021 ) 695 - 721 .

[5] T. Kliegr , Š. Bahník, J. Fürnkranz , A review of possible efects of cognitive biases on interpretation of rule-based machine learning models , Artificial Intelligence 295 ( 2021 ).

[6] J. Rabold , M. Siebers , U. Schmid, Generating contrastive explanations for inductive logic programming based on a near miss approach , Machine Learning 111 ( 2022 ) 1799 - 1820 .

[7] J. Žárskỳ , G. Lopez, T. Kliegr, Explainability of text clustering visualizations-twitter disinformation case study , IEEE Computer Graphics and Applications 42 ( 2022 ) 8 - 19 .

[8] M. Iferroudjene , C. Lonjarret , C. Robardet , M. Plantevit , M. Atzmueller , Methods for explaining top-n recommendations through subgroup discovery, Data Mining and Knowledge Discovery ( 2022 ) 1 - 40 .

[9] D. Müller , M. März , S. Scheele , U. Schmid , An interactive explanatory ai system for industrial quality control , in: Proc. AAAI , volume 36 , 2022 , pp. 12580 - 12586 .

[10] L. Beranová , M. P. Joachimiak , T. Kliegr , G. Rabby , V. Sklenák , Why was this cited? explainable machine learning applied to covid-19 research literature , Scientometrics 127 ( 2022 ) 2313 - 2349 .

[11] T. Silver , K. R. Allen , A. K. Lew , L. P. Kaelbling , J. Tenenbaum , Few-shot bayesian imitation learning with logical program policies , in: Proc. AAAI , volume 34 , 2020 , pp. 10251 - 10258 .