Contrastive explanation: a structural-model approach

The topic of causal explanation in artificial intelligence has gathered interest in recent years as researchers and practitioners aim to increase trust and understanding of intelligent decision-making and action. While different sub-fields have looked into this problem with a sub-field-specific view, there are few models that aim to capture explanation in AI more generally. One general model is based on structural causal models. It defines an explanation as a fact that, if found to be true, would constitute an actual cause of a specific event. However, research in philosophy and social sciences shows that explanations are contrastive: that is, when people ask for an explanation of an event -- the fact --- they (sometimes implicitly) are asking for an explanation relative to some contrast case; that is, "Why P rather than Q?". In this paper, we extend the structural causal model approach to define two complementary notions of contrastive explanation, and demonstrate them on two classical AI problems: classification and planning. We believe that this model can be used to define contrastive explanation of other subfield-specific AI models.

[1]  David-Hillel Ruben,et al.  Explaining contrastive facts , 1987 .

[2]  Petri Ylikoski,et al.  THE IDEA OF CONTRASTIVE EXPLANANDUM , 2007 .

[3]  B. Chandrasekaran,et al.  Explaining control strategies in problem solving , 1989, IEEE Expert.

[4]  WRTG,et al.  ? ' Why ? ' Questions , 2011 .

[5]  Dennis Temple,et al.  The Contrast Theory of Why-Questions , 1988, Philosophy of Science.

[6]  Carlos Guestrin,et al.  "Why Should I Trust You?": Explaining the Predictions of Any Classifier , 2016, ArXiv.

[7]  Seth Chin-Parker,et al.  Contrastive Constraints Guide Explanation-Based Category Learning , 2017, Cogn. Sci..

[8]  Germund Hesslow,et al.  The problem of causal selection , 1988 .

[9]  Michael A. Rupp,et al.  Intelligent Agent Transparency in Human–Agent Teaming for Multi-UxV Management , 2016, Hum. Factors.

[10]  Frank E. Ritter,et al.  Designs for explaining intelligent agents , 2009, Int. J. Hum. Comput. Stud..

[11]  Song-Chun Zhu,et al.  CoCoX: Generating Conceptual and Counterfactual Explanations via Fault-Lines , 2020, AAAI.

[12]  Trevor Darrell,et al.  Generating Counterfactual Explanations with Natural Language , 2018, ICML 2018.

[13]  Alex Kean.  A Characterization of Contrastive Explanations Computation , 1998, PRICAI.

[14]  Michael Winikoff,et al.  Debugging Agent Programs with Why?: Questions , 2017, AAMAS.

[15]  Germund Hesslow,et al.  Explaining differences and weighting causes , 2008 .

[16]  Joseph Y. Halpern,et al.  Causes and Explanations: A Structural-Model Approach. Part I: Causes , 2000, The British Journal for the Philosophy of Science.

[17]  D. Hilton.  Conversational processes and causal explanation. , 1990 .

[18]  Mandy Eberhart,et al.  The Scientific Image , 2016 .

[19]  Bruce G. Buchanan,et al.  The MYCIN Experiments of the Stanford Heuristic Programming Project , 1985 .

[20]  Pamela J. Hinds,et al.  Autonomy and Common Ground in Human-Robot Interaction: A Field Study , 2007, IEEE Intelligent Systems.

[21]  Denis J. Hilton,et al.  Contemporary science and natural explanation : commonsense conceptions of causality , 1988 .

[22]  Tim Miller,et al.  Model-based contrastive explanations for explainable planning , 2019 .

[23]  John D. Lee,et al.  Human-Automation Collaboration in Dynamic Mission Planning: A Challenge Requiring an Ecological Approach , 2006 .

[24]  Qian Yang,et al.  Designing Theory-Driven User-Centric Explainable AI , 2019, CHI.

[25]  Joseph Y. Halpern.  A Modification of the Halpern-Pearl Definition of Causality , 2015, IJCAI.

[26]  Siobhan Chapman.  Logic and Conversation , 2005 .

[27]  Amit Sharma,et al.  Explaining machine learning classifiers through diverse counterfactual explanations , 2020, FAT*.

[28]  James R. Webb,et al.  Forms of Explanation; Rethinking the Questions in Social Theory , 1983 .

[29]  Anind K. Dey,et al.  Assessing demand for intelligibility in context-aware applications , 2009, UbiComp.

[30]  Chris Russell,et al.  Counterfactual Explanations Without Opening the Black Box: Automated Decisions and the GDPR , 2017, ArXiv.

[31]  Mark A. Neerincx,et al.  Contrastive Explanations for Reinforcement Learning in terms of Expected Consequences , 2018, IJCAI 2018.

[32]  Tim Miller,et al.  Explanation in Artificial Intelligence: Insights from the Social Sciences , 2017, Artif. Intell..

[33]  Roger Lamb,et al.  Attribution in conversational context: Effect of mutual knowledge on explanation‐giving , 1993 .

[34]  Erik Weber,et al.  Remote causes, bad explanations? , 2002 .

[35]  Tim Miller,et al.  Explainable Reinforcement Learning Through a Causal Lens , 2019, AAAI.

[36]  Amit Dhurandhar,et al.  Explanations based on the Missing: Towards Contrastive Explanations with Pertinent Negatives , 2018, NeurIPS.

[37]  Joseph Y. Halpern,et al.  Causes and explanations: A structural-model approach , 2000 .

[38]  Johanna D. Moore,et al.  Explanation in second generation expert systems , 1993 .

[39]  Subbarao Kambhampati,et al.  Hierarchical Expertise Level Modeling for User Specific Contrastive Explanations , 2018, IJCAI.

[40]  Joseph Y. Halpern,et al.  Causes and Explanations: A Structural-Model Approach. Part II: Explanations , 2001, The British Journal for the Philosophy of Science.