Smart systems that apply complex reasoning to make decisions and plan behavior are often difficult for users to understand. While research to make systems more explainable and therefore more intelligible and transparent is gaining pace, there are numerous issues and problems regarding these systems that demand further attention. The goal of this workshop is to bring academia and industry together to address these issues. The workshop includes a keynote, poster panels, and group activities, towards developing concrete approaches to handling challenges related to the design, development, and evaluation of explainable smart systems.
INTRODUCTION
Smart systems that apply complex reasoning to make
decisions and plan behaviour, such as clinical decision
support systems, personalized recommendations, home
automation, machine learning classifiers, robots and
autonomous vehicles, are difficult for users to understand
[
MOTIVATION, TOPICS AND CONTRIBUTION
Research to make smart systems explainable is gaining
pace, partly stimulated through a recent DARPA call on
Explainable AI (XAI) [
How to glean explainable information from machine learning systems – “The design and validation of an intuitive confidence measure” (van der Waa et al.), “An Axiomatic Approach to Linear Explanations in Data Classification” (Sliwinski et al.), “Explaining Contrasting Categories” (Pazzani et al.), Explaining Complex Scheduling Decisions” (Ludwig et al.). Explainable/semantically meaningful features – “Explainable Movie Recommendation Systems by using Story-based Similarity” (Lee and Jung), “Labeling images by interpretation from Natural Viewing” (Guo et al.) How to design and present explanations – “Normative vs. Pragmatic: Two Perspectives on the Design of Explanations in Intelligent Systems” (Eiband et al.), “Explaining Recommendations by Means of User Reviews” (Donkers et al.), “What Should Be in an XAI Explanation? What IFT Reveals” (Dodge et al.), “Interpreting Intelligibility under Uncertain Data Imputation” (Lim et al.) Explanations’ impact on user behavior and experience – “Explanation to Avert Surprise” (Gervasio et al.), “Representing Repairs in Configuration Interfaces: A Look at Industrial Practices” (Leclercq et al.), “Explaining smart heating systems to discourage fiddling with optimized behavior” (Stumpf et al.) User feedback/interactive explanations – “Working with Beliefs: AI Transparency in the Enterprise” (Chander et al.), “The Problem of Explanations without user Feedback” (Smith and Nolan) The main part of the workshop is structured around two hands-on activity sessions in small subgroups of 3-5 participants. The activities are grounded in example systems provided by industry participants. The first session identifies challenges and high-level approaches for the example systems while the second session in explores concrete explanation or study designs for the example systems. Each of the subgroups works on the activities in parallel, and the outcomes are shared in a final presentation session.
Workshop Organizers
Dr. Brian Lim is an Assistant Professor in the Department
of Computer Science at the National University of
Singapore (NUS), Singapore, where he researches
ubiquitous computing and intelligible data analytics for
healthcare and smart cities [
Health Research & Technology (BIGHEART) and the
Sensor-enhanced Social Media Centre (SeSaMe) at NUS.
Alison Smith is the Lead Engineer of the Machine
Learning Visualization Lab for Decisive Analytics
Corporation, where her focus is on enhancing end users’
understanding and analysis of complex data without
requiring expertise in data science or machine learning. She
is also a PhD student at the University of Maryland,
College Park, and her research focuses on human-centred
design for interactive machine learning [
Dr. Simone Stumpf is a Senior Lecturer (Associate
Professor) at City, University of London, UK, where she
researches designing end-user interactions with intelligent
systems [