Figure 1. A sampling of Zoom portraits of ICCBR2020 attendees
Bridging Case-Based Reasoning, DL and XAI at the First Virtual ICCBR Conference (ICCBR2020)
Ian Watson, Rosina O Weber, David Leake
Case-based reasoning is reasoning from experience, solving new problems and interpreting new situations by retrieving and adapting prior cases. The Twenty-Eight International Conference on Case-Based Reasoning (ICCBR2020) was held from June 8-12, 2020, with program chairs Ian Watson and Rosina Weber. The conference was originally scheduled for Salamanca, Spain, a World Heritage site, under the auspices of local chair Juan Manuel Corchado and the University of Salamanca. Its theme, “CBR Across Bridges”, reflected the goal of bringing together researchers and practitioners with relevant work across various AI areas. Before the conference, the pandemic struck, with tragic effects. The conference chairs resolved to continue with a safe alternative: the first virtual ICCBR. With researchers unable to travel, the virtual conference not only bridged AI areas but geographic ones: 141 conference attendees participated from 23 countries.
The conference program included 22 paper presentations, two invited talks, and nine doctoral consortium presentations. The papers covered a wide range of areas, including traditional CBR areas, emerging areas, and applications. CBR has been applied and researched based both on its data-driven and reasoning aspects. These characteristics place CBR in a unique position to be explored for data-driven learning tasks. Due to its transparency, it has been investigated for interpretability. For these reasons, CBR and deep learning, and CBR in explainable AI were areas reflected in major submission clusters. Both have been strongly represented in workshops at prior ICCBR conferences and continue to attract great interest. Deep learning methods have been developed, for example, to learn feature representations, similarity criteria, and case adaptation knowledge for CBR. Explainability has been seen as a core CBR benefit from the early days of CBR, because the cases on which CBR systems base their solutions can serve as explanations. Recent research has studied questions such as how to pair CBR systems with black-box systems for explanation and how to select the best explanatory cases. The ICCBR2020 best paper award went to Mark Keane and Barry Smyth for the paper “Good Counterfactuals and Where to Find Them: A Case-Based Technique for Generating Counterfactuals for Explainable AI (XAI)”. This year’s award was renamed in honor of Miltos Petridis, a long-time contributor to CBR and friend to the community, who was lost to Covid-19 and will be deeply missed.
Other strongly represented areas included recommender systems, workflows, and health applications. The conference inaugurated a special track “Challenges and Promises”, aimed at spurring CBR progress by presenting challenges for novel CBR methods or promising areas that create bridges between domains, such as identifying challenges faced by other communities where CBR may have a role, CBR problems in AI tasks or new domains. This track presented a challenge proposal for applying CBR to a new task domain, personal productivity support, and for exploiting CBR within deep learning through use of CBR-inspired workflows in deep learning and in the application of CBR lessons to automated machine learning.
The conference included two keynote talks. CBR methods have a long history in recommender systems. Francesco Ricci’s “Computing useful recommendations: still requires knowledge” argued for the importance of utility and presented an approach to constructing the utility function from user choices. CBR is a powerful method for reasoning from limited data. Nirmalie Wiratunga’s “Learning to Compare with Few Data for Personalised Human Activity Recognition,” presented meta-learning methods and illustrated their application to the leading cause of years lost to disability in Europe – low back pain.
The online experience was overall quite positive. Following the community’s core principle, it was full of learned experiences. We learned that attendees were eager for interaction even in virtual form, and thus virtual social time every day was needed and well attended. Thirty-minute breaks between sessions with breakout rooms enabled extensive online discussions, and short paper presentations replaced posters. Sessions closed with a community meeting in which a major topic was how to advance diversity and inclusion for the conference. After the sessions, instead of traveling to the local train stations, attendees enjoyed two well-attended virtual social events and looked forward to their next meeting in person.
Full details of ICCBR2020, including videos of conference presentations, are available on the conference web site, https://iccbr20.org/ ICCBR-21, with Program Chairs Michael Floyd and Antonio A. Sánchez-Ruiz, is scheduled for beautiful Salamanca in September 2021 and welcomes relevant contributions from all areas of AI and cognitive science. Information is available at https://iccbr21.org.