Optimisation and AI are two vast complimentary fields that are nowadays seeing more and more collaboration. Their intersection includes such fundamental areas as constraint satisfaction, planning and scheduling, which have seen applications in many real-world domains, from resource allocation to software testing to robot navigation, as well as the pervasive use of optimisation techniques in statistical AI research with far-reaching applications all around us. Much research has been devoted to develop expressive planning languages, powerful scheduling methods and efficient optimisation solvers. Recently, the need for developing intelligent explainable systems has been raised, echoing the call for realising explainable AI. Explainable AI seeks building machine-led decision making that is transparent, interpretable, explainable and accountable. Given the pervasive industrial use of increasingly complex optimisation algorithms on both site and enterprise scales, as well as the potential benefits of more advanced AI and optimisation techniques, explainability is critical in industrial applications. At the same time, in order to see large-scale deployment, explainable AI methods have to be highly efficient. Synthesis of optimisation and explanation therefore promises lucrative two-way advances of both explainable optimisation and scalable explanations in AI. 

In this context, the workshop on Optimisation and Explanations in AI will aim to address the need for explainable optimisation techniques by employing formal AI methods and explore the possibility of scaling and improving explainable AI using optimisation techniques. The workshop will bring researchers and industry experts from (traditional) optimisation domains such as operations research, planning and scheduling together with experts working on explainable AI, including but not limited to symbolic AI. We will specifically incorporate methodological experts from the domains of optimisation, e.g. mixed-integer programming, combinatorial optimisation and constraint programming, and AI, e.g. knowledge representation, reasoning and knowledge engineering. The workshop aims to support collaboration among researchers from the relevant AI and optimisation areas as well as industry partners seeking explainable optimisation solutions and/or scalable explanation methods. The attendees will share expertise, identify key research questions and deployment challenges, and propose techniques for developing explainable optimisation systems and optimising explanation mechanisms. The workshop will establish entirely new interdisciplinary collaborations and advance state-of-the-art computational models of explanations and optimisation. 

We seek works on explaining computational optimisation with formal AI methods and scaling explainable AI approaches with optimisation methods. In particular, we solicit works on:

• Techniques
  • Computational optimisation and constraint programming
  • Planning
  • Scheduling
  • Knowledge engineering
  • Knowledge representation and reasoning
  • Symbolic methods
  • Logic-based approaches
  • Argumentation
  • Abduction
  • Verification
• Applications of explainable optimisation and/or scalable explanations for
  • Manufacturing scheduling
  • Autonomous agents
  • Resource allocation
  • Software testing
  • Robot planning
  • Machine learning

We welcome original contributions in the form of mature papers or work in progress.

Submission information:

We welcome papers in between 4 to 9 pages in IJCAI-19 format. At least one author of any accepted paper will need to register by the early registration deadline and attend the workshop. More information on https://www.doc.ic.ac.uk/~kc2813/OXAI/

Important Dates:

Paper Submission: May 9, 2019

Authors Notification: May 30, 2019

Workshop Organisers:

Kristijonas Čyras, Xiuyi Fan, Dimitris Letsios

Program Committee:

Please check the program committee members at: https://www.doc.ic.ac.uk/~kc2813/OXAI/

Contact:

Questions about the workshop should be directed to the organisers through the email: oxai19@outlook.com