The proceedings are available! (Click to download)

Mission:

Explainable Artificial Intelligence (XAI) concerns the challenge of shedding light on opaque models in contexts for which transparency is important, i.e. where these models could be used to solve analysis or synthesis tasks. In particular, as AI is increasingly being adopted into application solutions, the challenge of supporting interaction with humans is becoming more apparent. Partly this is to support integrated working styles, in which humans and intelligent systems cooperate in problem-solving, but also it is a necessary step in the process of building trust as humans migrate greater competence and responsibility to such systems. The challenge is to find effective ways to characterise, and to communicate, the foundations of AI-driven behaviour, when the algorithms that drive it are far from transparent to humans. While XAI at large is primarily concerned with learning-based approaches, model-based approaches are well suited -- arguably better suited -- for explanation, and Explainable AI Planning (XAIP) can play an important role in addressing complex decision-making procedures.

After the success of previous workshops on XAI and XAIP, the mission of this workshop is to mature and broaden the XAIP community, fostering continued exchange on XAIP topics at ICAPS.

Topics

Topics of interest include (but are not limited to):

• Frameworks for defining meaningful explanations in planning and scheduling contexts;
• Representation, organization, and memory content used in explanation;
• The creation of such content during plan generation or understanding;
• Generation and evaluation of explanations;
• The explanation process, i.e. the way in which explanations are communicated to humans (e.g., plan summaries, answers to questions);
• The role of knowledge and learning in explainable planners;
• Human vs AI models in explanations;
• Links between explainable planning and other disciplines (e.g. social science, argumentation);
• Model differences and model reconciliation;
• Goal reasoning and plan explanations;
• Excuse generation, unsolvability and explanations;
• Use cases and applications of explainable planning.

Submissions:

Given the novelty of XAIP as an area, we invite several different kinds of submissions:

• full technical papers, making an original contribution; up to 9 pages including references;
• short technical papers, making an original contribution; up to 5 pages including references;
• position papers, proposing XAIP challenges, outlining XAIP ideas, debating issues relevant to XAIP; up to 5 pages including references.

Important Dates:

Invited Speaker:

Robert R. Hoffman, Institute for Human and Machine Cognition

Macrocognition: Foundations for Planning and Explanation
Macrocognition is how cognition adapts to complexity. The historical roots of macrocognition reach back to the late 1800s, and the essentials of the paradigm have been fairly well specified.
The models of sensemaking, flexecution, coordination, re-learning, and mental projection help clarify differences between macrocognitive and microcognitive approaches. Microcognitive models are based on causal chains having distinct start and stop (or input-output) points. On the other hand, macrocognitive models are cyclical and closed-loop. Microcognitive models are useful in hindsight, to tell stories; macrocognitive models are transcendent and anticipatory.
The primary macrocognitive functions correspond with the "Families of Laws of Complex Cognitive Systems" developed by David Woods. The Families are based on five fundamental bounds on complex human-machine work systems.
Noteworthy aspects of macrocognition are pertinent to planning systems technology:

• Flexecution emphasizes the fact that goals morph even as they are being pursued.
• Re-grounding embraces the fact that planning and plan execution are team activities.
• Projection elaborates on how planning is anticipatory.

Program:

Accepted Papers:

• Human-Understandable Explanations of Infeasibility for Resource-Constrained Scheduling Problems
  Niklas Lauffer, Ufuk Topcu
• Online Explanation Generation for Human-Robot Teaming
  Mehrdad Zakershahrak, Ze Gong, Akkamahadevi Hanni, Yu Zhang
• Towards Model-Based Contrastive Explanations for Explainable Planning
  Benjamin Krarup, Michael Cashmore, Daniele Magazzeni, Tim Miller
• Combining Cognitive and Affective Measures with Epistemic Planning for Explanation Generation
  Ronald P. A. Petrick, Sara Dalzel-Job, Robin L. Hill
• Challenges of Explaining Control
  Adrian Agogino, Ritchie Lee, Dimitra Giannakopoulou
• Design for Interpretability
  Anagha Kulkarni, Sarath Sreedharan, Sarah Keren, Tathagata Chakraborti, David Smith, Subbarao Kambhampati
• Towards Explainable Planning as a Service
  Michael Cashmore, Anna Collins, Benjamin Krarup, Senka Krivic, Daniele Magazzeni, David Smith
• (How) Can AI Bots Lie?
  Tathagata Chakraborti, Subbarao Kambhampati
• Varieties of Explainable Agency
  Pat Langley
• Towards an argumentation-based approach to explainable planning
  Anna Collins, Daniele Magazzeni, Simon Parsons
• Feature-directed Active Learning for Learning User Preferences
  Sriram Gopalakrishnan, Utkarsh Soni, Subbarao Kambhampati
• Model-Free Model Reconciliation
  Sarath Sreedharan, Alberto Olmo, Aditya Prasad Mishra, Subbarao Kambhampati
• A General Framework for Synthesizing and Executing Self-Explaining Plans for Human-AI Interaction
  Sarath Sreedharan, Tathagata Chakraborti, Christian Muise, Subbarao Kambhampati
• Why Can't You Do That HAL? Explaining Unsolvability of Planning Tasks
  Sarath Sreedharan, Siddharth Srivastava, David Smith, Subbarao Kambhampati
• Bayesian Inference of Temporal Specifications to Explain How Plans Differ
  Joseph Kim, Christian Muise, Ankit Shah, Shubham Agarwal, Julie Shah
• A Preliminary Logic-based Approach for Explanation Generation
  Stylianos Loukas Vasileiou, William Yeoh, Tran Cao Son
• When Agents Talk Back: Rebellious Explanations
  Ben Wright, Mark Roberts, David W. Aha, Ben Brumback
• Branching-Bounded Contingent Planning via Belief Space Search
  Kevin McAreavey, Kim Bauters, Weiru Liu, Jun Hong
• Domain-independent Plan Intervention When Users Unwittingly Facilitate Attacks
  Sachini Weerawardhana, Darrell Whitley, Mark Roberts
• Robust Goal Recognition with Operator-Counting Heuristics
  Felipe Meneguzzi, André Grahl Pereira, Ramon F. Pereira
• Explaining the Space of Plans through Plan-Property Dependencies
  Rebecca Eifler, Michael Cashmore, Jörg Hoffmann, Daniele Magazzeni, Marcel Steinmetz

Sister Workshops:

Please consider our sister workshops at AAMAS and IJCAI!
EXTRAAMAS: visit website
XAI: visit website

Organizing Chairs:

• Tathagata Chakraborti (IBM Research AI, USA)
• Dustin Dannenhauer (Navatek, LLC, USA)
• Joerg Hoffmann (Saarland University, Germany)
• Daniele Magazzeni (King's College London, UK)

Program Committee:

• David Aha (Naval Research Laboratory)
• Susanne Biundo (University of Ulm)
• John Bresina (NASA)
• Michael Cashmore (King's College London)
• Tathagata Chakraborti (Arizona State University)
• Dustin Dannenhauer (Navatek, LLC)
• Jeremy Frank (NASA)
• Joerg Hoffmann (Saarland University)
• Senka Krivic (King's College London)
• Pat Langley (University of Auckland)
• Daniele Magazzeni (King’s College London)
• Fabio Mercorio (University of Milan-Bicocca)
• Tim Miller (The University of Melbourne)
• Matthew Molineaux (Wright State Research Institute)
• Mark Roberts (Naval Research Laboratory)
• David Smith
• Siddharth Srivastava (ASU)
• Kartik Talamadupula (IBM)
• Ben Wright (Naval Research Laboratory)