-
Call for Participation
Bring one concrete case of explaining: one you designed, studied, or found wanting.
-
Programme
Two sessions: cases of explaining, then designing an explaining process together.
Outline
This page describes the CHI 2027 Workshop From Explain-ability to Explain-activity, the third edition of the series started as Adaptive XAI.
- Venue: Pittsburgh, PA, USA (in person)
- Workshop Date: May 10 or 11, 2027 TBA
- Submission Deadline: February 11, 2027
- Notification: TBA
- Submission Website: opening soon
- Submission Template: CEUR-ART 1-column
AI models explain why they flagged a case, interfaces explain their own state, teachers explain concepts, clinicians explain diagnoses, and people now explain tasks and corrections to AI agents. Each rests on the same choices: what to say, how much, in what form, at what moment, and what to do when the first attempt fails.
Research on explainable AI has largely treated explainability as a property of a system, judged by examining the system alone, with nobody there to receive it and nothing following the first attempt. This workshop moves to explain-activity: explaining as an activity done with a recipient, shaped by what comes back, and accumulating across encounters.
The same questions are studied in XAI, education, clinical communication and end-user development, with little contact between them. We bring those communities together to work out what varies when someone explains, whether the explainer is a person or a system.
Topics and Themes
We welcome cases of explaining from any setting, and are particularly interested in the following threads:
- Explaining over time
-
What a single encounter loses is the history around it
- What was established last time and can now be assumed
- Approaching a misunderstanding from another angle
- What to withdraw as the recipient becomes competent
- Longitudinal studies of explanation
- Explaining in both directions
-
People now explain tasks, intentions and corrections to AI agents
- Prompting, instructing and correcting as explaining
- Constraining features, supplying local rules, rejecting reasoning
- Instructable agents that ask about what was left out
- Grounding between people and systems
- Repairing explanations
-
What the explainer does when the first attempt fails
- Recognising that an explanation did not land
- Recovery strategies in classrooms, clinics and interfaces
- When to stop, and when to volunteer something unasked
- Measuring explaining
-
Beyond whether the decision improved
- What the recipient ends up knowing
- Self-reported versus demonstrated understanding
- Explanations that build skill rather than reliance
Submissions may come from:
- Explainable AI and human-AI decision-making
- Education and instructional explanation
- Clinical communication
- End-user development and instructable agents
- Technical writing, accessibility and interface design
Workshop Format
The workshop runs over two 90-minute sessions. In the first, authors present their cases of explaining as short talks and posters. In the second, organisers each open a setting with a case from their own work in which explaining failed: an interface, a classroom, a clinic, an AI agent, a person instructing an agent. Attendees join one setting, design an explaining process for it on a shared grid covering a first encounter and a second one weeks later, and compare the designs across settings. See the programme for details.
Contributing Your Work
We accept papers of 5 to 10 pages and posters of 1 to 2 pages, excluding references, each built around one concrete case of explaining: an explanation you designed, studied, or found wanting. Submissions follow the CEUR-WS instructions for single column papers and do not need to be anonymized. Accepted submissions will be published in CEUR-WS proceedings, and results will feed a jointly authored design framework. See the full call for participation.
The deadline for submissions is February 11, 2027 AoE. At least one author of each accepted submission must attend in person.
Please send any comments or questions to Tommaso Turchi, tommaso.turchi@unipi.it.
Organisers
-
Tommaso Turchiis an Assistant Professor at the University of Pisa (Italy) working on human-AI decision-making, over-reliance, and end-user development. He led both previous editions of this series.
-
Simone Borsciis an Associate Professor at the University of Twente (Netherlands), working on human factors and on standardised measurement instruments for usability and user experience, with applied work in healthcare and medical devices.
-
Alan Chamberlainis a Principal Research Fellow in the Mixed Reality Laboratory at the University of Nottingham (United Kingdom), working on ethnographic studies of technology in everyday and creative settings and on explainable AI for the arts.
-
Cristina Conatiis a Professor of Computer Science at the University of British Columbia (Canada), working on user modelling, adaptive interaction and user-adaptive explainable AI.
-
Alan Dixis a Professor Emeritus at Cardiff Metropolitan University and Swansea University (United Kingdom), who has worked on the theory of human-computer interaction for four decades, most recently on human input to interactive AI systems and on the long-run assessment of AI-assisted decisions.
-
Toby Jia-Jun Liis an Assistant Professor at the University of Notre Dame (USA), working on end-user programming and instructable agents, including systems that learn a concept by asking the user about what they left out.
-
Alessio Maliziais a Full Professor at the University of Pisa (Italy) and an associate member of the HIIS Laboratory at CNR-ISTI. His work spans visual languages, intelligent environments and design-led approaches to interactive systems.
-
Fabio Paternòis Research Director at CNR-ISTI in Pisa (Italy), with long-standing contributions to end-user development, human control in IoT settings, and accessibility.
-
Kaisa Väänänenis a Professor of Human-Technology Interaction at Tampere University (Finland), where she leads the Human-Centered Technology research group. Her work spans three decades of interdisciplinary research in industry and academia, most recently on human-centred AI and embodied interaction for sustainability.
Previous Editions
This is the third edition of a series held at ACM IUI under the Adaptive XAI name: IUI 2024 in Greenville and IUI 2025 in Cagliari. Both editions drew participants from healthcare, cybersecurity, smart environments and recommender systems, and both produced CEUR-WS proceedings. This edition widens the scope beyond AI-generated explanation, because the difficulties participants reported, particularly around granularity and timing, were not specific to machine learning, and it now covers explanation running from people to systems as well. Members of the team also organise the SYNERGY workshop series on designing hybrid human-AI systems.