Conference track · AISoLA 2026

Responsible and Trusted AI

An Interdisciplinary Perspective · AISoLA 2026

When
Tuesday 27 to Friday 30 October 2026
Where
Mitsis Selection Blue Domes, Kos, Greece
All track sessions take place in Room 1
Conference
AISoLA 2026 website · Conference programme · Track page
Organisers
Kevin Baum (DFKI, Hamburg University of Technology)
Thorsten Helfer (CISPA)
Felix Jahn (DFKI)
Sophie Kerstan (University of Freiburg)
Andreas Sesing-Wagenpfeil (Saarland University)
Timo Speith (University of Bayreuth)
Sarah Sterz (Saarland University)
Co-organised by
CERTAIN, the Centre for European Research in Trusted AI
Supported by
Lamarr Institute
Proceedings
LNCS 17146, open access

The track brings together researchers from philosophy, law, psychology, computer science and neighbouring fields to work on the conceptual, normative and regulatory side of trustworthy AI. It has run at AISoLA since 2023.

Programme

This page shows the most recent schedule of the RTAI talks, including last-minute changes. If the interactive conference programme differs for an RTAI talk, please check here. Talks with a paper in the proceedings link to it. The shaded boxes list central AISoLA events in the same room.

Tuesday, 27 October 11:00–15:00 · 6 talks

Human Oversight11:00–12:30

Lunch break12:30–13:30

Shorter than on the other days: because the outing starts at 15:30, lunch ends at 13:30 instead of 14:30.

Meeting Risks of AI (1)13:30–15:00

  • 13:30When Rules Run Out: Formal Methods for Open-Textured Legal Terms
    Frederike Zufall, Maike Schwammberger
  • 14:00Interaction-Level Monitoring: A Holistic Strategy for Mitigating Human–AI Double Anchoring
    Mariela Destefano
  • 14:30When the Red Team Gets an AI Intern – Ethical Continuities and New Challenges in AI-Assisted Vulnerability & Exploitation Research
    Thorsten Helfer, Jannik Zeiser

Wednesday, 28 October 11:00–16:00 · 5 talks

Meeting Risks of AI (2)11:00–12:00

  • 11:00AI Literacy as System Capacity: Socio-Technical Principles for AI-Literate AI Use
    Sophie Kerstan, Richard Uth, Markus Langer
  • 11:30Between Prohibition and Permission: How the EU AI Act Tackles the Dark Side of AI
    Andreas Sesing-Wagenpfeil

AI for Good (1)12:00–12:30

  • 12:00Trustworthy AI and the Democratization of Philosophical Inquiry
    Michael Hemmingsen
Lunch break12:30–14:30

AI for Good (2)14:30–15:00

  • 14:30Stitching social robots: Fostering Youth Critical AI Engagement through Tangible and Digital Co-Creation
    Xiaoyu Wang, Anna Keune

XAI (1)15:00–16:00

  • 15:00Trustworthy, Traceable, and Compliant: Run-Time Validation of AI Explanations for Autonomous Vehicles under the EU AI Act
    Dominik Grundt, Maike Schwammberger
  • 15:30Conceptual Infrastructure for Explainable AI Cancelled
    Timo Speith, Astrid Schomäcker

Thursday, 29 October 11:00–16:00 · 6 talks

XAI (2)11:00–12:30

Lunch break12:30–14:30

AI, Responsibility and Authorship14:30–16:00

  • 14:30Students' Judgments of Seriousness of LLM-Based vs. Conventional Plagiarism: A Qualitative Interview Study
    Hernán González Cruz, Tanja Fritz, Martin Daumiller, Stefan Janke
  • 15:00Should Alignment be about Reasons or Preferences? An interactive discussion
    Felix Jahn, Lisa Dargasz, Andre Steingrueber
  • 15:30Responsible Conversational AI: Shifting Focus from Alignment to Gatekeepers
    Ron Aboodi

Friday, 30 October 11:00–12:30 · 3 talks

AI Alignment (2)11:00–12:30

  • 11:00Reason-based Normative Reinforcement Learning: Embedding Defaults into the Reward Structure via Reward Machines
    Yannic Muskalla, Lisa Dargasz, Felix Jahn
  • 11:30Value Capture and the Structure of AI Alignment: A Case for Rule-Based Approaches
    Andre Steingrüber, Kevin Baum
  • 12:00Synthesizing Moral Reasoning Theories from Live Audience Judgments: An interactive session with the AISoLA-RTAI Community
    Lisa Dargasz, Felix Jahn, Yannic Muskalla
Lunch break12:30–14:30

Proceedings

The track chapters appear in Bridging the Gap Between AI and Reality (AISoLA 2026, ed. Bernhard Steffen, Lecture Notes in Computer Science 17146, Springer), online since 13 October 2026. All chapters are open access. The descriptions are short summaries; the authors’ abstracts are on the Springer pages.

Responsible and Trusted AI: An Interdisciplinary Perspective (2026)

Kevin Baum, Thorsten Helfer, Felix Jahn, Sophie Kerstan, Andreas Sesing-Wagenpfeil, Timo Speith, Sarah Sterz

The track introduction. It asks whether the safeguards that regulation and industry practice have produced (humans in decision loops, oversight duties, documentation, explanations) do what they are meant to do, and places the track’s papers within that question.

pp. 107–114 · doi:10.1007/978-3-032-41269-0_7

Constitutive vs. Corrective: A Causal Taxonomy of Human Runtime Involvement in AI Systems

Kevin Baum, Johann Laux

Distinguishes human-in-the-loop (constitutive: the output cannot come about without the human) from human-on-the-loop (corrective: the human can prevent, modify or override it) by causal structure rather than position. Keeps these descriptive categories apart from the normative requirement of effective human oversight, and treats cases where one person holds both roles as a design problem.

pp. 115–130 · doi:10.1007/978-3-032-41269-0_8

Compliance-by-Design with Lingua Franca: Toward Meeting EU AI Act Requirements

Matthias Wagner, Sriram Bhimaraju, Markus Borg, Per Runeson, Edward A. Lee

Shows how the coordination language Lingua Franca turns selected AI Act requirements into architectural properties of AI-enabled cyber-physical systems: generated architecture diagrams, deterministic and explicitly timed behaviour, anomaly monitoring and traceable logging, demonstrated on an AI-enabled traffic-light controller.

pp. 131–145 · doi:10.1007/978-3-032-41269-0_9

Enhancing AI Interpretability Through Localised Architectures

Ian Seet, Jonas Bozenhard, Simon Ostermann

Argues that localised models are easier to interpret, and that today’s dense, high-bandwidth hardware pushes models the other way. Surveys alternative hardware substrates for more localised architectures by per-node expressivity, energy efficiency and maturity.

pp. 146–160 · doi:10.1007/978-3-032-41269-0_10

Earlier editions: AISoLA 2025 · AISoLA 2024

  • Lamarr Institute for Machine Learning and Artificial Intelligence
  • DFKI
  • CERTAIN
  • University of Freiburg
  • Algoright e.V.
  • CISPA Helmholtz Center for Information Security
  • Saarland University
  • University of Bayreuth
  • Hamburg University of Technology