TRUST-AI is organized as a two-day workshop, August 15-16, at IJCAI-ECAI 2026. The tentative draft program is provided below - changes may occur.
0900 Welcome
0910 Paper session 1: Fairness, bias and explainability
Evaluating Trustworthy Multilingual Alignment in LLMs via Synthetic Natural Language Inference. Samir Abdaljalil, Erchin Serpedin, Khalid Qaraqe and Hasan Kurban
Human Rationale Aggregation for Balancing Performance and Explainability in Text Classification. Hideki Tsukamoto and Jiyi Li
Evaluating and Mitigating Gender Bias in Pre-trained Embeddings for ML-based Recruitment. Farnaz Faramarzi Lighvan and Lynn Houthuys
Fairness-Aware Knowledge Distillation via Symbolic Knowledge Extractors. Federico Sabbatini and Roberta Calegari
1030 Coffee break
1100 Poster session
In the poster session, all poster presenters are invited to make an initial 3-4 minute plenary overview of their presentation. Following this, all poster presentert will be available for presentation and discussion of their work by their posters. This is a great opportunity workshop participants to engage with the presented research.
Trustworthy AI and Mechanical Expressivism: The Systemic Risk of Ontological Hermeneutical Injustice. Marios Thomas Dogkas
Do Not Compare Apples and Offsets: An ESG Alignment Gate for Trustworthy Report Analysis. Zexi Wang and Zewen Shen
From AI output to human decision: a framework for verifiable AI-based tools (position paper). Pavlo Barkhayev and Mikhail Tchernikov
Rule-State Inference: A Bayesian Framework for Trustworthy Compliance Monitoring in Rule-Governed Domains (position paper). Abdou-Raouf Atarmla
Simulating the Clinic: Causal Digital Twins for Physicians' Training in Telemedicine (position paper). Matteo Gregorini, André Meyer-Vitali, Lorenzo Valerio, Chiara Boldrini and Lorenzo Lorusso
Spatial-Temporal Graph Kronecker Convolutional Network with Cauchy Activation Function: Trustworthy Skeleton‑Based Tennis Stroke Recognition Architecture. Paweł Powroźnik, Maria Skublewska-Paszkowska, Marcin Barszcz, Krzysztof Dziedzic and Sylwester Korga
Trustworthy Adversarial Robustness via Adaptive Learning-Driven Multi-Teacher Knowledge Distillation. Hayat Ullah, Talha Zaidi and Arslan Munir
Human Preference aligned Tabular Similarity (position paper). Frederik Hoppe, Astrid Franz, Marianne Michaelis, Lars Kleinemeier and Udo Göbel
Exploring the boundaries of AI use in academic writing: An autoethnography case study of a novice academic writer using a learning journal for transparency (position paper). Glykeria Karagiorgou
Aligning AI Reliance and Perceived Trustworthiness for Responsible Human-AI Interaction (position paper). Sebastian Krügel, Anja Bodenschatz, Eda Ismail-Tsaous, Ute Schmid, Celine Spannagl and Matthias Uhl
When Ground Truths Disagree: Human-Variability-Aware Uncertainty and Explainability for Trustworthy Oncology Segmentation (position paper). Monika Pytlarz, Fabian Umeh, Nazanin Maleki, Nikolay Yordanov, Raisa Amiruddin, Marina Ivory, Mohamed Ghonim, Mohanad Ghonim, Albara Alotaibi, Melisa Guelen, Nathan Page, Pascal Fehringer, Sedra Mhana, Bojan Petrovic, Fatima Memon, Basimah Albalooshy, Elizabeth Schrickel, Mark Krycia, Justin Cramer, Ahmed Moawad, Spyridon Bakas, Jan K. Argasiński and Mariam Aboian
1245 Lunch
1400 Groupwork - General purpose AI in knowledge work.
AI is currently taken up for a broad range of knowledge work tasks – often by users with limited technical training, something that entail substantial trustworthiness issues. Groups identify and prioritize key challenges and research needs to enable trustworthy application of general purpose AI in knowledge work. Group size: 6-7 persons. The groupwork output is presented on Day 2.
1530 Coffee break
1600 Paper session 2: Confidence and hallucination detection
A Voter-Based Stochastic Rejection-Method Framework for Asymptotically Safe Language Model Outputs. Jake Watts and Joel Sokol
The First Token Knows: Single-Decode Confidence for Hallucination Detection. Mina Gabriel and Pei Wang
A Trustworthy Federated Framework for Automated Medical Reporting via Two-Stage OCR and NLI-Based Risk Management. Yu-Chih Wei, Wei Shan Lin and Chia-Yun Hsu
1700 Closing the day
0900 Welcome
0910 Paper session 3: Risk assessment, governance, and regulation
An Empirical Study of Assessing and Mitigating Trustworthiness of Educational AI. Evangelia Anagnostopoulou, Nikolaos Antonios Grammatikos, Dimitris Apostolou and Gregoris Mentzas
Assessment of Trustworthiness of an Industrial AI System Using ALTAI – A Feasibility Study. Shukun Tokas, Aida Omerovic and Ragnhild Halvorsrud
Towards a Trustworthy AI Solution for Disinformation Detection. Eleni Tsalapati
The EU AI Act Definition of "AI System": Interpretive Challenges in Real-World Application. María J. Villalobos Quesada
1030 Coffee break
1100 Groupwork presentations
Groups present the outcomes of the Day 1 groupworks. Each group with 5 minutes presentation time and 5 minutes for Q&A.
1145 Paper session 4: Robustness, safety and security
Assessing Calibration and Conformal Coverage of SAR Perception Models under Sim-to-Real Shift. Diego Argüello Ron, Christyan Cruz Ulloa, Naresh Ravichandran, Anders Lansner, Pawel Herman, Jaime del Cerro and Oscar García Perales
Fuzzy Logic Interacts with Attention: Trustworthy Tennis Movements Identification for Motion Capture Data. Pawel Powroznik, Maria Skublewska-Paszkowska, Krzysztof Dziedzic and Marcin Barszcz
DAPHne: A Trust Layer for Adversarial Detection in Histopathology via Gradient–Confidence Monitoring. Diogen Babuc
1245 Lunch
1400 Paper session 5: Trustworthy agents and LLMs
S³LoRA: Spectral Sharpness–Guided Pruning for Trustworthy Adaptation of Agentic LLMs. Shuang Ao, Shoaib Ehsan and Sarvapali D. Ramchurn
Trustworthy-by-Design Generative AI Assistants for Industrial Troubleshooting: A Knowledge Graph Grounded Architecture. Rohan Jadhav, Emmanuel Papadakis and George Baryannis
Legitimate Delegation and the Preservation of Human Agency in Trustworthy Agentic AI. Alexandru Mateescu
Do LLMs Explain or Remember? Evaluating Language Model Explanations of Reinforcement Learning Failures. Nivetha Suntharamoorthy, Jørn Eirik Betten, Dennis Gross, Eirik Valseth and Helge Spieker
1530 Coffee break
1600 Closing session and next TRUST-AI workshop (ECAI, Athens 2027?)
Reflections on the workshop and discussion on how to align next year's TRUST-AI 2027 with the ECAI 2027 conference theme of human-centred AI.
1630 Closing the day