We invite you to participate in the PRICAI 2025 workshop "HIDDEN-RAD: Unlocking Causal Explanations in Medical AI and Beyond." As general AI reasoning capabilities continue to advance, developing systems capable of articulating hidden causal reasoning and generating accurate explanations in specialized domains remains a critical challenge for building trustworthy AI. Building on the successful case study of the NTCIR-18 HIDDEN-RAD challenge, which specifically focused on radiology, our workshop invites research contributions addressing causal reasoning, explanation generation, and hallucination detection across various domains.
Our aim is to bring together researchers working on explanation frameworks, domain-specific knowledge integration, evaluation methodologies, and cross-domain applications. By sharing insights and methodologies across fields, we seek to advance the state-of-the-art in creating AI systems that can not only reach accurate conclusions but explicitly explain their reasoning in ways that domain experts find meaningful and trustworthy.
Bring together researchers working on causal reasoning and explanation generation in AI across domains.
Share methodologies for enhancing AI explanations through reasoning path generation (e.g., Chain-of-Thought) and knowledge integration.
Compare evaluation approaches for measuring explanation quality and detecting hallucinations.
Facilitate cross-domain knowledge transfer of successful techniques.
Create a community interested in advancing explainable AI systems for specialized applications.
Topics of Interest: We welcome submissions that explore a wide range of topics, including but not limited to:
Causal reasoning frameworks in specialized domains (medical, legal, scientific, financial).
Approaches for generating explicit explanations that articulate reasoning steps.
Methodologies for detecting and correcting hallucinations in domain-specific explanations.
Evaluation frameworks for measuring explanation quality (e.g., Human Evaluation, LLM-judge, GPT-based evaluation, RAG-based verification systems).
Knowledge integration techniques to enhance factual accuracy.
Cross-domain applications and knowledge transfer.
Insights from explanation systems like those developed for HIDDEN-RAD and other domains.
We invite two types of submissions:
New Research Papers: Original contributions describing novel research. These submissions should be up to 8 pages excluding references, formatted in the PRICAI 2025 proceedings format (Springer LNCS format). Submissions will undergo a double-blind peer review process.
Recently Published or Work-in-Progress Papers: Papers that have been recently published in a relevant venue (e.g., another conference or journal) or represent ongoing work. These submissions can be in their original format (e.g., PDF of the published paper) or a short paper format (2-4 pages) for work-in-progress. These submissions will be reviewed primarily for their relevance to the workshop's themes and their potential to stimulate discussion.
All submissions must be in English and in PDF format. Accepted papers will be non-archival and will not be formally published in the PRICAI proceedings. Authors are highly encouraged to upload their papers to platforms like arXiv.org, and links to these papers will be provided on the HIDDEN-RAD workshop website. This policy allows authors to present their work at HIDDEN-RAD without precluding future submission to archival venues.
This workshop will feature a separate track dedicated to the Hidden-Rad Shared Task. Datasets will be made available, including those from the NTCIR-18 HIDDEN-RAD challenge, for researchers interested in benchmarking their approaches. This track welcomes contributions on system development and evaluation focusing on discovering and explaining hidden causal relationships. Shared Task participants will have the opportunity to submit their solutions and results and present them at the workshop.