The workshop will invite participation through paper submissions and poster presentations in the area of multimodal representation learning in healthcare. The topic of the workshop is very relevant to the current discourse of investigation of CV, NLP, and ML in healthcare being the immediate impacts on society and the biggest challenges faced by the community. We list prospective topics (but are not limited to) with specific applications in the healthcare domain:
Domain generalization and adaptation for multimodal models; especially new findings / insights that agree or disagree with previous findings in vision-only tasks
Adversarial robustness of multimodal systems
Calibration and uncertainty estimation for multimodal models
Detection and mitigation of fake and misleading images
Watermarking, fingerprinting, or other preventative approaches for misuse of generative models
Reliability from a human perspective, including investigations of bias, fairness, malicious use, users' perception of open-source multimodal systems
Human-AI teaming for multimodal learning
Ideas from human-computer interaction research that seek to improve reliability
New benchmarks, datasets, evaluation metrics for testing open-domain reliability of multimodal systems
Commentary and analysis of failure cases or negative results
Out-of-Distribution detection for multimodal problems and for tasks beyond classification, detection, segmentation
Fairness, explanability, interpretability of multimodal reasoning models
Data-centric AI methods for cleaning, curating, and generating biomedical training data
Our workshop invites 4-page papers that describe innovative ideas and developments. We will also accept ongoing work. The reviewing process will be double-blind. Authors will have an option to (1) opt into ICHI workshop proceedings or (2) a non-archival route that allows papers to be subsequently or concurrently submitted to other venues. Submissions and reviews will not be public. Only accepted papers will be made public and publish in ICHI workshop proceedings.