Call for Papers

Topics ℹ️

The past few years have seen AI, particularly Deep Learning, achieve remarkable capabilities, transforming fields from scientific discovery to everyday applications.
Following breakthroughs in model performance across various tasks, it is clear that the next crucial frontier lies in ensuring these systems are fundamentally Unbiased, Interpretable, and Trustworthy (UIT).
There has recently been significant interest in addressing these aspects, yet a deep, methodological investigation into the core science of UIT remains paramount, especially for high-stakes domains.

One challenging but compelling aspect unique to building truly reliable AI is the rigorous definition, measurement, and validation of its Unbiasedness: from uncovering deep-seated biases in complex data and algorithmic processes to developing foundational mitigation strategies.
Similarly, true Interpretability demands advancing the science of explanation itself, understanding why models behave as they do, and formally validating these insights, not just generating plausible narratives.
Furthermore, genuine Trustworthiness requires establishing principled frameworks for robustness, fairness, and accountability that go beyond empirical performance on specific benchmarks.

Given the profound societal implications of AI and the urgent need for robust, foundational methodologies, we believe now is the perfect time to engage the broader AI and computer vision community in this critical research area.
The first BISCUIT workshop seeks to bring together researchers focused on the fundamental science and engineering of UIT: from those developing core machine learning theories, to experts in algorithmic fairness, model interpretability, robust AI, formal methods, and human-computer interaction, especially those whose work develops generalizable UIT methods with potential for or inspiration from challenges in critical domains.

Important dates 📅