Deceptive bias measurement in deep learning: Assessing shortcut reliance in TCGA cancer models

Machine learning bias is a persistent challenge because it can create unfair outcomes, limit generalization, and reduce trust in real-world applications. A key source of this problem is shortcut learning, where models exploit signals linked to sensitive attributes, such as the data source or collection site, instead of relying on task-relevant features. To address this, we propose the Deceptive Signal metric, a novel quantitative measure designed to assess the extent of a model's reliance on hidden shortcuts during the learning process. This metric is derived via the Deceptive Bias Detection pipeline, which isolates shortcut dependence by contrasting the model's behavior under two controlled conditions: (1) Full Exclusion, where a sensitive subgroup is completely removed from training; and (2) Partial Exclusion, where the model has limited access to specific classes within that subgroup. By calculating the behavioral shift between these settings, the Deceptive Signal metric provides a quantitative estimate of the model's susceptibility to learning task-irrelevant patterns. In experiments with the TCGA histopathology dataset, our metric successfully quantified substantial dependencies on center-specific artifacts in models trained for cancer classification.