Fairness is Not Flat: Geometric Phase Transitions Against Shortcut Learning
For practitioners in fairness and robustness, this provides a computationally efficient a priori method to reduce demographic bias and improve out-of-distribution generalization.
The paper proposes a geometric method to mitigate shortcut learning in deep neural networks, reducing counterfactual gender vulnerability from 21.18% to 7.66% by pruning linear shortcuts and forcing higher geometric capacity.
Deep Neural Networks are highly susceptible to shortcut learning, frequently memorizing low-dimensional spurious correlations instead of underlying causal mechanisms. This phenomenon not only degrades out-of-distribution robustness but also induces severe demographic biases in sensitive applications. In this paper, we propose a geometric \textit{a priori} methodology to mitigate shortcut learning. By deploying a zero-hidden-layer ($N=1$) Topological Auditor, we mathematically isolate features that monopolize the gradient without human intervention. We empirically demonstrate a Capacity Phase Transition: once linear shortcuts are pruned, networks are forced to utilize higher geometric capacity ($N \geq 16$) to curve the decision boundary and learn ethical representations. Our approach outperforms L1 Regularization -- which collapses into demographic bias -- and operates at a fraction of the computational cost of post-hoc methods like Just Train Twice (JTT), successfully reducing counterfactual gender vulnerability from 21.18\% to 7.66\%.