An Assessment of Human vs. Model Uncertainty in Soft-Label Learning and Calibration
For researchers in human-aligned AI and uncertainty calibration, this work clarifies the distinct role of human soft-labels, providing a diagnostic testbed for human-AI uncertainty alignment.
This paper disentangles the benefits of human soft-labels from label mode shifts, showing that while human soft-labels improve accuracy, their primary value is as a regularizer that enhances calibration on difficult samples and training stability. Experiments on MNIST and a synthetic variant demonstrate that models trained on human soft-labels align with human uncertainty, unlike those trained on synthetic labels.
Central to human-aligned AI is understanding the benefits of human-elicited labels over synthetic alternatives. While human soft-labels improve calibration by capturing uncertainty, prior studies conflate these benefits with the implicit correction of mislabeled data (mode shifts), obscuring true effects of soft-labels. We present a controlled audit of soft-label learning across MNIST and a synthetic variant, re-annotating subsets to extract human uncertainty. By decoupling soft-label supervision from underlying label mode shifts, we show that while human soft-labels do provide accuracy gains, their larger value lies in acting as a regularizer that improves model calibration on difficult samples and promotes stable convergence across training runs. Dataset cartography reveals models trained on human soft-labels mirror human uncertainty, whereas those trained on synthetic labels fail to align with humans. Broadly, this work provides a diagnostic testbed for human-AI uncertainty alignment.