When to Accept Automated Predictions and When to Defer to Human Judgment?
This work addresses the reliability and safety of automated decision-making for applications where distribution shifts occur, but it is incremental as it builds on existing clustering and confidence methods.
The paper tackled the problem of unreliable automated predictions under data distribution shifts by proposing a new metric based on clustering distances from class centroids. The results showed consistency across MNIST and CIFAR-10 datasets with different neural network models, indicating the metric can efficiently determine when to accept or defer predictions.
Ensuring the reliability and safety of automated decision-making is crucial. It is well-known that data distribution shifts in machine learning can produce unreliable outcomes. This paper proposes a new approach for measuring the reliability of predictions under distribution shifts. We analyze how the outputs of a trained neural network change using clustering to measure distances between outputs and class centroids. We propose this distance as a metric to evaluate the confidence of predictions under distribution shifts. We assign each prediction to a cluster with centroid representing the mean softmax output for all correct predictions of a given class. We then define a safety threshold for a class as the smallest distance from an incorrect prediction to the given class centroid. We evaluate the approach on the MNIST and CIFAR-10 datasets using a Convolutional Neural Network and a Vision Transformer, respectively. The results show that our approach is consistent across these data sets and network models, and indicate that the proposed metric can offer an efficient way of determining when automated predictions are acceptable and when they should be deferred to human operators given a distribution shift.