An Epistemic Position-Based Click Model: From Interactions to Epistemic Distributions of Relevance and Bias

2026-07-21

AI Summary

This work addresses the limitation of existing click models in quantifying epistemic uncertainty and effectively disentangling the credibility of relevance signals from position bias in user clicks. The paper proposes the first evidential deep learning–based position-aware click model, which explicitly represents both relevance and position bias as Beta distributions. By introducing Bayesian epistemic uncertainty into contextual click modeling for the first time, the model outputs full probability distributions rather than point estimates. Through a combination of position-bias decoupling, variance reduction, and numerically stable optimization strategies, the approach accurately captures epistemic uncertainty on unseen data and learns distributional representations with greater semantic meaning compared to standard policy gradient methods.

Abstract

User interactions with rankings are affected by both items' relevances and display positions. Accordingly, click probabilities are often modeled as a product of relevance and position factors; and for improving recommendation and search, one needs to disentangle relevance from position bias. However, existing click models only provide frequentist point-estimates that do not capture any measure of epistemic uncertainty. Consequently, there is no indication of how much confidence one should have in their predictions. In this work, we introduce the first evidential deep-learning approach to form an epistemic alternative to the important position-based click model. Our learned model takes as input item and position features and outputs a beta-distribution for every relevance and position-bias variable of the position-based model. These distributions capture epistemic uncertainty about click probabilities and the underlying effects of attraction and position bias. The main challenge of our approach is its optimization for which we propose approximation and conditioning techniques to provide numerical stability and variance reduction. Our experiments indicate that our approach captures epistemic uncertainty in predictions on previously-unseen data, whereas standard policy gradients fail to learn meaningful distributions. We believe our contribution of the first contextual epistemic click model constitutes an important step in incorporating Bayesian uncertainty into click modeling.

Problem

• click model
• epistemic uncertainty
• position bias
• relevance estimation
• Bayesian uncertainty

Innovation

• evidential deep learning
• epistemic uncertainty
• click model
• position bias
• beta distribution