Improving user satisfaction is at the forefront of industrial recommender systems. While significant progress has been made by utilizing logged implicit data of user-item interactions (i.e., clicks, dwell/watch time, and other user engagement signals), there has been a recent surge of interest in measuring and modeling user satisfaction, as provided by orthogonal data sources. Such data sources typically originate from responses to user satisfaction surveys, which explicitly ask users to rate their experience with the system and/or specific items they have consumed in the recent past. This data can be valuable for measuring and modeling the degree to which a user has had a satisfactory experience on the recommendation platform, since what users do (engagement) does not always align with what users say they want (satisfaction as measured by surveys). We focus on a large-scale industrial system trained on user survey responses to predict user satisfaction. The predictions of the satisfaction model for each user-item pair, combined with the predictions of the other models (e.g., engagement-focused ones), are fed into the ranking component of a real-world recommender system in deciding items to present to the user. It is therefore imperative that the satisfaction model does an equally good job on imputing user satisfaction across slices of users and items, as it would directly impact which items a user is exposed to. However, the data used for training satisfaction models is biased in that users are more likely to respond to a survey when they will respond that they are more satisfied. When the satisfaction survey responses in slices of data with high response rate follow a different distribution than those with low response rate, response rate becomes a confounding factor for user satisfaction estimation. We find positive correlation between response rate and ratings in a large-scale survey dataset collected in our case study. To address this inherent response rate bias in the satisfaction data, we propose an inverse propensity weighting approach within a multi-task learning framework. We extend a simple feed-forward neural network architecture predicting user satisfaction to a shared-bottom multi-task learning architecture with two tasks: the user satisfaction estimation task, and the response rate estimation task. We Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for third-party components of this work must be honored. For all other uses, contact the owner/author(s).

In this work we study the problem of measuring the fairness of a machine learning model under noisy information. Focusing on group fairness metrics, we investigate the particular but common situation when the evaluation requires controlling for the confounding effect of covariate variables. In a practical setting, we might not be able to jointly observe the covariate and group information, and a standard workaround is to then use proxies for one or more of these variables. Prior works have demonstrated the challenges with using a proxy for sensitive attributes, and strong independence assumptions are needed to provide guarantees on the accuracy of the noisy estimates. In contrast, in this work we study using a proxy for the covariate variable and present a theoretical analysis that aims to characterize weaker conditions under which accurate fairness evaluation is possible. Furthermore, our theory identifies potential sources of errors and decouples them into two interpretable parts 𝛾 and 𝜀. The first part 𝛾 depends solely on the performance of the proxy such as precision and recall, whereas the second part 𝜀 captures correlations between all the variables of interest. We show that in many scenarios the error in the estimates is dominated by 𝛾 via a linear dependence, whereas the dependence on the correlations 𝜀 only constitutes a lower order term. As a result we expand the understanding of scenarios where measuring model fairness via proxies can be an effective approach. Finally, we compare, via simulations, the theoretical upper-bounds to the distribution of simulated estimation errors and show that assuming some structure on the data, even weak, is key to significantly improve both theoretical guarantees and empirical results. CCS CONCEPTS• Computing methodologies → Machine learning.