CounterFair is a group counterfactual search algorithm that detects and minimizes biases among sensitive groups and identifies relevant subgroups inside these sensitive groups based on shared counterfactual instances. We investigate the latter capability, analyzing the found subgroups from the perspective of fairness based on counterfactual reasoning, in order to evaluate whether they present different biases with respect to each other and to the sensitive feature groups they belong to. We perform these measurements on the subgroups extracted by CounterFair over six binary classification datasets, providing figures and their respective analysis on the presence of bias.

Counterfactual explanations can be used as a means to explain a models decision process and to provide recommendations to users on how to improve their current status. The difficulty to apply these counterfactual recommendations from the users perspective, also known as burden, may be used to assess the models algorithmic fairness and to provide fair recommendations among different sensitive feature groups. We propose a novel model-agnostic, mathematical programming-based, group counterfactual algorithm that can: (1) detect biases via group counterfactual burden, (2) produce fair recommendations among sensitive groups and (3) identify relevant subgroups of instances through shared counterfactuals. We analyze these capabilities from the perspective of recourse fairness, and empirically compare our proposed method with the state-of-the-art algorithms for group counterfactual generation in order to assess the bias identification and the capabilities in group counterfactual effectiveness and burden minimization.

Reinforcement learning algorithms have a dependency on observing rewards for actions taken. The relaxed setting of having fully observable rewards, however, can be infeasible in certain scenarios, due to either cost or the nature of the problem. Of specific interest here is the challenge of learning a policy when rewards are delayed, aggregated, and anonymous (DAAF). A problem which has been addressed in bandits literature and, to the best of our knowledge, to a lesser extent in the more general reinforcement learning (RL) setting. We introduce a novel formulation that mirrors scenarios encountered in real-world applications, characterized by intermittent and aggregated reward observations. To address these constraints, we develop four new algorithms: one employs least squares for true reward estimation; two and three adapt Q-learning and SARSA, to deal with our unique setting; and the fourth leverages a policy with options framework. Through a thorough and methodical experimental analysis, we compare these methodologies, demonstrating that three of them can approximate policies nearly as effectively as those derived from complete information scenarios, albeit with minimal performance degradation due to informational constraints. Our findings pave the way for more robust RL applications in environments with limited reward feedback.

On Spotify, multiple recommender systems enable personalized user experiences across a wide range of product features. These systems are owned by different teams and serve different goals, but all of these systems need to explore and learn about new content as it appears on the platform. In this work, we describe ongoing efforts at Spotify to develop an efficient solution to this problem, by centralizing content exploration and providing signals to existing, decentralized recommendation systems (a.k.a. exploitation systems). We take a creator-centric perspective, and argue that this approach can dramatically reduce the time it takes for new content to reach its full potential.

Despite the increased attention to explainable AI, explainability methods for understanding reinforcement learning (RL) agents have not been extensively studied. Failing to understand the agent’s behavior may cause reduced productivity in human-agent collaborations, or mistrust in automated RL systems. RL agents are trained to optimize a long term cumulative reward, and in this work we formulate a novel problem on how to generate explanations on when an agent could have taken another action to optimize an alternative reward. More concretely, we aim at answering the question: What does an RL agent need to do differently to achieve an alternative target outcome? We introduce the concept of a counterfactual policy, as a policy trained to explain in which states a black box agent could have taken an alternative action to achieve another desired outcome. The usefulness of counterfactual policies is demonstrated in two experiments with different use-cases, and the results suggest that our solution can provide interpretable explanations.

In reinforcement learning, an agent makes decisions to maximize rewards in an environment. Rewards are an integral part of the reinforcement learning as they guide the agent towards its learning objective. However, having consistent rewards can be infeasible in certain scenarios, due to either cost, the nature of the problem or other constraints. In this paper, we investigate the problem of delayed, aggregated, and anonymous rewards. We propose and analyze two strategies for conducting policy evaluation under cumulative periodic rewards, and study them by making use of simulation environments. Our findings indicate that both strategies can achieve similar sample efficiency as when we have consistent rewards.