Recommender systems are a ubiquitous feature of online platforms. Increasingly, they are explicitly tasked with increasing users' long-term satisfaction. In this context, we study a content exploration task, which we formalize as a multi-armed bandit problem with delayed rewards. We observe that there is an apparent trade-off in choosing the learning signal: Waiting for the full reward to become available might take several weeks, hurting the rate at which learning happens, whereas measuring short-term proxy rewards reflects the actual long-term goal only imperfectly. We address this challenge in two steps. First, we develop a predictive model of delayed rewards that incorporates all information obtained to date. Full observations as well as partial (short or medium-term) outcomes are combined through a Bayesian filter to obtain a probabilistic belief. Second, we devise a bandit algorithm that takes advantage of this new predictive model. The algorithm quickly learns to identify content aligned with long-term success by carefully balancing exploration and exploitation. We apply our approach to a podcast recommendation problem, where we seek to identify shows that users engage with repeatedly over two months. We empirically validate that our approach results in substantially better performance compared to approaches that either optimize for short-term proxies, or wait for the long-term outcome to be fully realized.

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.

Online controlled experiments, in which different variants of a product are compared based on an Overall Evaluation Criterion (OEC), have emerged as a gold standard for decision making in online services. It is vital that the OEC is aligned with the overall goal of stakeholders for effective decision making. However, this is a challenge when the overall goal is not immediately observable. For instance, we might want to understand the effect of deploying a feature on long-term retention, where the outcome (retention) is not observable at the end of an A/B test.

Algorithmic recommendations shape music consumption at scale, and understanding the impact of various algorithmic models on how content is consumed is a central question for music streaming platforms. The ability to shift consumption towards less popular content and towards content different from user's typical historic tastes not only affords the platform ways of handling issues such as filter bubbles and popularity bias, but also contributes to maintaining a healthy and sustainable consumption patterns necessary for overall platform success.

We consider the problem of predicting users’ preferences on online platforms. We build on recent findings suggesting that users’ preferences change over time, and that helping users expand their horizons is important in ensuring that they stay engaged. Most existing models of user preferences attempt to capture simultaneous preferences: “Users who like A tend to like B as well”. In this paper, we argue that these models fail to anticipate changing preferences. To overcome this issue, we seek to understand the structure that underlies the evolution of user preferences. To this end, we propose the Preference Transition Model (PTM), a dynamic model for user preferences towards classes of items. The model enables the estimation of transition probabilities between classes of items over time, which can be used to estimate how users’ tastes are expected to evolve based on their past history. We test our model’s predictive performance on a number of different prediction tasks on data from three different domains: music streaming, restaurant recommendations and movie recommendations, and find that it outperforms competing approaches. We then focus on a music application, and inspect the structure learned by our model. We find that the PTM uncovers remarkable regularities in users’ preference trajectories over time. We believe that these findings could inform a new generation of dynamic, diversity-enhancing recommender systems.

Recommender systems play an important role in providing an engaging experience on online music streaming services. However, the musical domain presents distinctive challenges to recommender systems: tracks are short, listened to multiple times, typically consumed in sessions with other tracks, and relevance is highly context-dependent. In this paper, we argue that modeling users’ preferences at the beginning of a session is a practical and effective way to address these challenges. Using a dataset from Spotify, a popular music streaming service, we observe that a) consumption from the recent past and b) session-level contextual variables (such as the time of the day or the type of device used) are indeed predictive of the tracks a user will stream—much more so than static, average preferences. Driven by these findings, we propose CoSeRNN, a neural network architecture that models users’ preferences as a sequence of embeddings, one for each session. CoSeRNN predicts, at the beginning of a session, a preference vector, based on past consumption history and current context. This preference vector can then be used in downstream tasks to generate contextually relevant just-in-time recommendations efficiently, by using approximate nearest-neighbour search algorithms. We evaluate CoSeRNN on session and track ranking tasks, and find that it outperforms the current state of the art by upwards of 10% on different ranking metrics. Dissecting the performance of our approach, we find that sequential and contextual information are both crucial.

On many online platforms, users can engage with millions of pieces of content, which they discover either organically or through algorithmically-generated recommendations. While the short-term benefits of recommender systems are well-known, their long-term impacts are less well understood. In this work, we study the user experience on Spotify, a popular music streaming service, through the lens of diversity—the coherence of the set of songs a user listens to. We use a high-fidelity embedding of millions of songs based on listening behavior on Spotify to quantify how musically diverse every user is, and find that high consumption diversity is strongly associated with important long-term user metrics, such as conversion and retention. However, we also find that algorithmically-driven listening through recommendations is associated with reduced consumption diversity. Furthermore, we observe that when users become more diverse in their listening over time, they do so by shifting away from algorithmic consumption and increasing their organic consumption. Finally, we deploy a randomized experiment and show that algorithmic recommendations are more effective for users with lower diversity. Our work illuminates a central tension in online platforms: how do we recommend content that users are likely to enjoy in the short term while simultaneously ensuring they can remain diverse in their consumption in the long term?

Inspired by applications in sports where the skill of players or teams competing against each other varies over time, we propose a probabilistic model of pairwise-comparison outcomes that can capture a wide range of time dynamics. We achieve this by replacing the static parameters of a class of popular pairwise-comparison models by continuous-time Gaussian processes; the covariance function of these processes enables expressive dynamics. We develop an efficient inference algorithm that computes an approximate Bayesian posterior distribution. Despite the flexbility of our model, our inference algorithm requires only a few linear-time iterations over the data and can take advantage of modern multiprocessor computer architectures. We apply our model to several historical databases of sports outcomes and find that our approach outperforms competing approaches in terms of predictive performance, scales to millions of observations, and generates compelling visualizations that help in understanding and interpreting the data.

As the number of contributors to online peer-production systems grows, it becomes increasingly important to predict whether the edits that users make will eventually be beneficial to the project. Existing solutions either rely on a user reputation system or consist of a highly specialized predictor that is tailored to a specific peer-production system. In this work, we explore a different point in the solution space that goes beyond user reputation but does not involve any content-based feature of the edits. We view each edit as a game between the editor and the component of the project. We posit that the probability that an edit is accepted is a function of the editor's skill, of the difficulty of editing the component and of a user-component interaction term. Our model is broadly applicable, as it only requires observing data about who makes an edit, what the edit affects and whether the edit survives or not. We apply our model on Wikipedia and the Linux kernel, two examples of large-scale peer-production systems, and we seek to understand whether it can effectively predict edit survival: in both cases, we provide a positive answer. Our approach significantly outperforms those based solely on user reputation and bridges the gap with specialized predictors that use content-based features. It is simple to implement, computationally inexpensive, and in addition it enables us to discover interesting structure in the data.