Dating used to be about the end result. Its shift to an online business has made it about the journey. That might not be great for the longevity of consumers' relationships, but it should continue to benefit investors' love affair with publicly traded companies like Match Group and Bumble. Match's apps had nearly 100 million collective monthly active users as of the end of the first quarter. Meanwhile, the number of people willing to pay for so-called "freemium" dating apps continues to climb.

Many modern sequential recommender systems use deep neural networks, which can effectively estimate the relevance of items but require a lot of time to train. Slow training increases expenses, hinders product development timescales and prevents the model from being regularly updated to adapt to changing user preferences. Training such sequential models involves appropriately sampling past user interactions to create a realistic training objective. The existing training objectives have limitations. For instance, next item prediction never uses the beginning of the sequence as a learning target, thereby potentially discarding valuable data. On the other hand, the item masking used by BERT4Rec is only weakly related to the goal of the sequential recommendation; therefore, it requires much more time to obtain an effective model. Hence, we propose a novel Recency-based Sampling of Sequences training objective that addresses both limitations. We apply our method to various recent and state-of-the-art model architectures - such as GRU4Rec, Figure 1: The SASRec [18] model trained with our proposed Caser, and SASRec. We show that the models enhanced with our training method outperforms BERT4Rec on the MovieLens-method can achieve performances exceeding or very close to stateof-the-art 20M dataset [14] and requires much less training time.

We present a series of two studies conducted to understand user's affective states during voice-based human-machine interactions. Emphasis is placed on the cases of communication errors or failures. In particular, we are interested in understanding "confusion" in relation with other affective states. The studies consist of two types of tasks: (1) related to communication with a voice-based virtual agent: speaking to the machine and understanding what the machine says, (2) non-communication related, problem-solving tasks where the participants solve puzzles and riddles but are asked to verbally explain the answers to the machine. We collected audio-visual data and self-reports of affective states of the participants. We report results of two studies and analysis of the collected data. The first study was analyzed based on the annotator's observation, and the second study was analyzed based on the self-report.

They allow users to discover new songs or artists they may like within large music catalogs, and they are known to improve the overall user experience on these services [5, 22]. In particular, the French music streaming service Deezer [7], offering 90 million music tracks to 16 million active users from 180 countries, extensively relies on its homemade Flow feature to recommend music. Flow materializes as a simple button, proposed to Deezer users on the homepage of the service. A click on this button launches a personalized and virtually infinite radio-style playlist of songs, computed internally using collaborative filtering methods [3, 16]. However, despite promising results over the past years, Flow used to ignore the moods of users when generating playlists.

As time passes by, technology continues to evolve at an astonishing rate. This has been partly driven by the past few years due to the pandemic, which pushed organisations to adopt new technology and digitally transform at a faster rate, much faster than anyone thought possible within that frame of time. At this height of innovation, the constant galloping acceleration of technology is unrestrained. You might be asking yourself whether all these new developments are actually making life easier or making it more complex, especially as each year there's a constant stream of new features or functions produced by companies making it hard to stay on top of the new technology. It's fine to admit that it can be confusing to comprehend the purpose of the new technology and what it does.

Shopping for a new way to enjoy your favorite music, podcasts, or TV shows? Below you'll find the best headphones and speakers we've seen on sale this Prime Day, which ends in a few hours (11:59 pm PT, to be exact). Be sure to check out our guides to the Best Wireless Headphones, Best Noise-Canceling Headphones, Best Workout Headphones, and Best Bluetooth Speakers for more information about what's hot right now. The WIRED Gear team tests products year-round. We sorted through hundreds of thousands of deals by hand to make these picks.

Modeling users' dynamic preferences from historical behaviors lies at the core of modern recommender systems. Due to the diverse nature of user interests, recent advances propose the multi-interest networks to encode historical behaviors into multiple interest vectors. In real scenarios, the corresponding items of captured interests are usually retrieved together to get exposure and collected into training data, which produces dependencies among interests. Unfortunately, multi-interest networks may incorrectly concentrate on subtle dependencies among captured interests. Misled by these dependencies, the spurious correlations between irrelevant interests and targets are captured, resulting in the instability of prediction results when training and test distributions do not match. In this paper, we introduce the widely used Hilbert-Schmidt Independence Criterion (HSIC) to measure the degree of independence among captured interests and empirically show that the continuous increase of HSIC may harm model performance. Based on this, we propose a novel multi-interest network, named DEep Stable Multi-Interest Learning (DESMIL), which tries to eliminate the influence of subtle dependencies among captured interests via learning weights for training samples and make model concentrate more on underlying true causation. We conduct extensive experiments on public recommendation datasets, a large-scale industrial dataset and the synthetic datasets which simulate the out-of-distribution data. Experimental results demonstrate that our proposed DESMIL outperforms state-of-the-art models by a significant margin. Besides, we also conduct comprehensive model analysis to reveal the reason why DESMIL works to a certain extent.

Counterfactual explanations interpret the recommendation mechanism via exploring how minimal alterations on items or users affect the recommendation decisions. Existing counterfactual explainable approaches face huge search space and their explanations are either action-based (e.g., user click) or aspect-based (i.e., item description). We believe item attribute-based explanations are more intuitive and persuadable for users since they explain by fine-grained item demographic features (e.g., brand). Moreover, counterfactual explanation could enhance recommendations by filtering out negative items. In this work, we propose a novel Counterfactual Explainable Recommendation (CERec) to generate item attribute-based counterfactual explanations meanwhile to boost recommendation performance. Our CERec optimizes an explanation policy upon uniformly searching candidate counterfactuals within a reinforcement learning environment. We reduce the huge search space with an adaptive path sampler by using rich context information of a given knowledge graph. We also deploy the explanation policy to a recommendation model to enhance the recommendation. Extensive explainability and recommendation evaluations demonstrate CERec's ability to provide explanations consistent with user preferences and maintain improved recommendations. We release our code at https://github.com/Chrystalii/CERec.

The content that a recommender system (RS) shows to users influences them. Therefore, when choosing a recommender to deploy, one is implicitly also choosing to induce specific internal states in users. Even more, systems trained via long-horizon optimization will have direct incentives to manipulate users: in this work, we focus on the incentive to shift user preferences so they are easier to satisfy. We argue that - before deployment - system designers should: estimate the shifts a recommender would induce; evaluate whether such shifts would be undesirable; and perhaps even actively optimize to avoid problematic shifts. These steps involve two challenging ingredients: estimation requires anticipating how hypothetical algorithms would influence user preferences if deployed - we do this by using historical user interaction data to train a predictive user model which implicitly contains their preference dynamics; evaluation and optimization additionally require metrics to assess whether such influences are manipulative or otherwise unwanted - we use the notion of "safe shifts", that define a trust region within which behavior is safe: for instance, the natural way in which users would shift without interference from the system could be deemed "safe". In simulated experiments, we show that our learned preference dynamics model is effective in estimating user preferences and how they would respond to new recommenders. Additionally, we show that recommenders that optimize for staying in the trust region can avoid manipulative behaviors while still generating engagement.

Recommender systems play a key role in shaping modern web ecosystems. These systems alternate between (1) making recommendations (2) collecting user responses to these recommendations, and (3) retraining the recommendation algorithm based on this feedback. During this process the recommender system influences the user behavioral data that is subsequently used to update it, thus creating a feedback loop. Recent work has shown that feedback loops may compromise recommendation quality and homogenize user behavior, raising ethical and performance concerns when deploying recommender systems. To address these issues, we propose the Causal Adjustment for Feedback Loops (CAFL), an algorithm that provably breaks feedback loops using causal inference and can be applied to any recommendation algorithm that optimizes a training loss. Our main observation is that a recommender system does not suffer from feedback loops if it reasons about causal quantities, namely the intervention distributions of recommendations on user ratings. Moreover, we can calculate this intervention distribution from observational data by adjusting for the recommender system's predictions of user preferences. Using simulated environments, we demonstrate that CAFL improves recommendation quality when compared to prior correction methods.