Psychological research shows that enjoyment of many goods is subject to satiation, with enjoyment declining after repeated exposures to the same item. Nevertheless, proposed algorithms for powering recommender systems seldom model these dynamics, instead proceeding as though user preferences were fixed in time. In this work, we adopt a multi-armed bandit setup, modeling satiation dynamics as a time-invariant linear dynamical system. In our model, the expected rewards for each arm decline monotonically with consecutive exposures and rebound towards the initial reward whenever that arm is not pulled. We analyze this model, showing that, when the arms exhibit deterministic identical dynamics, our problem is equivalent to a specific instance of Max K-Cut. In this case, a greedy policy, which plays the arms in a cyclic order, is optimal. In the general setting, where each arm's satiation dynamics are stochastic and governed by different (unknown) parameters, we propose an algorithm that first uses offline data to estimate each arm's reward model and then plans using a generalization of the greedy policy.

Recent technological advancements paved the way for Artificial Intelligence (AI) to be integrated into our daily lives. Innovations in the field have greatly disrupted the business landscape, changing consumer behavior, and redefining customer service. According to the 2019 report of the U.S. think-tank Centre for Data Innovation, AI is applied in 32% of Chinese businesses. Meanwhile, in the EU and U.S., AI application is at 18% and 22%, respectively. As businesses become more competitive, there is a proportionate increase in the demand for AI to help simplify complex tasks.

There's no doubt that retail banking leaders understand the potential of artificial intelligence technology to improve customer experience. Nearly every one (94%) of more than 300 banking and insurance executives surveyed by The Capgemini Research Institute agreed that improving CX is the key objective behind launching new AI-enabled initiatives. In fact, more than half of the international sample say that at least 40% of customer interactions are already enabled by various AI applications, including conversational agents, prescriptive modeling, process automation, and complex analytics. That would be impressive -- except for one thing: Half of more than 5,000 consumers polled by Capgemini worldwide feel that the value they receive from AI-powered financial interactions was "non-existent or less than expected." What about in the U.S., the land of "Erica" and "Eno" and other digital assistants, and the many advanced mobile banking apps?

It's hard not to like a Nest Thermostat, but as the folks at Google learned over the years, the primary reason people cite for not buying one is they're too expensive. It's not as slick or sophisticated as the top-of-the-line Nest Learning Thermostat, but it carries enough of that device's DNA to be an excellent value at $130. Buy one and you'll get the familiar round form factor, a bright display with sharp visuals, and the ease of installation and day-to-day use that made the original product such as star. You'll adjust your HVAC system's target temperatures and the new Nest Thermostat's various settings using its outer ring, too. But instead of spinning a mechanism, you'll slide and tap your fingertip on the responsive touch-sensitive surface of the thermostat's outer bezel.

Rasa, the customizable open source machine learning framework to automate text and voice based AI assistants, has released version 2.0 with significant improvements to dialogue management, training data format, and interactive documentation. In addition, the latest release reduces the learning curve to get started while expanding configuration options for advanced users. Rasa Open Source 2.0 simplifies dialogue policy configuration, draws a clearer distinction between policies that use rules and those that use machine learning, and makes it easier to enforce business logic. Previously, rule-based logic in Rasa Open Source was controlled by a combination of 3 or more dialogue policies. The new RulePolicy allows users to implement forms, map actions to intents, and specify fallback logic, using a single policy.

Deep learning has proved an effective means to capture the non-linear associations of user preferences. However, the main drawback of existing deep learning architectures is that they follow a fixed recommendation strategy, ignoring users' real time-feedback. Recent advances of deep reinforcement strategies showed that recommendation policies can be continuously updated while users interact with the system. In doing so, we can learn the optimal policy that fits to users' preferences over the recommendation sessions. The main drawback of deep reinforcement strategies is that are based on predefined and fixed neural architectures. To shed light on how to handle this issue, in this study we first present deep reinforcement learning strategies for recommendation and discuss the main limitations due to the fixed neural architectures. Then, we detail how recent advances on progressive neural architectures are used for consecutive tasks in other research domains. Finally, we present the key challenges to fill the gap between deep reinforcement learning and adaptive neural architectures. We provide guidelines for searching for the best neural architecture based on each user feedback via reinforcement learning, while considering the prediction performance on real-time recommendations and the model complexity.

In this paper, we aim to mine the cue of user preferences in resource-limited recommendation tasks, for which purpose we specifically build a large used car transaction dataset possessing resource-limitation characteristics. Accordingly, we propose an interest-behavior multiplicative network to predict the user's future interaction based on dynamic connections between users and items. To describe the user-item connection dynamically, mutually-recursive recurrent neural networks (MRRNNs) are introduced to capture interactive long-term dependencies, and meantime effective representations of users and items are obtained. To further take the resource limitation into consideration, a resource-limited branch is built to specifically explore the influence of resource variation on user preferences. Finally, mutual information is introduced to measure the similarity between the user action and fused features to predict future interaction, where the fused features come from both MRRNNs and resource-limited branches. We test the performance on the built used car transaction dataset as well as the Tmall dataset, and the experimental results verify the effectiveness of our framework.

Online recommendation is an essential functionality across a variety of services, including e-commerce and video streaming, where items to buy, watch, or read are suggested to users. Justifying recommendations, i.e., explaining why a user might like the recommended item, has been shown to improve user satisfaction and persuasiveness of the recommendation. In this paper, we develop a method for generating post-hoc justifications that can be applied to the output of any recommendation algorithm. Existing post-hoc methods are often limited in providing diverse justifications, as they either use only one of many available types of input data, or rely on the predefined templates. We address these limitations of earlier approaches by developing J-Recs, a method for producing concise and diverse justifications. J-Recs is a recommendation model-agnostic method that generates diverse justifications based on various types of product and user data (e.g., purchase history and product attributes). The challenge of jointly processing multiple types of data is addressed by designing a principled graph-based approach for justification generation. In addition to theoretical analysis, we present an extensive evaluation on synthetic and real-world data. Our results show that J-Recs satisfies desirable properties of justifications, and efficiently produces effective justifications, matching user preferences up to 20% more accurately than baselines.

Artificial intelligence (AI) is everywhere. In a typical day, people likely use AI multiple times without even knowing it: Alexa and Siri, Google Maps, Uber and Lyft, autopilot on commercial flights, spam filters, and smart email categorization (so anyone using Gmail, Yahoo, or Office 365/outlook), mobile check deposits, plagiarism checkers, online searches, personalized recommendations, Facebook, Instagram, and Pinterest are all examples of AI. But what happens when people are being introduced to a new AI technology? How likely are they to trust the new technology? With an interdisciplinary team of researchers from the University of Kansas, we set to find out.

Machine learning is one of the most promising technologies for the coming decades, a branch of artificial intelligence that studies how to make machines learn, and that could completely change the world as we know it today. If we manage to develop machines capable of learning by themselves, they will probably do so much faster than humans. They will also be able to make more efficient use of the information acquired, getting closer and closer to executive intelligence. Machine learning promises to bring more intelligence to all the software of the machines and devices that surround us, from a smartphone to a coffee machine or a home device, such as Amazon's Echo or Google Home. With that ability to learn, machines will gradually acquire new skills and abilities, achieving previously unthinkable things.