Interactive artificial intelligence systems employ preferences in both their reasoning and their interaction with the user. This survey considers preference handling in applications such as recommender systems, personal assistant agents, and personalized user interfaces. We survey the major questions and approaches, present illustrative examples, and give an outlook on potential benefits and challenges.

We address user system interaction issues in product search and recommender systems: how to help users select the most preferential item from a large collection of alternatives. As such systems must crucially rely on an accurate and complete model of user preferences, the acquisition of this model becomes the central subject of our paper. Many tools used today do not satisfactorily assist users to establish this model because they do not adequately focus on fundamental decision objectives, help them reveal hidden preferences, revise conflicting preferences, or explicitly reason about tradeoffs. As a result, users fail to find the outcomes that best satisfy their needs and preferences. In this article, we provide some analyses of common areas of design pitfalls and derive a set of design guidelines that assist the user in avoiding these problems in three important areas: user preference elicitation, preference revision, and explanation interfaces. For each area, we describe the state-of-the-art of the developed techniques and discuss concrete scenarios where they have been applied and tested.

In this paper, we consider collaborative filtering as a ranking problem. We present a method which uses Maximum Margin Matrix Factorization and optimizes ranking instead of rating. We employ structured output prediction to optimize directly for ranking scores. Experimental results show that our method gives very good ranking scores and scales well on collaborative filtering tasks.

Content-based image suggestion (CBIS) targets the recommendation of products based on user preferences on the visual content of images. In this paper, we motivate both feature selection and model order identification as two key issues for a successful CBIS. We propose a generative model in which the visual features and users are clustered into separate classes. We identify the number of both user and image classes with the simultaneous selection of relevant visual features using the message length approach. The goal is to ensure an accurate prediction of ratings for multidimensional non-Gaussian and continuous image descriptors. Experiments on a collected data have demonstrated the merits of our approach.

In this paper, we consider collaborative filtering as a ranking problem. We present a method which uses Maximum Margin Matrix Factorization and optimizes ranking instead of rating. We employ structured output prediction to optimize directly for ranking scores. Experimental results show that our method gives very good ranking scores and scales well on collaborative filtering tasks.

Content-based image suggestion (CBIS) targets the recommendation of products based on user preferences on the visual content of images. In this paper, we motivate both feature selection and model order identification as two key issues for a successful CBIS. We propose a generative model in which the visual features and users are clustered into separate classes. We identify the number of both user and image classes with the simultaneous selection of relevant visual features using the message length approach. The goal is to ensure an accurate prediction of ratings for multidimensional non-Gaussian and continuous image descriptors. Experiments on a collected data have demonstrated the merits of our approach.

Experiments on a collected data have demonstrated the merits of our approach.

In this paper, we consider collaborative filtering as a ranking problem. We present a method which uses Maximum Margin Matrix Factorization and optimizes ranking insteadof rating. We employ structured output prediction to optimize directly for ranking scores. Experimental results show that our method gives very good ranking scores and scales well on collaborative filtering tasks.

Social tags are user-generated keywords associated with some resource on the Web. In the case of music, social tags have become an important component of Web2.0" recommender systems, allowing users to generate playlists based on use-dependent terms such as "chill" or "jogging" that have been applied to particular songs. In this paper, we propose a method for predicting these social tags directly from MP3 files. Using a set of boosted classifiers, we map audio features onto social tags collected from the Web. The resulting automatic tags (or "autotags") furnish information about music that is otherwise untagged or poorly tagged, allowing for insertion of previously unheard music into a social recommender. This avoids the ''cold-start problem'' common in such systems. Autotags can also be used to smooth the tag space from which similarities and recommendations are made by providing a set of comparable baseline tags for all tracks in a recommender system."

The advent of the Semantic Web necessitates paradigm shifts away from centralized client/server architectures towards decentralization and peer-to-peer computation, making the existence of central authorities superfluous and even impossible. At the same time, recommender systems are gaining considerable impact in e-commerce, providing people with recommendations that are personalized and tailored to their very needs. These recommender systems have traditionally been deployed with stark centralized scenarios in mind, operating in closed communities detached from their host network's outer perimeter. We aim at marrying these two worlds, i.e., decentralized peer-to-peer computing and recommender systems, in one agent-based framework. Our architecture features an epidemic-style protocol maintaining neighborhoods of like-minded peers in a robust, selforganizing fashion. In order to demonstrate our architecture's ability to retain scalability, robustness and to allow for convergence towards high-quality recommendations, we conduct offline experiments on top of the popular MovieLens dataset.