Daniel Tunkelang

Recommendations as a conversation with the user

Recommender systems provide users with products or content intended to satisfy their information needs. The primary evaluation measures for recommender systems emphasize either the perceived relevance of the recommendations or the actions driven by those recommendations (e.g., purchases on ecommerce sites or clicks on news and social networking sites). Unfortunately, this transactional emphasis neglects the inherently interactive nature of the user experience. This tutorial explores recommendations as part of a conversation between users and systems. A conversational approach should provide transparency, control, and guidance. Transparency means that users understand why systems offer particular recommendations. Control means that users can explicitly manipulate the behavior of recommender systems based on personal needs and preferences. Guidance means that systems offers plausible and predictable next steps rather than requiring users to guess the consequences of their interactions. Finally, there are psychological factors -- in particular, the faith that users place in the recommender systems effectiveness. Since users develop mental models of recommender systems, the system should become more predictable with repeated use. The tutorial does not require any special background in interfaces or usability. Rather, it summarizes the best lessons from research and industry, offering concrete examples and practical techniques to make recommender systems most effective for users.

The role of network distance in linkedin people search

LinkedIn is the worlds largest professional network, with over 300 million members. One of the primary activities on the site is people search, for which LinkedIn members are both the users and the corpus. This paper presents insights about people search behavior on LinkedIn, based on a log analysis and a user study. In particular, it examines the role that network distance plays in name searches and non-name searches. For name searches, users primarily click on only one of the results, and closer network distance leads to higher click-through rates. In contrast, for non-name searches, users are more likely to click on multiple results that are not in their existing connections, but with whom they have shared connections. The results show that, while network distance contributes significantly to LinkedIn search engagement in general, its role varies dramatically depending on the type of search query.

Editorial: Introduction to special issue on human-computer information retrieval

Contemporary search engines are optimized for look-up scenarios where the information target is well-defined and human–machine interaction is limited to queries and search-result selections. In this role the search system serves as a cognitive prosthetic, temporarily enhancing people’s mental capabilities to provide access to additional information not known to the searcher or not readily accessible to them. However, this type of support is insufficient for tasks requiring more involved information interaction (e.g., where cross-query learning may be important) and situations where information behavior encompasses more than just information seeking. People may often possess more aspirational goals such as augmenting their intellect (Engelbart, 1962) and applying their knowledge to explore, learn, and make sense of the information they encounter (Marchionini, 2006a). To meet these demands search systems need to form a symbiotic relationship with their users, providing support for fluid and meaningful interaction between human and machine, and promoting critical thinking and increased interaction with information to facilitate cognitive development. Human–computer information retrieval (HCIR) is the study of methods that integrate human intelligence and algorithmic search to help people better search, explore, learn from, and leverage information (Marchionini, 2006b). It comprises a number of sub-disciplines including exploratory search (White & Roth, 2009) which combines querying and browsing strategies to foster learning and investigation; information retrieval in context (Ingwersen & Järvelin, 2005) which considers aspects of the user or environment that are typically not reflected in query statements; and interactive information retrieval (IIR), which is primarily focused on the interactive communication processes between the main actors in retrieval operations: users, systems, and optionally, intermediaries (Ingwersen, 1992). All of these sub-disciplines share a common goal: understanding and improving the way that people leverage technology to derive value from information. HCIR affords searchers self-actualization, deep involvement in the search process, and predominant roles in system design, use, and evaluation. Over the past few decades, search behavior has converged on a small number of tactics that transform the user into a passive information receiver rather than an active information seeker (Allan et al., 2012). HCIR systems empower searchers to be more proactive, critical thinkers during the information search process. To achieve this they must help users acquire better information search skills (through tutorials, tips, etc. (Bateman, Teevan, & White, 2012; Moraveji, Russell, & Mease, 2011)) and provide better system support for information interaction and understanding. Marchionini (2006b) proposed the following design goals for search systems where users have more control in determining relevant results: (i) help people making sense of information (cf. Dervin, 1998); (ii) amplify and reward good intellectual effort; (iii) have flexible architectures; (iv) situate themselves within a broader information ecology of memories and tools; (v) support the information lifecycle from creation to use; (vi) support tuning by users, and (vii) should be engaging and enjoyable to use. By realizing many of these goals, HCIR systems can help people become more effective information seekers and consumers, and more informed individuals generally. Existing capabilities of search systems already provide some support for the objectives listed above. Beyond presenting a ranked list of search results chosen with respect to the provided query, features such as spelling suggestions and suggested queries provide mechanisms to lead the user to potentially relevant content. Importantly from an HCIR perspective, control over selection and interpretation of the suggestions offered still remains with the searcher. Systems are also becoming increasingly aware of their users and their situations that extend beyond the query statements they issue. Personalization employs user models to individualize search results to a particular user’s interests (Teevan, Dumais, & Horvitz, 2005). Relevance feedback (including implicit feedback from interaction behavior and biometrics from physiological sensors) enables search systems to develop richer representations of users’ tasks and search interests by capturing relevance judgments explicitly from searchers (Rocchio, 1971) or mining them from their search interactions (Kelly & Teevan, 2003) or physiological signals (Feild, Allan, & Jones, 2010). The search behavior of other users (primarily queries and clickthrough) can also be mined in the aggregate and used to help rank results (Agichtein, Brill, & Dumais, 2006; Joachims, 2002) or suggest paths to follow through document collections (Wexelblat & Maes, 1999; White, Bilenko, & Cucerzan 2007). Recent research on mining finer-grained search interactions, such as mouse movements to estimate searcher’s gaze attention, is a promising direction for richer behavioral modeling on the Web (Guo & Agichtein, 2010; Huang, White, & Dumais, 2011).

Symposium on Human-Computer Information Retrieval.

Human-computer information retrieval (HCIR) is the study of information retrieval techniques that integrate human intelligence and algorithmic search to help people explore, understand, and use information. Since 2007, we have held an annual gathering of researchers and practitioners to advance the state of the art in this field. This meeting report summarizes the history of the HCIR symposium and emphasizes its relevance to the data science community.

HCIR 2013: the seventh international symposium on human-computer interaction and information retrieval

This report describes the 2012 Symposium on Human-Computer Interaction and Information Retrieval. Now in its sixth year, the two-day symposium (formerly a one-day workshop) was held in October in Cambridge, MA. The event brought together researchers and practitioners from academia, industry, and government and a range of disciplines for in-depth discussions in an informal atmosphere. The symposium attracted 75 attendees, over a third of which were from industry. New for this year, we accepted full papers that will be archived and published in the ACM Digital Library. We continued the HCIR Challenge, this year focusing on the problem of people and expertise finding, five in-depth system demonstrations, and audience selection of a challenge winner.

HCIR 2010: the fourth international workshop on human-computer interaction and information retrieval

This report describes the 2010 workshop on Human-Computer Interaction and Information Retrieval. Now in its fourth year, the event was held in August 2010 in conjunction with the Information Interaction in Context Symposium. The workshop brought together researchers from academia, industry, and government and a range of disciplines to present and discuss their research. We had a record 70 attendees, making this the largest of our workshops to date. New for this year, we ran a challenge, and six research groups participated.