This perspective paper analyzes the rising threat of retracted scientific works and the challenges of preventing the continued spreading and use of the retracted science; further, a framework is proposed for research and actions to effectively manage retractions in the information ecosystem. The precipitous increase in retractions of scientific publications is real and the complexity of retracting publications challenges current IR systems and people's information behaviors. Retracting published, especially peer-reviewed, papers in prestigious venues is a complex phenomenon involving various entities through often time-consuming processes. These publications may be accessible from the original venues, digital archives, or free-access databases, but these systems differ in retrievability and output. Many systems do not identify the retractions or reasons for retractions; most systems do not treat the retracted paper and its related notices (retraction or correction) as an integrated entity. Studies found that many retracted publications continue to be cited post-retraction as valid science. A new threat is the widely spread of retracted publications on social media. Retracting invalid scientific publications has serious implications in the real world. Based on current findings, we propose (1) a framework for further research; (2) a DOI resolution to integrate the documents related to retraction/correction; (3) a structured facet taxonomy for representing and indexing the retracted, corrected, or republished publications in databases; (4) a retraction registry or database with personalized AI helper for researchers to tract retracted publications; (5) an approach for understanding how retracted publications are circulated on social media.

This study compares the 45 “Best Paper” award articles with nonaward articles published in the Journal of Association for Information Science and Technology (JASIST) to observe the differences in citations. The results show that, in most cases, the citations of the award articles are more numerous than the median, belonging to the Top‐50% stratum. Only 15.6% of the award articles have the status of being the most‐cited article of the year in which the article was published; 24.4% belong to the Top‐5% stratum of the publication year; 44.4% belong to the Top‐10% stratum of the publication year; and 73.3% belong to the Top‐25% stratum of the publication year. Surprisingly, from 2000 to 2012, none of the award articles made it to the Top‐10% stratum, apart from the year 2004; the least‐cited award article received only three citations during a 5‐year period. The results show a wide range of citations among the Best JASIST Papers. This study also observes that the number of articles changed little from 1969 to 1995 but grew rapidly from 1996 to 2012. Suggestions for possible ways to better meet the challenges of the journal's growth in size and scope in selecting award articles are provided.

Abstract Session characteristics taken from large transaction logs of three Web search environments (academic Web site, public search engine, consumer health information portal) were modeled using cluster analysis to determine if coherent session groups emerged for each environment and whether the types of session groups are similar across the three environments. The analysis revealed three distinct clusters of session behaviors common to each environment: “hit and run” sessions on focused topics, relatively brief sessions on popular topics, and sustained sessions using obscure terms with greater query modification. The findings also revealed shifts in session characteristics over time for one of the datasets, away from “hit and run” sessions toward more popular search topics. A better understanding of session characteristics can help system designers to develop more responsive systems to support search features that cater to identifiable groups of searchers based on their search behaviors. For example, the system may identify struggling searchers based on session behaviors that match those identified in the current study to provide context sensitive help.

Abstract The authors investigated 11 sports‐related query keywords extracted from a public search engine query log to better understand sports‐related information seeking on the Internet. After the query log contents were cleaned and query data were parsed, popular sports‐related keywords were identified, along with frequently co‐occurring query terms associated with the identified keywords. Relationships among each sports‐related focus keyword and its related keywords were characterized and grouped using multidimensional scaling (MDS) in combination with traditional hierarchical clustering methods. The two approaches were synthesized in a visual context by highlighting the results of the hierarchical clustering analysis in the visual MDS configuration. Important events, people, subjects, merchandise, and so on related to a sport were illustrated, and relationships among the sports were analyzed. A small‐scale comparative study of sports searches with and without term assistance was conducted. Searches that used search term assistance by relying on previous query term relationships outperformed the searches without the search term assistance. The findings of this study provide insights into sports information seeking behavior on the Internet. The developed method also may be applied to other query log subject areas.

Abstract A multidimensional‐scaling approach is used to analyze frequently used medical‐topic terms in queries submitted to a Web‐based consumer health information system. Based on a year‐long transaction log file, five medical focus keywords (stomach, hip, stroke, depression, and cholesterol) and their co‐occurring query terms are analyzed. An overlap‐coefficient similarity measure and a conversion measure are used to calculate the proximity of terms to one another based on their co‐occurrences in queries. The impact of the dimensionality of the visual configuration, the cutoff point of term co‐occurrence for inclusion in the analysis, and the Minkowski metric power k on the stress value are discussed. A visual clustering of groups of terms based on the proximity within each focus‐keyword group is also conducted. Term distributions within each visual configuration are characterized and are compared with formal medical vocabulary. This investigation reveals that there are significant differences between consumer health query‐term usage and more formal medical terminology used by medical professionals when describing the same medical subject. Future directions are discussed.

Abstract Eleven middle school children constructed hierarchical maps for two science categories selected from two Web directories, Yahooligans! and KidsClick! For each category, children constructed a pair of maps: one without links and one with links. Forty‐four maps were analyzed to identify similarities and differences. The structures of the maps were compared to the structures employed by the directories. Children were able to construct hierarchical maps and articulate the relationships among the concepts. At the global level (whole map), children's maps were not alike and did not match the structures of the Web directories. At the local levels (superordinate and subordinate), however, children shared similarities in the conceptual configurations, especially for the concrete concepts. For these concepts, substantial overlap was found between the children's structures and those employed in the directories. For the abstract concepts the configurations were diverse and did not match those in the directories. The findings of this study have implications for design of systems that are more supportive of children's conceptual structures.

Abstract This project analyzed 541,920 user queries submitted to and executed in an academic Website during a four‐year period (May 1997 to May 2001) using a relational database. The purpose of the study is three‐fold: (1) to understand Web users' query behavior; (2) to identify problems encountered by these Web users; (3) to develop appropriate techniques for optimization of query analysis and mining. The linguistic analyses focus on query structures, lexicon, and word associations using statistical measures such as Zipf distribution and mutual information. A data model with finest granularity is used for data storage and iterative analyses. Patterns and trends of querying behavior are identified and compared with previous studies.

Abstract This is Part II of an experimental study investigating students' perceptions of research and research paper anxiety. The study integrates quantitative and qualitative designs to collect complimentary data. The participants were students in four sections of an upper division undergraduate course on technical and professional writing during the fall of 1999. A survey instrument used the Critical Incident Technique to solicit writings in students' own words about a memorable past research and writing experience at the beginning of the semester and the current research and writing at the end of the semester. The quantitative part of the survey measured students' perceptions about research using a questionnaire with five‐point Likert scale, and students' anxiety levels using a standard state anxiety test (STAI Y‐1). The first article, Part I, provides a detailed description of the experimental design and reports on quantitative results. This article reports on content analysis of students' writings about their experiences of the two research projects. Analysis of the data confirmed Kuhlthau's Information Search Process (ISP) model and revealed additional affective and cognitive aspects related to research and writing.

This article reports on the follow-up study of a two-part project designed to study the decision-making process underlying how academic researchers select documents retrieved from online databases, consult or read, and cite documents during a research project. The participants are 15 of the 25 agricultural economics users who participated in the original study of document-selection conducted in 1992. They were interviewed about subsequent decisions on documents considered relevant and selected in 1992, as well as documents cited in their written products but not in the original searches. Of particular interest in this article are the decision criteria and rules they apply to documents as they progress through the project. The first study in 1992 emphasized the selection processes and resulted in a document selection model; the 1995 study concentrates on the reading and citing decisions. The model derived from this project shows document use as a decision-making process with decisions occurring at three points or stages during a research project: selecting, reading, and citing. It is an expansion of the document selection model developed in the 1992 study, identifies more criteria, and clarifies the criteria and rules that are in use at each stage. The follow-up study not only found that all but one of the criteria identified in selection re-occur in connection with reading and citing decisions, but also identified 14 new criteria. It also found that decision rules applied in selection decisions are applied throughout the project.

This article proposes a model of document selection by real users of a bibliographic retrieval system. It reports on Part 1 of a longitudinal study of decision making on document use by academics during an actual research project. (Part 2 followed up the same users on how the selected documents were actually used in subsequent stages.) The participants are 25 self-selected faculty and graduate students in Agricultural Economics. After a reference interview, the researcher conducted a search of DIALOG databases and prepared a printout. The users selected documents from this printout; they were asked to read and think aloud while selecting documents. Their verbal reports were recorded and analyzed from a utility-theoretic perspective. The following model of the decision-making in the selection process emerged: document information elements (DIEs) in document records provide the information for judging documents on 11 criteria (including topicality, orientation, quality, novelty, and authority); the criteria judgments are combined in an assessment of document value along five dimensions (epistemic, functional, conditional, social, and emotional values), leading to the use decision. This model accounts for the use of personal knowledge and decision strategies applied in the selection process. The model has implications for the design of an intelligent document selection assistant. © 1998 John Wiley & Sons, Inc.