Chemical Literature Databases: Conflict of Interest?

IntroductionMany articles have been written that compare and contrast the primary databases used by chemical researchers (Baykoucheva 2011; Gavel & Iselid 2007; Grey et al. 2012; Li et al. 2010). These and other articles have considered similarities, differences, advantages, and disadvantages in databases in areas such as citation analysis, h-factors, number of journals indexed, number of articles retrieved, user friendliness and much more. One aspect of chemical information retrieval for which there is a lack of discussion is the possible conflict of interest that arises when the publisher of scientific journals is also the publisher of a scientific database. Does such a publisher intentionally (or unintentionally) establish algorithms within its database that steer users toward its own publications? Even if one had access to the proprietary details of a database's algorithms, can such a question be answered with any degree of certainty? Discovery through a database depends on both the algorithms of the database and the metadata attached to each journal article. A publisher of a database and a journal indexed in that database obviously has significant control over both ends of this process and could introduce a bias that favors discovery of its own publications. This paper attempts to raise an awareness of this possible conflict of interest.First and Foremost, Awareness of the IssueBecause this paper is unique in the idea it explores, a note on the underlying message is appropriate. Unlike much of the literature related to research databases, this paper is not meant to evaluate how a given research database works (its algorithms), nor is it about the options a database presents to users, nor about the results that could be obtained if the database were used in a proper (or even moderately proper) manner. The data presented below look at the results that a typical user would obtain during a typical search and how those results might vary from database to database. The intent of this paper is not to reexamine other papers' studies on the mechanics of using databases. Search results and whether those results show any evidence of publisher bias is the sole focus. Certainly, the algorithms and options presented by a research database have a significant effect on the results obtained. However, it is no secret that users of databases display less than desirable habits when searching for literature and database providers are most definitely aware of user habits. In short, the results obtained by a typical user are pertinent to this work; how or why those results are obtained is not pertinent to this work, if one accepts that the majority of database users approach searches in a similar (less than ideal) manner. The foremost aim of this work is to bring attention to a possible conflict of interest and in doing so it looks at the results that a typical user would obtain, rather than the results a librarian would obtain or the results librarians believe users could be obtaining. There is ample library science literature indicating that the typical user puts forth as little effort as possible when performing a search in a research database. In fact, as Markey (2007) specifically noted in her review of literature related to end-user searching habits, "[m]ost end users accept the IR system's default values" (Bishop et al. 2000; Cooper 2001; Jones et al. 2000; Markey 2007). This holds not only for less experienced searchers, but also for more experienced searchers. Jones et al. (2000) gathered results from over 30,000 queries over a 61-week period using transaction log analysis of searches performed in the New Zealand Digital Library interface. Similar to most database interfaces, this user-friendly interface provides users with several query options in the default search screen and additional query options in the advanced search screen. Jones et al. concentrated their analysis on searches of Computer Science Technical Reports with the assumption that the "computer science research community could be thought of as 'best case' users . …