Exploring the Navigation Tools: A Sociocultural Perspective on Online Search Technologies

(1)

EXPLORING THE NAVIGATION TOOLS

Fleur Praal; lecturer and PhD candidate (docentpromovendus) at Book and Digital Media Studies, Leiden University Centre for the Arts in Society; her research features the relationships

between authors, publishers and readers in Humanities scholarly communication

A Sociocultural Perspective on Online Search Technologies

Ease of access, increased scale, and improved searchability have been listed as the major advantages of academic publishing’s turn to the World Wide Web from the beginning of that endeavour, around the turn of the century.¹ However, these three newly advantageous characteristics of online scholarly publications can also be viewed as main challenges for users: handling the number of academic publications available has always been a chal- lenge to scholars, but the vast and criss-cross connected web of publications online over- whelms them more than print libraries did, and presses the need for assistance in navigation.

Unlike most other reading audiences, academics cannot afford to miss any text that is relevant in their research field: knowledge production builds on existing knowledge, and the records of preceding academic analy- ses also point out yet uncharted territory. The search for and gathering of relevant scholarly literature is therefore a crucial step in the research process towards formulating new

scholarly ideas. Academics’ need to read and the abundance of publications available have given rise to extensive technical mediation in the process of academic literature search since the rise of modern science: from the eigh- teenth century, scholars have increasingly sys- tematically used publishers’ lists, library catalogues, and each others’ references to relevant texts as instruments to identify and access literature. These traditional tools for the search and discovery of publications have now been overtaken by online equivalents, of which general search engines and social networks are perhaps the most visible and well-known. In this sense, the transition to the digital medium is simply an extension of the interference of technology that has always been present in the search and discovery of literature.

However, as it is used by humans, technology never operates in a vacuum, and therefore innovations always bring about consequences that were unforeseen.² In the case of literature navigation aids, the academic users’ behaviour is also shaped by the existing social

(2)

and cultural habits of the scholarly communication process. Academia’s conventions and the norms of ‘good research’, institutional pressures and expectations, and the intellectual properties particular to research disciplines all interplay with the technology scholars use.

It should therefore be acknowledged that the online tools for searching and discovering texts have not just replaced the instruments of the print era, but rather affect the literature research process itself because they bring new technological affordances.

Such a realization then merits an analysis of the new, online technologies for literature search and retrieval, not on a technological- ly detailed level, but from a conceptual perspective that sheds light on the interactions of these tools with the intellectual and cultural properties inherent to research disciplines.

Assessing the online tools grouped by their technological functionality can bring to light unintended consequences for Humanities scholarship, through information research. To a degree, it can be assumed that all academics search for literature with the same utilitarian approach: they want to quickly select and col- lect texts, to find as much relevant material as possible in as little time as possible.³ However, within this pragmatic intellectual approach, precisely what qualifies as relevant material varies with the cultures of the academic disciplines.

The Humanities disciplines seek knowledge that is not, as in the Sciences, based on discovery and explanation, but rather on un- derstanding and interpreting aspects of the human experience.⁴ This analysis is provided in qualitative, argumentative, and often highly value-laden analytical texts. Articles in the Humanities are on average longer than in the STEM-fields, and besides articles, Hu-

manities scholars use monographs to com- municate deep exploration of and sustained analysis on a particular niche interest. In the publications themselves, that specific analysis has to be situated in the context of preceding and concurrent research: individual Human- ities scholars choose to work in a wide array of interests. Contrasting with science fields, in which collective efforts are concentrated on a specific set of problems that drive them, the scholarly endeavour evolves around connecting new arguments to existing ones, making interpretative voices relevant by contextual- isation. Although in the research on information-seeking there is very little consensus about ‘typical’ strategies and behaviour in the Humanities, there has long been general agreement that the intrinsic characteristics of these disciplines translate into Humanities scholars’ use of search and retrieval tools.⁵ The use of multiple text types, the conceptual and non-literal arguments that shape its discourses, and the wide scattering of research interests render information seeking in the Humanities fundamentally complex.

Cognitively, metadata-based search is the process of literature retrieval in which technology can most straightforwardly assist: if users have already selected criteria to which their targeted publications should conform, they can direct a search based on those – for instance, all articles from a particular author, in a specific year, or with a precise term in the title. For users who are less sure what they want to find, subject classification systems are the traditional technological support, as they allow browsing categorized publications increasingly specifically based on their topic. Publishers’ practice to organise articles in research journals, and monographs in series based on a common topic, has traditionally helped users, too: these channels highlight the topical relations between otherwise separate

(3)

publications. Once users identify a channel that provides them with a relevant text, exploring the other texts in the same channel (journal, series) is a logical next step. This provision of a relevant context for publications would be conceptually grouped under what Michael Bhaskar has named “amplification activities”: efforts of content providers in ensuring a text reaches its intended audience.⁶ Catalogues and channels, found in libraries, publishers, and disciplinary aggregators such as JSTOR and ProjectMuse are the traditional instruments of amplification provided by the incumbent agents

in the realm of scholarly communication.

Publishers and libraries have always had a clear incentive for providing the technology that assists users navigating the wide array of available literature:

their production

and acquisition of content is only effective if those texts actually reach their reading audiences. To a degree, publishers and libraries legitimise their existence by their amplification and curation services. The introduction of the online medium, however, opened the market for navigation services produced by parties that do not have direct links with content production or ownership. Connecting, perhaps, with the popular adage that the internet should provide access for all, the new search and retrieval tools increasingly move towards business models that do not depend on content amplification, but rather on offering users free range in all online texts available. Whereas publishers and libraries subside

by providing content and produce navigation technology as tools to approach that content with, the new and independent navigation tools seem to depend mostly on user traffic – to any content at all. User data has become an important commodity, but, perhaps due to undisclosed business models, users seem to develop loyalty to such new navigation instruments quickly and more sustainably than their relations with content providers.⁷ This renders analysis of all the more pressing.

The most visible competitor of the traditional instruments for search and retrieval, and one that has gen- erated unparalleled loyalty among its users, is Google’s academic platform Google Scholar.

Since its introduction in 2004, Goo- gle Scholar crawls the web of publishers’ databases, scholarly networks and institutional reposito- ries for academic publications on an unprec- edentedly large set of parameters: it indexes the metadata and full text of any academic publication it can find, weighs in the channel in which it appears and the number of times it was already cited, and based on this allows searching on criteria and on terminology that occurs in the text – through one search bar.⁸ This ease of querying and the convenient result lists have tremendous appeal among Humanities scholars: in a 2012 UK survey, 40% favoured to begin their search for literature through Google Scholar – which sur- passes the share of them who initially turn to disciplinary aggregators such as JSTOR and

“Whereas publishers and libraries subside by providing content

and produce navigation technology as tools to approach that content with,

the new and independent navigation tools seem to

depend mostly on user traffic – to any content at all.”

(4)

to library catalogues, combined.⁹ Although it must be noted that these responses only reflect users’ primary choice of channel, and that many respondents, especially from the Social Sciences & Humanities, indicate that they often combine several sources to navigate scholarly literature, this finding seems to demonstrate the quick uptake of Google in research practices.¹⁰

Academia’s rapid and relatively unencum- bered embrace of Google Scholar is surpris- ing, because the generic search engine technology has inherent cognitive limitations for Humanities information retrieval. The fact that the search engine so broadly indexes all academic materials is a complication: it does not only retrieve Humanities articles, but also crawls databases of literature from the STEM-fields – which produce the absolute overwhelming majority of publications.¹¹ Generic search engines therefore inundate Humanities texts with other search results, much more than discipline-specific search platforms. This could render it difficult for scholars to retrieve relevant publications.

This problem is compounded by the fact that many Humanities keywords are quite generic terms that can carry many different mean- ings, as well as the added difficulty that Hu- manities texts come in other languages than English. Although the search results can be fine-tuned in a number of ways – by using Boolean operators, filters, specific domains, and the use of subject categories – scholars on Google report on using simple and advanced keyword searches mostly, and thus retrieve a haystack of results that contain certain terminology which they have to manually sift in search for their proverbial needle.¹²

The encroaching popularity of the search engine as a multipurpose tool for use in

any information context has been dubbed

“Googlization”: scholars increasingly approach academic literature search via the same access point as all other media for all other occasions, automatically and almost in- tuitively.¹³ Although it presents the scholar’s task as easier, it has been argued that this approach through everyday tools runs the risk of rendering the search for literature as a simple, mundane process, instead of the complex specialists’ practice it really is.¹⁴ As Bowker et al. assert, our interactions with information technologies influence our perception of their properties, and the most intensively used tools and infrastructures tend to become transparent or invisible; the pitfall with fre- quently using the same tool or navigation aid is that users stop considering its precise technical workings upon every use.¹⁵ Follow- ing this model, the increasing omnipresence of Google would desensitise users towards its limitations. Indeed, it has been noted that Humanities scholars are aware of the limit- ed accuracy of Google Scholar’s search results and that they trust the search engine less than libraries and discipline-specific content portals such as JSTOR – but despite these misgivings, they continue to use the search engine as their primary access point.¹⁶ Goo- gle’s pervasiveness in everyday life seems thus to override concerns for intellectual rigour in the academic information retrieval process;

its technological properties might not fit Hu- manities research that well, but mundane sociocultural factors account for it being heavily used regardless.

Social factors also underpin scholarly communication networks, the second type of new information retrieval technologies used widely in academic communication. Google Scholar rolled out its social network features in a major update of functionalities in 2012, by which time Mendeley, ResearchGate and

(5)

Academia.edu had already been available for several years.¹⁷ These academic networks, perhaps partly due to the parallel popularity of Facebook and Twitter, have been rapidly assimilated in the communication practices of academics: ResearchGate counts over 11 million user profiles, and approximately two- thirds of researchers report to have an account on this network.¹⁸ Academia.edu is even larger, perhaps because it does not require an institutional email address to verify account registration: with 36 million users, it is also the preferred platform among Humanities scholars, 59% of whom maintain a profile there.¹⁹ These “Swiss army knives of scholarly communication”, as Bosman and Kramer aptly call them, offer their users a myriad of functionalities for uploading papers, sharing data, asking questions, and collaborating in projects. No doubt these are aimed at rendering the platforms attractive for use throughout the research process; however, researchers indicate that they primarily use the platforms by uploading the texts they have produced, and for navigating literature created by others in their network.²⁰

Using academia’s social ties to discover relevant literature is not at all a novel phenomenon in research. Since long before the advent of the web, standard scholarly strategies in- clude “citation chaining” and “monitoring”, both strategies that depend on identifying authoritative nodes in networks of texts and people (defined under these terms in information behaviour research).²¹ When a user finds a text authoritative, relevant, and of quality, he will probably infer that the texts upon which it builds will be of similar nature;

linking back to these references thus allows browsing through a network of related, authoritative texts on a certain research topic.

Monitoring, on the other hand, is the widely spread practice of identifying a scholar as au-

thoritative on a subject, and then checking all new materials by that person or in that channel. Chaining and monitoring have always existed as viable navigation strategies, but computer networking technology stretched them to massive scale with innovative ease and reduced user effort: instead of having to actively identify and trace patterns of expertise between texts and scholars, users just log on to a network and connections present themselves – together with quantified indicators for authority: counts of followers for a person, and shares, likes and links for texts.

It is inherently befitting for academia to allow authoritative voices to direct the navigation through scholarly literature: the most valuable contributions to knowledge should attract many readers. Yet the assessment of authority, the decision whether to second a particular author’s analysis, is one of the readers’ fundamental responsibilities. The fact that others think highly of a particular text justifiably contributes to readers’ assessment of quality; others’ appreciation may be a relevant sign of authority. Such inference is a constructive mechanism. However, researchers run the risk of over-emphasizing others’

appreciation in their own assessment: they tend to read what others cite, and thus indi- rectly recommend. This tendency is named the Matthew effect, after the gospel parable where the rich get richer, and the poor get poorer: frequent use can lead publications to attract even more readers.²² As this effect was already present before the online medium, the presence of quantitative indicators of reading, downloading, sharing, citing, and linking can only exacerbate it on scholarly communication networks. This is especially relevant for the Humanities, because the individual interpretation of others’ trustwor- thiness and quality itself is a principal part of research. Rigour of research does not de-

(6)

the user’s network is favoured in the display of results, and its periphery remains out of sight.

Although the technological properties of new online navigation tools perhaps fit researchers’ usage preferences, they clash with the epistemological requirements of research in general, and of the Humanities more specifically: generic search engines are deceptively easy to use, but result in too many results based on generic keywords, or in none at all for specific or non-English ones. Scholarly communication networks, on the other hand, offer the certainty of browsing only relevant results through webs of relations, but make users disinclined to navigate away from the trodden paths.

Interestingly, alternatives rise that aim to counter some of these negative side-effects, by accommodating users’ preferences for easy interfaces, and yet offering a breadth of quality results, among which unexpected ones. Yewno, for instance, is an experimental semantic-analysis search engine: the user provides keywords as for a regular Google search, but instead of string-matching these words in a text, the engine looks for the concepts and ideas they point to, and presents results that also cover those concepts, even if expressed in different terms.²⁵ The user can then influence pend on falsifiable hypotheses or repeatable

empirical results, but on arguments that are analysed, pondered, and poked by the reader.

Potential de-activation of readers’ assessment, as provoked by social networks’ quantitative proxies for authority, therefore put Human- ities research fundaments under pressure.

Moreover, the use of online platforms for navigating scholarly literature, although ap- preciated for its increased efficiency, is likely to narrow results of search and browsing.²³ As the serendipitous discovery of related literature is important in knowledge creation, this is problematic – and, once again, perhaps most strongly evident in the Human- ities. In the specific, problem-driven settings of most STEM-fields, any literature not di- rectly pertaining to the problem can perhaps be dismissed, but on their remote islands of Humanities scholarship, researchers should remain aware of “adjacent areas of interest or even potential areas of unknown interest”.²⁴ Situating one’s own research in the contextual landscape of related argumentation is an important cognitive aspect of Humanities research, but this can be considered to become more difficult with the reliance on scholarly network technology for the navigation of that landscape, as literature from the centrality of

“Instead of having to actively identify and trace patterns of expertise between texts and scholars, users just log on to a network and connections present themselves – to- gether with quantified indicators for

authority: counts of followers for a person,

and shares, likes and links for texts.”

(7)

the navigation process, by indicating which retrieved relations the tool should explore further, and which it should ignore. Yewno’s founders hope that their tool, with its basic keyword search and visual results display, can be used in creating awareness of the scope of a research topic and its situatedness in a context of navigable literature, especially for students and junior researchers at the beginning of a new research project.²⁶ Further in the research process, scholars can use JSTOR’s Text Ana- lyzer that also runs a semantic analysis: this instrument extracts a set of related, relevant keywords from any uploaded text, and sug- gests literature based on co-occurrence of these terms.²⁷ The process is not fully auto- mated, but requires iterative refining actions by the user, who must indicate the relevance of indexed keywords on sliders. The literature suggestions then shift accordingly. Comput- er-aided navigation alternatives such as these accommodate the users’ wish for a straight- forward interface, but also force them to take an active approach in narrowing or directing initial results, instead of leaning back and see- ing where the network or search engine takes them.

These advanced search and discovery tools combine harnessed computing power with a reaffirmed emphasis on user control. This reasserts the notion that innovative ways to navigate scholarly literature are not merely engineered with ever-advancing technological instruments, but rather iteratively construct-

ed in interplay between such technology and the sociocultural and cognitive processes that drive its academic users, often tacitly. Schol- ars should not necessarily be wary, but always aware of the technological capacities of the instruments they use and the side-effects those may cause in their retrieval of scholarly literature; they should be equipped to assess to what extent such navigation tools fit the research culture in their field, and encouraged to explicate such epistemological incentives.

Attentive scholarly users, publishers, and libraries should also critically examine the business models of novel navigation tools.

The traditional stakes in the scholarly communication process have always been clear:

publishers and libraries want to produce and amplify content, so that it reaches the targeted scholarly audience. New search tools are created by incumbents that do not share all of these traditional goals, because they do not produce or host content themselves, they merely point to it. They do therefore not have any interest in matching content and reader, except to the extent that the reader should be sufficiently satisfied to return to the platform:

user traffic itself, in clicks, shares, and likes, is what drives commercial enterprises such as Google and Academia.edu alike.²⁸ Most tools are not, by intent or purpose, launched to re- direct use from the business of libraries and publishers.²⁹ However, the increased com- petition among navigation services calls for diligent assessment and conscientious use, to ensure every text finds its reading audience, and every user finds relevant texts.

1. See for instance J. Thompson, Books in the Digital Age: The Transformation of Academic and Higher Education Publishing in Britain and the United States (Cambridge: Polity, 2005), ch. 13.

2. A. van der Weel, Changing Our textual minds: Towards a Digital Order of Knowledge (Manchester: Manchester University Press, 2011).

3. A. van der Weel notes this in ‘Reading the Scholarly Monograph: From TL,DR to LBW;PR’, TXT Magazine (2015), pp. 75-81, and refers to C.

Tenopir & D. W. King’s observations in: ‘Electronic Journals and Changes in Scholarly Article Seeking and Reading Patterns’, D-Lib Magazine 14.11/12 (2008).

4. I will use Science, Technology, Engineering and Medicine (STEM-fields) as a contrast to Social Sciences & Humanities (SS&H) throughout this text, but it should be noted that this dichotomy is an oversimplification: although practices differ more between STEM and SS&H in general, some Social Sciences behave less like the Humanities than some theoretical STEM-fields do. Moreover, the Humanities disciplines are by no

(8)

means a homogeneous group – although here, they will be treated as such. See also: T. Becher & P. R. Trowler, Academic Tribes and Territories, 2nd ed. (Buckingham: SHRE & Open University Press, 2001).

5. E. Whitmire, ‘Disciplinary Differences and Undergraduate Information-Seeking Behaviour’, Journal of the Association for the Study of Informa- tion, Science, and Technology 53.8 (2002), pp. 631-638.

6. M. Bhaskar, The Content Machine (London: Anthem Books, 2013), especially ch. 5 ‘Models’.

7. O. Sundin, ‘The Search-Ification of Everyday Life and the Mundane-Ification of Search’x Journal of Documentation 73.2 (2017), pp. 224–243.

DOI: http://dx.doi.org/10.1108/JD-06-2016-0081.

8. Google, ‘About Google Scholar’, <https://scholar.google.com/intl/en/scholar/about.html>.

9. R. Housewright, R. C. Schonfeld & K. Wulfson, UK Survey of Academics 2012 (Ithaka S+R | Jisc | RLUK: 16 May 2013), pp. 21-28, ‘Discovery’;

especially fig. 7, p. 23.

10. Ibid., p. 27; confirmed in: M. Kemman, M. Kleppe & S. Scagliolia, ‘Just Google It: Digital Research Practices of Humanities Scholars’, in: C. Mills, M. Pidd & E. Ward, Proceedings of the Digital Humanities Congress 2012 (Sheffield: HRI Online Publications, 2014).

11. Figures like these are notoriously difficult to compile, but the number of articles is approximated at 1.5 million annually, and academics in the STEM-fields produce more publications than Humanities scholars. The market for SS&H-publishing is assessed at a quarter of the value of STEM-publishing – and within that share, Social Sciences are much larger than the Humanities. See: J. Esposito, ‘The Market for Social Science and Humanities Publications’, The Scholarly Kitchen, 28 January 2014. Online: <https://scholarlykitchen.sspnet.org/2014/01/28/the-market-for-social-sciences-and-humanities-publications/>.

12. Kemman, Kleppe & Scagliola, ‘Just Google It’, fig. 6.

13. S. Vaidnahyanatan, The Googlization of Everything (And Why We Should Worry) (Oakland: University of California Press, 2011).

14. A. Halavais, Search Engine Society (Cambridge: Polity, 2009); O. Sundin et al., ‘The Search-Ification of Everyday Life and the Mundane-Ification of Search’, Journal of Documentation 73.2 (2017) pp. 224 – 243. DOI, http://dx.doi.org/10.1108/JD-06-2016-0081.

15. G. C. Bowker et al., ‘Toward Information Infrastructure Studies: Ways of Knowing in a Networked Environment’, in: J. Hunsinger et al. (eds.), International Handbook of Internet Research (New York/London/ Dordrecht: Springer, 2010), pp. 97-117.

16. Kemman, Kleppe & Scagliola, ‘Just Google It’, fig. 5.

17. Mendeley was founded as an independent commercial company in 2007 (and acquired by Reed Elsevier in 2013); Academia.edu and Re- searchGate were both founded in 2008 and continue as independent businesses until today.

18. Number of profiles as self-reported by the platforms, March 2017. User preferences in: J. Bosman & B. Kramer, ‘Swiss Army Knives of Scholarly Communication - ResearchGate, Academia, Mendeley and Others’, Presentation for STM Innovations Seminar, 7 December 2016. Online available: https://doi.org/10.6084/m9.figshare.4290428.v1.

19. The disparity between survey respondents is striking: Arts & Humanities scholars in majority prefer Academia.edu, and those from STEM and Social Sciences overall use ResearchGate. This merits further analysis, but not in this article.

20. Bosman & Kramer, ‘Swiss Army Knives of Scholarly Communication’.

21. D. Ellis, ‘A behavioural Approach to Information Retrieval Design’, Journal of Documentation 45.3 (1989), pp. 171–212; via: T.D. Wilson, ‘Models in Information Behaviour Research’, Journal of Documentation 55.3 (1999), pp. 249–270. DOI:10.1108/EUM0000000007145.

22. R. K. Merton, ‘The Matthew Effect in Science’, Science 159 (1968), pp. 56–63. DOI: 10.1126/science.159.3810.56.

23. This argument is explored further in: F. Praal & A. van der Weel, ‘Taming the Digital Wilds: How to Find Authority in an Alternative Publication Paradigm’, TXT Magazine (2016), pp. 97-102; it is demonstrated for hyperlinks in: J. E. Evans, ‘Electronic Publication and the Narrowing of Science and Scholarship’, Science 321 (2008), pp. 385-398. DOI: 101126/science.1150473.

24. R. C. Schonfeld calls this the “penumbra” in: ‘Personalizing Discovery without Sacrificing Serendipity’, The Scholarly Kitchen (30 March 2015).

<https://scholarlykitchen.sspnet.org/2015/03/30/personalizing-discovery-without-sacrificing-serendipity/>.

25. At the time of writing this article (March 2017), Yewno is available for beta testing on a selected number of institutions only, but J, O’Neill provides a detailed account of her experience with a Humanities query, and explains the underlying technologies: ‘Have You Looked at This?

Yewno’, The Scholarly Kitchen (13 July 2016). <https://scholarlykitchen.sspnet.org/2016/07/13/have-you-looked-at-this-yewno/>.

26. O’Neill, ‘Have You Looked at This? Yewno’.

27. JSTOR ‘Text Analyzer Bèta’: <http://www.jstor.org/analyze/analyzer> (version 0.2.58, accessed 20 March 2017).

28. Despite its top-level domain extension, Academia.edu is not an educational institution. See: K. Fitzpatrick, ‘Academia, not Edu’, Planned Obso- lescence, 26 October 2015. <http://www.plannedobsolescence.net/academia-not-edu/> (20 March 2017).

29. Ironically, pirate aggregator SciHub, founded in 2011, is one of the few not-for-profit navigation and access platforms, but it is designed to bypass libraries and publishers, as it illegally harvests research publications – over 58 million at the time of writing this article – and allows direct access via an intuitive interface. Understandably, publishers are vexed by the quick rise of this illegal phenomenon: in a major survey by Science, 40% of the more than 11,000 respondents confirmed that they use the platform, despite the fact that many of them have legal tools at their disposal, too. Its uptake, however, seems to lag behind in the Humanities: although no tailored use patterns are available, only two scholarly articles feature in the top 100 of SciHub downloads, and then at the very bottom of the list, based on download counts. See: J. Travis,

‘In Survey, Most Give Thumbs-Up to Pirated Papers’, Science, 6 May 2016. DOI: 10.1126/science.aaf5704; M. Parkhill, ‘SciHub: The Academic Cat Is Out of the Bag’, Plum Analytics, 16 May 2016 <http://plumanalytics.com/sci-hub-academic-cat-bag-post/>.