OVERVIEW OF THE CONTRIBUTIONS

The five essays in this special section result from the Society for American Archaeology forum discussions. Four are authored by forum discussants, and one (Cook et al.) is an article we solicited in order to include insights from colleagues who teach with digital archaeological datasets (including, as a coauthor, Timothy Goddard, who was a named panelist but was unable to attend the forum). We would also like to highlight two additional publications, in press at the time of the forum and authored by forum participants (Richards and Brin), which, in discussing the Archaeology Data Service and tDAR, touch on some of the themes discussed in this special section (McManamon et al. Reference McManamon, Kintigh, Ellison and Brin2017; Richards Reference Richards2017).

This special section considers digital archaeology beyond the current status quo of “data management” to better situate the sharing and reuse of data in archaeological practice. Within this theme, articles discuss data stewardship and preservation, new pathways for interpretation and science, the place of “Big Data” in archaeology, public engagement, transparency, public policy, compliance, and improving digital literacy. Contributions grapple with the realities of working with digital data, ensuring access to archaeological data in the future, and considering the ethical implications of data access.

Since Huggett's blog post helped inspire the Society for American Archaeology forum, we begin this special section with his contribution. Huggett, Marwick and Pilaar Birch, and Sobotkova all provide invaluable discussions about what we mean by data, use, and reuse. Different definitions of data help shape our understanding about reuse practices. Huggett highlights the fact that reports (presumably in PDF format), and not data, dominate content and use in many digital repositories. He sees far greater use of these reports than of files containing structured data (databases, spreadsheets, tables, and the like). To Huggett, the greater use of reports over structured datasets illustrates how data reuse remains limited in archaeology. In contrast, reports (as PDF documents) represent a key source of “data” to Sobotkova, even though this format requires a great deal of largely manual effort to process in order to extract analytically usable information.

The (stubborn) centrality of the article or report seems evident in citation practices. As noted by Huggett and by Marwick and Pilaar Birch, archaeologists are habituated to citing literature, not structured, readily computable data. If, in citing data, researchers fail to use digital object identifiers (DOIs), a persistent identifier tracked in library and publishing systems, then reference to datasets’ citations will hide in obscurity. Even altmetrics, which typically measure references to DOIs, may miss data reuse. Interestingly, an important reuse of Open Context (the system we manage) published data centers on aggregation in another scholarly information system, Pelagios, as “Linked Data.” The Linked Data paradigm emphasizes Web identifiers (Uniform Resource Identifiers) more than DOIs. Similarly, in order to contextualize data with rich metadata, Open Context regularly references Uniform Resource Identifiers for concepts in data stores such as PeriodO (for chronology), the Getty Art and Architecture Thesaurus (for general typology), Pleiades (for ancient Mediterranean places), and more. Yet these linked data methods of reuse would not register in DOI-centered data citation metrics. Regardless, the lack of widely applied data citation practices reflects the relative novelty and rarity of citing data.

At least in the context of the United Kingdom, Huggett notes that archaeologists are much more likely to deposit digital data in a repository than they are to make use of digital data in a repository. Marwick and Pilaar Birch report that fewer than 1% of archaeological datasets have been cited. Digital data management has come into archaeology as an “archival” activity. Data repositories archive and preserve data. It may be that many archaeologists see their main obligation with respect to data to center on preservation. A perspective of “data are for preservation” may impede reuse. As Huggett notes, if conventional physical repositories see little use, then why should we expect digital repositories to see much use?

Faniel and colleagues also touch on this issue of archival vs. analytic purposes of data. They report findings from two excavation sites where excavation teams created data to record field observations without necessarily optimizing their data creation practices to support analysis. They document problems managing identifiers, ensuring consistency in recording (including heavy use of “free-text” fields), and difficulties in integrating specialist data. To Faniel and colleagues, data quality and modeling problems can be better identified and addressed if archaeologists prioritize making more analytic use of their own databases.

Although both excavation sites provided field school training, Faniel and colleagues find that training in data management needs to be more systematic and formalized. They also note how field schools should encourage student use of excavation databases to better teach how data creation practices impact later reuse. Cook and colleagues expand on the instructional and training challenges of archaeological data management. They highlight that although students are habituated to digital social media, undergraduates need to develop basic computational skills before they can work with even simple archaeological datasets, much less sophisticated linked open data resources and Web application program interfaces (enabling software interactions with an information system). In their experiences teaching with Open Context, the Archaeology Data Service, and Data Archiving and Networked Services (the Dutch national digital research archive), the steep learning curves and complexity of the data and information systems made student engagement with archaeological data very difficult.

The points Cook and colleagues make about complexity barriers to reuse help us better understand some instances of data reuse that have occurred. For example, the past few years have seen the tremendous success of the Portable Antiquities Scheme in catalyzing many research programs, including more than 100 doctoral dissertations and 20 major peer-reviewed publications (see https://finds.org.uk/research). tDAR has powerful data integration tools (McManamon et al. Reference McManamon, Kintigh, Ellison and Brin2017), as demonstrated in a recent publication discussing research outcomes of the analysis of aggregated zooarchaeological data (Kintigh et al. Reference Kintigh, Spielmann, Brin, Selçuk Candan, Clark and Peeples2017). Open Context has used “Linked Open Data” approaches to integrate zooarchaeological datasets, as well as government-created archaeological site file data from the Digital Index of North American Archaeology (Anderson et al. Reference Anderson, Bissett, Yerka, Wells, Kansa, Kansa, Noack Myers, DeMuth and White2017). Some of these data have already supported multiple publications by different teams of researchers (Arbuckle et al. Reference Arbuckle, Whitcher Kansa, Kansa, Orton, Çakırlar, Gourichon, Atici, Galik, Marciniak, Mulville, Buitenhuis, Carruthers, De Cupere, Demirergi, Frame, Helmer, Martin, Peters, Pöllath, Pawłowska, Russell, Twiss and Würtenberger2014; Atici et al. Reference Atici, Pilaar Birch and Erdoğu2017) as well as instructional use described by Cook and her coauthors. These examples of reuse seem to center on data that have sufficient scale and consistency to attract attention and to motivate researchers to invest time and effort in understanding and analysis. Researchers will probably be more willing to invest time in navigating a complex information resource if it has sufficient relevance and significance.

Engaging with complex data requires time, appropriate skills (especially for technically challenging resources), and motivation. Sobotkova had the motivation to reuse “data” from PDFs, and she had the background in Bulgarian archaeology required to understand these reports. Sobotkova nonetheless regrets that archaeological training did not provide her with the skills to extract information out of large numbers of PDFs using computer-assisted text analysis, a lost opportunity for archaeology since so much information and infrastructure is built around long-form PDF reports. Unfortunately in the case of archaeology, the computer skills required to work with structured data, or text as “data,” seem more rarified. Conventional literacy, the kind of literacy required to read PDFs, is widespread, while the data literacy required to unlock value from structured data that would allow us to treat large amounts of PDFs as corpora for analysis is not. We cannot just assume that this skill gap will be solved as “digitally native” students advance in the profession.

On this last point, Cook and colleagues also note how students quickly adopt faculty's tacit opinions about the relative value and prestige of reusing other people's data. With so few examples of data reuse to emulate, students may doubt that they will gain much by investing the time and effort to develop the skills and background needed to work in data reuse. Indeed, every contributor notes that data still fit awkwardly into archaeology's system of professional rewards and evaluation. Marwick and Pilaar Birch note poor enforcement of the data availability statements required in one of archaeology's most prestigious journals. They also note a widespread reluctance to share data, even data of low risk and low sensitivity to stakeholders (including indigenous communities).

Marwick and Pilaar Birch note that an impediment to the uptake “of data sharing in archaeology and other ‘small’ sciences is that [it] is unfunded, unrewarded, and only rarely required.” While this is not a new observation, they do suggest several new strategies to address the problem. They suggest a formal citation structure and reforms to editorial policies, as well as stricter enforcement of such policies, to make data disclosure a routine requirement for peer-review publication. A focus on enforcing data availability and consistent data citation practices in conventional journal publication will indeed change the incentive landscape and help drive expectations and cultural changes around the role of data in scholarly communications.