One of the great things about spending quality time with the Western Argolid Regional Project datasets is that it gets me thinking about data and digital archaeology more broadly. It is merely a happy coincidence that an a trio of interesting articles on digital archaeology have appeared over the last few weeks.
So for this week, we can do a little three thing Thursday that hits one some intriguing new publications.
Thing the First
I try to read most things that Jeremy Huggett writes and to my mind, he is among the most thoughtful commentators in the field of digital archaeology. His most recent article, in Open Archaeology, titled “Algorithmic Agency and Autonomy in Archaeological Practice” explores the nature of agency in digital archaeology at the moment where we are moving toward more sophisticated and complex digital tools. Huggett considers not only the changing notion of agency in light of the increasingly sophisticated technology used by archaeologists, but also traces a future trajectory that frames the need to consider the ethical implications of digital tools that archaeologists use to make their arguments.
He emphasizes the way that complex algorithms create “black boxes” that obscure the workings of the technology that archaeologists use in their analysis. This is not a kind of luddite alarmism, but instead anchored in a thoughtful understanding of recent trends in our field. For example, Huggett notes that advance in algorithms already allow computers to scan massive numbers of satellite and aerial photographs for patterns that suggest cultural artifacts. Similar technologies may soon allow archaeologists to stitch together highly fragmentary wall painting or identify ceramic forms on the basis of broken sherds. These kinds of technologies rely on algorithms that process far more data and consider nearly infinitely more variables than a human could consider, and this allows them draw unanticipated conclusions that exceed the typical process of hypothesis testing at the core of archaeological inquiry.
These algorithmic processes not only have the potential to disrupt the conventional process of hypothesis testing at the core of academic archaeology, but also produce results in such a way that they far exceed the conventional terms of archaeological explanation. At this point, Huggett would argue, the archaeologist has ceded a good bit of interpretative agency to technologies and algorithms. By giving up an understanding of process, we run the risk of giving up ethical control over our inquiries. We need look no further than recent controversies around facial recognition software that drew on databanks that were overwhelming white and this has created unexpected biases in biometric recognition practices (that tend to discriminate against non-white individuals).
In short, Huggett’s work is pushing archaeology to anticipate the ethical implications of ceding agency to algorithms that often are far more complex than the kind of routine hypothesis testing at the core of conventional archaeological practices.
Thing the Second
Néhémie Strupler’s recent article in Internet Archaeology is a remarkable first step toward a more critical practice in publishing. Titled “Re-discovering Archaeological Discoveries. Experiments with reproducing archaeological survey analysis,” Strupler compares archived and published date from three archaeological projects to the published results from those projects. Needless to say, the results are eye-opening. The data from two of the three projects (including my own Pyla-Koutsopetria Archaeological Project) did not coincide with the results published in their more formal, paper publications.
This posed two problems for Strupler. First, it suggests that existing peer review practices do not extend to exploring the relationship between archived and published data and more traditional, predominantly textual results. This is particularly glaring in the case of the Pyla-Koutsopetria project where the data was published in advance of the formal survey publication (although perhaps not in advance of our manuscript being circulated for review).
The second problem is concerns about the reproducibility of data-driven archaeological argument making. How robust must datasets be – in terms of metadata and paradata – to allow for scholars to reasonably test the results of archaeological analysis. More importantly, how robust must datasets be to allow scholars to go beyond merely testing published arguments, but propose counter arguments or new research directions on the basis of publicly available data. As I am involved in preparing three new datasets for both conventional and digital publication, this article provided some substantial food for thought.
Thing the Third
Readers of my blog know that I’ve been dipping my toe into some local heritage work and CRM. One of things that this work produced was a substantial data set that describes mid-century housing in Grand Forks, North Dakota. The dataset was dutifully submitted to the State Historical Society of North Dakota as a table in a PDF (as they requested) and will for the foreseeable future languish on my hard drive as a flat table.
This all introduces the nice little summative statement offered by Christopher Nicholson, Rachel Fernandez and Jessica Irwin titled “Digital Archaeological Data in the Wild West: the challenge of practising responsible digital data archiving and access in the United States” from Internet Archaeology. As they point out, the current state of digital archiving of archaeological data in the US is a patchwork of practices. Many states, for example, continue to lack policies or procedures for archiving the digital datasets that back many of the reports that CRM and heritage processionals produce on a regular basis. Private CRM firms lack any motivation to make data that they archive available publicly. Local heritage units, such as our Historical Preservation Commission, lack the resources to archive data, reports, and studies that they have commissioned and often look to the state for this or beyond, to the federal government.
In any event, this isn’t meant as a criticism of underfunded state, local, and federal agencies, but rather to note that archaeology as field is still struggling to come to terms with its digital footprint.