How does one study the sheer volume of materials bought and sold online that are archaeological, osteological, or ethnographic in nature? The current answer is 'machine vision', but that raises serious ethical concerns. In this article, we describe our methodology, and the ethical issues we see emerging from this engagement with neural network approaches to machine vision as a way of drawing out the ethical dimensions.
Journalists and scientists alike are increasingly working to expose trafficking of cultural and natural history items of all sorts, both to help raise public awareness and aid in law enforcement efforts (e.g. Williams 2018), as well as to probe the complicated history of collecting of such material (e.g. Redman 2016). With regard to investigating trafficking occurring on social media platforms, Huffer and Graham (2017) detail one approach that focuses on the text of social media posts. However, to focus on the textual only is to miss another obvious trove of data: the photographs themselves accompanying these posts, exemplified by Tilton (2016) and the Yale Photogrammar project (Wexler et al. 2016). In a subsequent article (Huffer and Graham 2018) we explored one approach to examine the actual photographic materials in that corpus using computer vision to categorise these materials at scale.
The point of studying these materials at scale is not to create a definitive sample of the materials, but to have enough materials so that comparisons and useful deductions about broader patterns can be made. The method we used to categorise images based on already built models of object identification (ImageNet and Google Inception v3) is detailed in Huffer and Graham (2018). In the present study, we wish to explore the ways in which transfer learning - where we perturb the existing model so that it learns the categories of materials of interest to us as archaeologists - raises ethical issues that demand resolution before we can continue. To teach the machine to see like an archaeologist may be to teach it to learn the sins of past archaeological and anthropological practice, wrapped in the faux-objectivity of 'artificial intelligence'. What does the machine see? This research, therefore, feeds into broader efforts to understand the dangers and biases behind the use of AI (Noble 2018; O'Neill 2016; Onuoha and Mother Cyborg (Diana Nucera) 2018).
We are collecting social media data connected with the online trade in human remains on platforms such as Instagram, Facebook, Etsy, the Dutch e-commerce site Marktplaats and other sources, both international and country-specific, in order to understand why people buy and sell human remains, and perhaps allow collaboration with descendant communities from whom these remains were taken. Many Indigenous communities remain concerned about the theft, forgery or misappropriation of their cultural heritage, especially human remains either known (or claimed by dealers and collectors) to be their ancestors, at least in a general sense.
To contextualise what we will discuss in this article, we first present the story of Abraham Ulrikab. An Inuk living in the Moravian missionary settlement of Hebron, Labrador, in the late 19th century. Ulrikab was a devout Christian. He was literate, and kept a diary in Inuktitut. In the autumn of 1880, Ulrikab and his family were invited to travel to Germany by the Norwegian ethnologist Adrian Jacobsen, on behalf of Carl Hagenbeck, owner of Hagenbeck's Zoo in Hamburg. The Hebron missionaries discouraged them, but recorded that the promised daily earnings were difficult to refuse, especially since Ulrikab and his family owed significant debts to the missionary store (Ulrikab and Lutz 2005). His diary records the decision-making process, and reveals a desire to visit Europe 'and the [Moravian] communities there' (Ulrikab and Lutz 2005). Abraham, 35, along with his wife Ulrike, 24, and daughters, Sara, 4, and baby Maria, agreed to board Jacobsen's ship and travel to Europe to enrol themselves in a Völkerschau - human zoo - tour. They also convinced Ulrike's nephew Tobias to join them, and helped Jacobsen to recruit three Inuit in Nakvak who had refused to be Christianised (Ulrikab and Lutz 2005).
Jacobsen neglected to have the eight travellers immunised when they arrived, and they fell ill within weeks. Less than four months later, all had died of smallpox (Rivet 2014). With the help of his diary, and that of Jacobsen, researchers have been able to locate the travellers' mortal remains (Rivet 2014). After their deaths, the skeletons of Abraham, Ulrike, Maria, Tobias and Tigianniak from Nakvak were mounted and kept by the Musée National d'Histoire Naturelle. Tiganniak's wife Paingu and daughter Nuggasak died before reaching Paris, and Paingu's skullcap is also in the Paris collection (Rivet 2014). Sara's body was kept by a collection in Berlin (Rivet 2014). The translation of Ulrikab's diary, and further research with European institutions and Nunatsiavut communities in Labrador, has meant the initiation of a repatriation process (Rivet 2015; 2016). Though the Inuit have not yet been returned home, the story and fate of Ulrikab, and his family and fellow travellers prompts one to ask: how many other Abrahams are out there?
Contrast Abraham Ulrikab's story - the result of patient historical research - against the story told by the human remains collecting community (we discuss in depth the texts of collectors' and dealers' posts in Huffer and Graham 2017). Examining the posts themselves and their associated comments and discussion, it seems that it is the adventure story that can be told about these remains that is part of the 'cachet' generating the monetary value exchanged. But the story is not one of lives lived, and they are not stories tied to named individuals. Instead, the focal point is the 'artistic expression', the 'beauty' of the thing. The exotic. The anonymity is part of the point, because it allows the collector to project their fantasies. A heroic collector tells this story of the transformation of human remains into commodities, and a constant search for the allegedly old, authentic, rare and macabre.
The contrast between the 'actual' story and the collector's 'stock' narrative is why we presented the tragic case of Abraham Ulrikab. It is to highlight the colonial violence at the heart of these collections of human remains, many of which have (one way or another) been released into the market, a violence that depends crucially on ways of looking, of consuming, of constructing, an exotic 'other'. It is a literal dehumanising: a named individual becomes a mere thing. Ulrikab's story is not unique - others that spring to mind immediately include the way Sarah Baartman was put on display while living, and her body dissected and displayed after death by George Cuvier within the Musée de l'Homme in Paris; the story of White Fox, a Pawnee who died in Sweden and was eventually repatriated in 1996 (Jibréus 2014); and the thousands of skulls collected under the auspices of colonial governments in Africa (Tharoor 2016), which include such leaders of uprisings as Chief Songea Mbano (Gross 2018). This raises our crucial ethical question concerning using machine vision to study these materials and their contexts of collecting: are we not just replicating a neocolonial violence?
Archaeologists and osteologists must follow numerous codes of best practice and ethical guidelines when proposing or conducting new research (Scarre and Scarre 2006; Zimmerman et al. 2003; Alfonso and Powell 2007; Márquez-Grant and Squires 2018). This includes professional curators of human remains and the researchers and students who seek access to human remains collections (e.g. Márquez-Grant and Fibiger 2011). However, collectors and dealers on the private market are not so bound, regardless of whether they seek archaeological or ethnographic remains, and irrespective of the circumstances leading to those remains surfacing on the (on or off-line) private market (see e.g. Williams 2019, discussing the auction of Anglo-Saxon remains and artefacts, ongoing at the time of writing). Collectors can tell very good stories about why they are doing what they are doing and what makes it appropriate, creating a pronounced us vs them dichotomy between how collectors and scholars can use social media for commerce, research or even community representation (e.g. Huffer 2018). The human remains trade, like many other categories of antiquities trade, exists within complicated legal and moral grey zones (Bowman 2008; Mackenzie and Yates 2017). Our research is being done in part to assess the figurative damage already done to the living individuals represented by the skeletal remains being trafficked.
Finally, the larger goal of our research project is to understand why people are buying and selling these remains or, to put it another way, to understand the linguistic and visual rhetoric employed to 'move' this material. Using digital tools, we want to work out the how and why and where of this traffic, specifically those from digital humanities and data science. Understanding the human remains trade goes beyond simply tracking changes to dollar values or the type of human material being sold, but is as much about addressing the 'post-mortem violence' or metaphorical trauma that has been done to trafficked individuals commonly encountered by law enforcement and psychologists who are charged with rescuing and attempting to counsel the victims of these sectors of the 'red market' (Carney 2011; Fong and Cardoso 2010). When SG and DH applied to the Social Sciences and Humanities Research Council of Canada for funding to expand this research from the foundations laid by DH and other colleagues using manual search methods, one potential outcome that excited peer reviewers was the idea that we might be able to identify descendant communities from these images - to restore some measure of their humanity. Doing so, however, is quite problematic for many reasons (which we discuss in more detail later), not least of which are the difficulties inherent in determining ancestry or population affiliation from human remains using genetic or even osteological data.
Internet Archaeology is an open access journal based in the Department of Archaeology, University of York. Except where otherwise noted, content from this work may be used under the terms of the Creative Commons Attribution 3.0 (CC BY) Unported licence, which permits unrestricted use, distribution, and reproduction in any medium, provided that attribution to the author(s), the title of the work, the Internet Archaeology journal and the relevant URL/DOI are given.
Internet Archaeology content is preserved for the long term with the Archaeology Data Service. Help sustain and support open access publication by donating to our Open Access Archaeology Fund.