Buttons figure large in the world around us. Just in the last year we’ve seen everything from presidents boasting about the size of their nuclear buttons to Apple being faced with a class action over the failure of their new ‘improved’ butterfly keys to Amazon’s Dash buttons being barred in Germany for not providing information about price prior to being pressed. In archaeology, we’ve become accustomed to buttons and button-presses generating data, performing analyses, and presenting results, ranging across the digital instruments we employ and the software tools we rely on. So, to pick a random example, “researchers will be able to compare ceramics across thousands of sites with a click of the button.” (Smith et al 2014, 245).
Rachel Plotnick has recently discussed the place of buttons in our cultural imaginary:
… push a button and something magical begins. A sound erupts that seems never to have existed before. A bomb explodes. A vote registers. A machine animates, whirling and processing. A trivial touch of the finger sets these forces in motion. The user is all powerful, sending the signal that turns on a television, a mobile phone, a microwave. She makes everything go. Whether or not she understands how the machine works, she determines the fate of the universe. (Plotnick 2018, xiv).
This seemed fair enough at the time, if admittedly challenging. What I hadn’t appreciated, though, was the controversial nature of such a claim. For sure, in that piece I referred to Yoshua Bengio’s argument that we don’t understand human experts and yet we trust them, so why should we not extend the same degree of trust to an expert computer (Pearson 2016)? But it transpires this is quite a common argument posited against claims that systems should be capable of explaining themselves, not least among high-level Google scientists. For example, Geoff Hinton recently suggested in an interview that to require that you can explain how your AI systems works (as, for example, the GDPR regulations do) would be a disaster:
When we hear of augmentation in digital terms, these days we more often than not think of augmented or mixed reality, where digital information, imagery etc. is overlain on our view of the real world around us. This is, as yet, a relatively specialised field in archaeology (e.g. see Eve 2012). But digital augmentation of archaeology goes far beyond this. Our archaeological memory is augmented by digital cameras and data archives; our archaeological recording is augmented by everything from digital measuring devices through to camera drones and laser scanners; our archaeological illustration is augmented by a host of tools including CAD, GIS, and – potentially – neural networks to support drawing (e.g. Ha and Eck 2017); our archaeological authorship is augmented by a battery of writing aids, if not (yet) to the extent that data structure reports and their like are written automatically for us (for example).
In 2014 the European Union determined that a person’s ‘right to be forgotten’ by Google’s search was a basic human right, but it remains the subject of dispute. If requested, Google currently removes links to an individual’s specific search result on any Google domain that is accessed from within Europe and on any European Google domain from wherever it is accessed. Google is currently appealing against a proposed extension to this which would require the right to be forgotten to be extended to searches across all Google domains regardless of location, so that something which might be perfectly legal in one country would be removed from sight because of the laws of another. Not surprisingly, Google sees this as a fundamental challenge to accessibility of information.
As if the ‘right to be forgotten’ was not problematic enough, the EU has recently published its General Data Protection Regulation 2016/679 to be introduced from 2018 which places limits on the use of automated processing for decisions taken concerning individuals and requires explanations to be provided where an adverse effect on an individual can be demonstrated (Goodman and Flaxman 2016). This seems like a good idea on the face of it – shouldn’t a self-driving car be able to explain the circumstances behind a collision? Why wouldn’t we want a computer system to explain its reasoning, whether it concerns access to credit or the acquisition of an insurance policy or the classification of an archaeological object?
My employer has decided to send all those of us involved in recruitment and promotion on Unconscious Bias training, in recognition that unconscious bias may affect our decisions in one way or another. Unconscious bias in our dealings with others may be triggered by both visible and invisible characteristics, including gender, age, skin colour, sexual orientation, (dis)ability, accent, education, class, professional group etc.. That started me thinking – what about unconscious bias in relation to digital archaeology?
‘Unconscious bias’ isn’t a term commonly encountered within archaeology, although Sara Perry and others have written compellingly about online sexism and abuse experienced in academia and archaeology (Perry 2014, Perry et al 2015, for example). ‘Bias’, on the other hand, is rather more frequently referred to, especially in the context of our relationship to data. Most of us are aware, for instance, that as archaeologists we bring a host of preconceptions, assumptions, as well as cultural, gender and other biases to bear on our interpretations, and recognising this, seek means to reduce if not avoid it altogether. Nevertheless, there may still be bias in the sites we select, the data we collect, and the interpretations we place upon them. But what happens when the digital intervenes?
“Nobody lives with conceptual black boxes and the allure of revelation more than the philologist or the scholarly editor. Unless it’s the historian—or the archaeologist—or the interpreter of the aesthetic dimension of arts and letters. Okay, nobody lives with black boxes more than the modern humanities scholar, and not only because of the ever-more-evident algorithmic and proprietary nature of our shared infrastructure for scholarly communication. She lives with black boxes for two further reasons: both because her subjects of inquiry are themselves products of systems obscured by time and loss (opaque or inaccessible, in part or in whole), and because she operates on datasets that, generally, come to her through the multiple, muddy layers of accident, selection, possessiveness, generosity, intellectual honesty, outright deception, and hard-to-parse interoperating subjectivities that we call a library.” (Nowviskie 2015 – her emphases)
Leaving aside the textual emphasis that is frequently the focus of digital humanities, these “multiple, muddy layers” certainly speaks to the archaeologist in me. The idea that digital archaeologists (and archaeologists using digital tools for that matter) work with black boxes has a long history – for instance, the black-boxing of archaeological multivariate quantitative analyses in the 1960s and 1970s was a not uncommon criticism at the time. During the intervening forty-odd years, however, it has become a topic that we rarely discuss. What are the black boxes we use? Where do they appear? Do we recognise them? What is their effect? Nowviskie talks of black boxes in terms of the subjects of enquiry – which as archaeologists we can certainly understand! – and the datasets about them, but, as she recognises, black boxing extends far beyond this.