How much does poetry cost? What is the worth of language in a digital age? Is quality measured on literary value or exchange value, the beauty of hand-crafted, hard-wrung words, or how many click-throughs those (key)words can attract and how much money they earn the company who sells them? But haven’t words always been sold? As soon as they became written down, moveable and transferable words entered the market place, and then necessarily the political sphere. But these words gained an exchange value as integral parts of a text – a story, a poem, a book, for example. Removing or reordering these individual words – or ranking them based on external influences would change the meaning and devalue the text as a whole, in both a literary and monetary way. Can language retain its integrity once it becomes part of the digital economy? Is there even such a thing as the ‘integrity of language’?  Certainly the words Google auctions off have referential values unanchored to narrative context, and it is this new context and the politics surrounding it that I am attempting to examine and expose in my new project which I have called {poem}.py.

The project started out when I was required to provide a poster for the Information Security Group (ISG) Open Day at Royal Holloway later this month, which always make me nervous as – unlike most of my PhD contemporaries in the Cyber Security CDT – I don’t have a load of mathematical formulas, graphs and data to fill out the required poster template. So as I was thinking about how to represent and explain my PhD topic to an audience of cryptographers, mathematicians and computer scientists, I decided to see how much my favourite poem ‘cost’ if I put all the words through the Google AdWords Keyword Planner and outputted the results on a mock-up of a receipt – which I thought might look nice on a poster. In this way I discovered that, at 4:39 PM on 7th May 2016, my favourite poem At the Bomb Testing Site by William Stafford cost the princely sum of £45.88 (before tax).

To explain the logic behind this – the keyword planner is the free tool Google AdWords provides advertisers so they can plan their budgets and decide how much to bid for a particular keyword or key phrase to use in their advert. Google gives a ‘suggested bid’ price for each word, giving an advertiser some idea how much they will have to spend to win the mini auction which is triggered each time someone searches for that keyword. When an advertiser wins the auction, their advert will appear as a ‘paid’ (as opposed to organic) search result right at the top (and now right the bottom too) of the rankings with the small yellow ‘Ad’ box next to them. The advertiser then pays the winning bid (which, like eBay, will be one penny/cent above the second highest bid) each time someone clicks on their advert. Phrases such as ‘cheap laptop’ or ‘car insurance’ can cost as much as £50 per click. This is the basis of how Google makes its money, a form of ‘linguistic capitalism’ (Kaplan: 2014) or ‘semantic capitalism’ (Feuz, Fuller & Stalder: 2011) in which the contextual or linguistic value of language is negated at the expense of its exchange value.

One of the first problems I encountered with this method was that once I had fed the words of a poem through the keyword planner I then had to put them back into their narrative order to make the receipt ‘readable’ as a downward list, as Google churns the words back out according to their frequency of search rather than in their original order. With my test poem, I had to order the words back into the shape of the poem manually, which was time-consuming and fiddly. I have since been working with CDT colleagues Ben Curtis and Giovanni Cherubin using Python code to automate this  process. This union of poetry and code is where the project title {poem}.py comes from – .py being the file extension for Python.

Once I had a spreadsheet with the poem back in narrative list order, and with the corresponding ‘price’ of each word – including duplicates – I added up the total cost of the poem and then created a template which mirrored a paper receipt.

This first attempt revealed several really interesting points which not only illustrate what I am trying to examine and expose in my thesis, but it also gave me ideas about how I could use the project as a quantitative method of gathering data and also as a creative practice and artistic intervention.  A section of my thesis examines how the decontextualisation of words in the searchable database leads to anomalies in search results and autopredictions which not only reflect, but also therefore perpetuate stereotypical, sexualised, racist or sexist search results. The words of the poem on the receipt have likewise been taken out of context, and are instead imagined as how well they will do in the context of an advert. Their repeated use by advertisers and confirmatory clicks by users will also presumably increase their frequency within the wider database.

“the cost of a word to Google relates to the size and wealth of the industry it plays a part in advertising”

Once I had run a few more poems through this process I started to realise that words relating to health, technology, litigation and finance were particularly expensive. In At the Bomb Testing Site, I was initially puzzled as to why the word ‘it’ costs £1.96, which seemed disproportionate compared to other words. I then realised that, to Google, the word is ‘IT’ (as in information technology) – hence its price.

In Wordsworth’s Daffodils, the words ‘cloud’, ‘crowd’ and ‘host’ are expensive not because of their poetic merit or aesthetic imaginings, but because of ‘cloud computing’, ‘crowd-sourcing/funding’ and ‘website hosting’. Wilfred Owen’s Dulce et Decorum Est revealed that medical words such as ‘cancer’ ‘fatigue’ and ‘deaf’ had relatively high suggested bid prices, while ‘economical’, ‘accident’, ‘broken’, in Anne Carson’s Essay on What I Think About Most are all over £5.00 per click, and the suggested bid for the word ‘claim’ is £18.10. Perhaps unsurprisingly,  it seems the cost of a word to Google relates to the size and wealth of the industry it plays a part in advertising.

But as well as pricing individual word and phrases, Google’s Keyword Planner also tries to second-guess what you are trying to advertise by the words you enter. In the case of At the Bomb Testing Site, the Keyword Planner thought I was either trying to advertise road biking (presumably the words curve, road, panting, tense, elbows, hands and desert suggested this), or some kind of life coaching, career management service which was prompted by the phrase in the poem ‘ready for change’. Put a question mark on the end of that phrase and it becomes a highly profitable key-phrase in an advert. Similarly the high price of the word ‘o’er’ in Daffodils is explained in the context of OER (Open Educational Resources). The AdWords planner also suggested I might be trying to market a product relating to Game of Thrones due to the Rains of Castamere song in which ‘the rains weep o’er his hall’.

As I played around with the receipt template, I added a CRC32 checksum hash value to the receipt as an ‘authorisation code’. A checksum is a mathematical blueprint of a piece of text which is generated to ensure that a transmitted text is not altered. The sender sends the checksum with the text and the recipient generates the same checksum to make sure it has stayed the same in transit. Using this as an authorisation code on the poem receipt is therefore indicating that when protected by code or encrypted, the poem retains its integrity, but when it is decoded, it is then subject to the laws of the market – as shown on the receipt itself. I also added N/A to the tax field as a little dig at Google’s tax situation in the UK.

But the more poems and texts I analysed in this way, I began to suspect that there is something interesting to be learnt from understanding the geographical, political and cultural logics which might dictate the economic forces which apparently mediate and control this linguistic market place. I ran words such as ‘trump’, ‘war’ and ‘blair’ through the keyword planner over a period of two weeks and noticed how the suggested bid prices fluctuated, despite them not being what you might assume to be very ‘marketable’ words. The keyword planner also allows the user to target their campaign by location, so I could then measure the ‘value’ of war, for example, in the US and in the UK, and even down to tiny areas such as Egham, and I could record these values over a period of time to see how key national and international events might influence word prices.

As well as recording the fluctuations of specific words and names, I am keen to capture the changing uses and values of groups of words based loosely around a theme, and have decided that continuing to use poetry is the best (and most apt) way to do this. So I have selected a series of poems which are somewhat tangentially linked to events which are happening in the UK and world over the next few months such as the Olympics, the EU Referendum, the release of the Chilcot report and the US Presidential election. Gathering this data over the next few months will enable me to conduct a quantitative longitudinal study into the geopolitical and cultural influences which shape linguistic capitalism, and therefore potentially also the composition and weighting of the wider linguistic discourse.

But apart from the quantitative side to this project (which can be harvested in data spreadsheet form), I want to use the output of the receipt as an artistic intervention or critique to make the issues and politics around linguistic capitalism and the way Google treats language more visible and accessible. If there is a politics lurking within the algorithmic hierarchies and logic of the search engine industry (which I believe there is), then it is a politics hidden by the sheer ubiquity and in some way the aesthetics of the Google empire. My thesis is based loosely around Walter Benjamin’s Work of Art in the Age of Mechanical Reproduction essay, and as such, views the various ways in which Google controls, manipulates and exploits data (linguistic and otherwise) under the guise of ‘free’ tools and accessories as a kind of aestheticisation of politics. Following Benjamin, therefore, the final chapter of my thesis will examine ways of turning this power back around, and ‘making art political’, or more specifically to this project, reclaiming language as art.

I hope to be able to speak to and engage with various academics and artists who have attempted ‘Google Poetics’, Adwords ‘happenings’ (Christophe Bruno: 2002) or creative resistance (Mahnke & Uprichard: 2014), (Cabell & Huff: 2010), (Feuz, Fuller & Stalder: 2011) to explore the difficulties and successes of working within or without of the Google framework to produce interventions. It is in this chapter that I also want to use {poem}.py as my own artistic intervention and act of political art. I am aware that I am in effect mixing quantitative data gathering with qualitative methods and creative practise here, which is something I need to think through.

As I mentioned in my previous post, last week I co-organised a workshop on Living with Algorithms which aimed to let participants be creative and provocative in thinking about everyday life and algorithms. For my own contribution to the workshop I asked participants to send me their favourite poems in advance. I then bought a second-hand receipt printer and set about monetising their poems so I was able to print them off for them ‘live’ during my presentation at the workshop. At some stage I would like to use this group of poems to form the basis of an actual art exhibition, but this method has also proved really helpful in terms of beginning to answer some of the questions I asked at the beginning of this post. Because I didn’t tell the participants why I wanted them to send me a poem, some of them were only available in formats which unintentionally resisted the process of commodification so I had no option but to print out VOID receipts for two of them.  The first was an amazing spoken word poem called Bog Eye Man, by Jemima Foxtrot which is only accessible on YouTube or Vimeo. As the actual text of the poem does not appear on the web, I was unable to ‘scrape’ it. The other poem was contained within a Jpeg file from which I could not copy and paste. These two examples show how we might begin to envisage a way to maintain the integrity of poetry in a digital age dominated by linguistic or semantic capitalism, the example of the spoken word poem in particular harks back to Benjamin’s description of the loss of aura when a work of art becomes ‘reproducible’. For the time being, Bog Eye Man remains resolutely unmonetised (at least until spoken data starts being algorithmically scraped anyway…), and retains – as Benjamin wrote ‘its presence in time and space’.

But back to the poster, where all this started. This isn’t the one I’ll be presenting at the ISG open day – it doesn’t conform to the strict template and colour scheme – but is one I made for the Humanities and Arts Research Centre poster competition, which is a bit more aesthetically pleasing…