The work in The Underlying co-opts the financiers tool of live sentiment analysis of online news production and social media, relating to BPA’s (Bisphenol A*), to consider how surveillance, rather than a rogue element of capitalism, enmeshes with the effects of market forces upon the environment, happening at a molecular level.
Lag Lag Lag (8 screen video interface, with live sentiment/emotion analysis re mentions of BPA on twitter and online news)
8 screens hang from the ceiling held in a black metal frame, reminiscent of financiers monitors, showing videos syncing across all the screens (looped, approx. 5 mins), with live sentiment and emotion analysis of the pollutant BPA – Bisphenol A (a synthetic oestrogen flooding the planets water supplies) - applied to live twitter feeds and news updates.
The interface, blurs distinctions between what are pre-recorded video works and live data analysis, as videos interweave with live ‘sentiment’ and ‘emotion’ analysis of the pollutant Bisphenol A on twitter and online news updates.
The video works enmesh human cognitive as well as non-cognitive processes, blurring human/animal in-distinctions with soft computing, the molecular structure of Bisphenol A, and live data production, engendering the potential as well as the dangers of multiple cross-species hybridities, from a critical posthuman position, with an emphasis on any subject to speak of (as questions of authorship, also arise) emerging in synthesis with their environment.
I was interested to see if there was a way to turn the financier’s tool of ‘sentiment’ and ‘emotion’ analysis, on its head, so to speak, and map something that might be of interest to lay folk, instead, with regards the environmental challenges ahead.
In an analysis of the 2013 flash crash (‘Breaking News – Flash Crash’ Ami Clarke (2014)) Karppi and Crawford1 drew attention to the Dataminr software that mines Twitter’s ‘firehose’ to produce a sophisticated scoring of the relationships between words in play, to uncover grades of expressed ‘emotions’ as well as “importance and social meaning - in order to ‘predict the present’ and thus transform social media signals into economic information and value”. Here, value is accrued through an opaque, but meaningful process of assessment, in that the analysis in no doubt fuels decisions made by financial operators happening at speeds of data processing far beyond any human capacity. “A phase transition in cultural research, social scientists now analyze patterns in the massive datasets used to study emotional sentiments on Twitter, to deconstruct narrative tropes in the media.1 to allegedly identify anger, fear, disgust and sadness. ‘Emotion detection’ has grown from a research project to a $20bn industry.”
I consulted an ex-derivatives trader, and worked with a programmer to devise a speculative pricing model that takes the sentiment/emotion analysis regarding mentions of BPA’s on twitter, and online news sources, to map the rise and fall of reputation, in real time, of the top 100 polluting companies in the world. It utilises a financial quantitative model that takes BPA sentiment analysis, the FTSE (as a proxy for the general stock market), and Weather Futures contracts from the Chicago Mercantile Exchange (that map the weather at London Heathrow), as well as local pollution data taken from the longitude and latitude of the gallery. A list to the side of the graph shows the most polluting companies with the price next to them generated via the pricing model.
The sentiment analysis software Vader operates from local raspberry pi’s, whilst the emotional analysis is done by IBM Watson online, that both then inform other aspects of the graphs. The Google map pulls on regular twitter API, showing sentiment analysis of tweets mentioning BPA, whilst the fluctuating bars show emotional analysis of the twitter feeds as they scroll up the bottom two monitors. The scrolling news feed is informed by sentiment analysis on the two monitors to the right, showing a spectrum between 1 and minus 1, whilst the candlestick graph shows the opening and closing senti value, and the range that these reach per hour.
The sentiment/emotion analysis also informs what’s happening in the VR work Derivative, and the amount of airborne particles that are produced within the environment.
The Underlying - installation shots - close-up details of graphs, twitter feed / news analysis / pricing model, around 08.14
The work considers the multiple ways that form and medium, as well as the content of the information we receive, influences reception of what is being transmitted, often in combinations of text and image as looping viral feedback systems within an economy of attention.
The text in the video works emerges from a script edited over many years: Error-Correction: an introduction to future diagrams. It acts as a script reflecting on the influence of calculus, in which each articulation is just one of many takes, constantly re-edited, that references and includes openly borrowed texts from contemporary commentary, news items, anecdotal evidence; culminating in an interrelated convergence of many interwoven threads, whereby the voice (through language) is constituted between someone else's thoughts and the page. It openly acknowledges the multitude of influences on the choices we make. For the first time, the artists personal history interweaves with theoretical musings from writers and thinkers such as Katherine Hayles, Octavia Butler, Sylvia Wynter, Paul B Preciado, with online twitter production/news updates, and live data analysis.
The live sentiment and emotion analysis occurs throughout the duration of the work, shown live on-screen often interwoven with other screens showing video works, with a rolling news feed, and a series of graphs that gradually combine to make up an entire screen reminiscent of the financier style display that includes:
- data visualisations - from top right - clockwise:
live news feed updates (scrolling up) with live sentiment analysis of BPA mentions in online news production - shown on a spectrum between minus 1 and 1
live twitter feed (scrolling up) with live sentiment analysis shown on a spectrum between minus 1 and 1, and emotion analysis showing joy, anger, disgust, fear, sadness - via emoji’s, with live rolling news feed
fluctuating bars showing emotion analysis of the twitter feeds: joy, anger, disgust, sadness, fear
google map showing geo-location of tweets - mediated by english language usage, and the tweet being tagged with it’s location
the candles display an hourly value rate that shows the highest and lowest values via the wick/stick, whilst the body of the candle shows the opening and closing sentiment value for the tweets.