The Silent Revolution

How Digitalization Transforms Knowledge, Work, Journalism and Politics without Making Too Much Noise

M. Bunz

• Publisher: Palgrave Pivot
• Year: 2013
• Print ISBN: 9781137373496
• eBook ISBN: 9781137373502
• Topic: Languages & Linguistics
• Subtopic: Journalism
• Index: Languages & Linguistics

Critically engaging, illustrative and with numerous examples, The Silent Revolution delivers a philosophically informed introduction to current debates on digital technology and calls for a more active role of humans towards technology.

1

When Algorithms Learned How to Write

Abstract: This chapter contributes to discussions in sociology of media that critically debate the interaction of algorithms with our knowledge and brain. Following Nicholas Carr (2010) and Katherine Hayles (2012) who have discussed the effect of information overload on reading, this chapter looks into algorithms that started to write using the example of a sports report writing algorithm ‘Stats Monkey’. To add to the debate, the chapter delivers an overview of actual definitions of algorithms in the humanities, including Media Theory, Software Studies, and Digital Humanities. Then it traces today’s debate back to the problem of Artificial Intelligence and the misconception of a ‘Ghost in the Machine’ (Ryle 1949). Having refuted the ghost, technology emerges as a tool, which nothing but shifts the parameter of knowledge.

Keywords: algorithms, artificial intelligence, digital humanities, digital knowledge, fact, truth

Mercedes Bunz. The Silent Revolution: How Digitalization Transforms Knowledge, Work, Journalism and Politics Without Making Too Much Noise. Basingstoke: Palgrave Macmillan, 2014. DOI: 10.1057/9781137373502.

Digitalization changes how we know, and we do not know enough about it. That algorithms have learnt how to write, for example, wasn’t expected. Nobody anticipated it, and the report of a college baseball game wasn’t anything to grab anybody’s attention with anyhow. ‘The robots are coming. Oh, they are here!’ wrote David Carr of The New York Times on the issue, half a year after the algorithm Stats Monkey had covered a Northwestern Wildcat baseball game in Illinois. Here is the thing with innovations: When it is plain as the nose on your face that from now on we need to see the world with different eyes, everybody is in shock. But before innovations become annoyingly obtrusive, they are often already among us, and pass by unnoticed. In this way, groundbreaking inventions have taken place in back rooms such as garden sheds or US garages. There, the first Apple and Microsoft inventions have been built, as these have been the places where man and machine have combined forces. With algorithms it is different. When in March 1998 the search engine Google, for example, was tested at google.stanford.edu, it was a research project. Algorithms don’t need a place, but come into being as a rough idea or university project to be immediately launched and tested online. In a way, they don’t hide in back rooms, but in front of our eyes, and only when they change the world do they become apparent.

We have known for quite a while that algorithms do an outstanding job with calculations. We also got used to the fact that on the internet, they look up what we want to know. Finally, they also started to write, a moment that the French philosopher Jacques Derrida would have not let gone by unnoticed. But are they really writing? Let’s have a look into the procedure in order to understand what its social effect could potentially be.

The sports report writing algorithm Stats Monkey combines for his texts two techniques: first, it looks up the scores that are published online; second it investigates the most important players and the course of the game by making use of an algorithmic decision tree; and third it puts the outcome into a journalistic text using pre-written templates such as ‘Team X takes an early lead and never looks back’ or ‘Team Y tries to rally late, but doesn’t make it’. One click, and a bone dry, yet informative sport report is ready, faster than any journalist could even type a sentence. It reads as follows:

SOUTH BEND, Ind. – Tony Bucciferro put the Michigan State Spartans on his back Sunday and spurred them to a 3–0 win over the Notre Dame Fighting Irish (7–11) at Frank Eck Stadium. Bucciferro kept the Fighting Irish off the board during his nine innings of work for Michigan State (12–4). He struck out five and allowed one walk and three hits. Senior Matt Grosso was not able to take advantage of a big opportunity for the Irish in the ninth inning (Bunz 2010, further texts also cited by Carr 2009).

While we may criticize the text’s prosaic quality we can’t argue away that a cultural technique is automated: story telling. Up till then it had been exclusively a human skill. Now the algorithm Stats Monkey marks another historical moment of a digitalization that transforms our world as fundamentally as once industrialization did. For computers, search engines, and with them algorithms have already stolen into our daily lives, and the latter can be compared to the inventions that changed sluggish cotton manufacturing into a rapid textile industry at the beginning of the industrial revolution. Of course, this also came as a surprise: ‘Capitalism arrived unannounced,’ as the political economist Karl Polanyi puts it (Polanyi 1944, 89). In 1733 the Englishman John Kay patented his invention of the ‘Flying Shuttle’, a device that allowed the weft to be passed through the threads much faster, with ‘a speed that cannot be imagined’ as a contemporary noted (Wadsworth and de Lacy Mann 1931, 470). The device doubled the productivity of the weavers. Their increasing demand for yarn was answered in 1764 with the invention of the multi-spool machine, the ‘Spinning Jenny’. The machine enabled one worker to work eight spools or more simultaneously. From England, the industrial revolution was on its way. Can the program Stats Monkey be understood as the revenant of the Spinning Jenny? The Stats Monkey program makes apparent that digitalization has taken on a dramatic scale. Its influence in our lives started with the World Wide Web at the European Organization for Nuclear Research, CERN, where in 1990 the first prototype of the information managing system for sharing knowledge has been tested. While the machines of the industrial revolution have automated human work, the algorithms of the digital revolution assist human knowledge. However, in the age of skilled work this affects our jobs as the American economists Erik Brynjolfsson and Andrew McAfee have pointed out in their awakening book Race against the Machine (2011). And other than in the industrial revolution, this time the automation won’t hit the low-pay sector. Today, armies of expensive lawyers can be replaced by cheaper software, when algorithms help analyse 1.5 million documents for less than $100,000 (Markoff 2011). Some of the well-paid bond traders and bankers of the financial service such as Morgan Stanley will soon be exchanged with computers. Its head of interesting rates at this time, Glenn Hadden, is reported to tell colleagues that ‘the trading floor of the future will surround a few traders with the hum of powerful machines’ (Lucchetti and Brett 2012). Also the dream job of the middle class, journalism, is affected with the first sports report writing programs developed and working, a case which demonstrates the current dilemma.

Invented by Professor Kristian Hammond’s students at the Northwestern University in Illinois, the Stats Monkey is the fruit of a collaboration of the Department of Computer Science and its Medill School of Journalism, and was developed as an answer to media diversification and the financial crisis that have both made traditional journalism struggle (Intelligent Information Laboratory 2010). As print media had to follow their readers onto the internet, they had to find new revenue streams. Local coverage is especially troubled financially, after classifieds had left traditional media to find a new home online. Here is where the students wanted to help. Automatically generated sports reports could allow local media to expand coverage and offer more content to their readers, which would enable them to increase the advertising – more articles equal evermore possibilities to place ads. Additionally, it could also free reporters from the chore of writing reviews on all the games in the lower leagues instead of allowing them to focus on more sophisticated analysis and features.

However, much like their predecessors, the weavers, the journalists didn’t feel relieved by the automation of parts of their work. Two hundred and seventy-seven years later they too didn’t have the impression that the algorithms would disburden them, but were afraid they would replace them. Unlike the weavers they didn’t destroy the machines. Instead, the rise of algorithmic help was answered by a wave of indignation. Journalists from Russia to India, from the UK to the US, from Belgium to Italy wrote on the fact that algorithms had learned how to write. Several editorial desks flirted with the end of human journalism altogether. The American magazine Business Week worried: ‘Are sportswriters really necessary?’ La Stampa in Italy described journalists ‘besieged’ by intelligent software. And the Parisian Le Monde claimed: ‘The era of robot-journalism has begun.’ It has.

But what is happening here to journalists will soon happen to everyone: digitalization. Journalism is simply the first profession to experience a change that is much more profound. It will shake up our expertise in general. Far beyond sports reports, algorithms can gather information available in data sets or online. Faster than any human, it can restructure the information in a chart, or can even transform its findings into a textual overview. As knowledge has also freed itself from being only available at a specific point, a book in the library or the computer at home, to be available online, a lot of research, reports, and evaluations could be delegated to algorithms. Bankers, lawyers, journalists, all skilled jobs will be affected. Nearly everywhere an employee keeps track of a development; algorithms can draw up an overview instead. It is of no wonder that the university project of a sports report writing algorithm has quickly turned into a start up called ‘Narrative Science’. Unsurprisingly it also formed a technology development agreement with the investment firm In-Q-Tel, which supports the CIA and the US Intelligence Community; in the era of digitalization, they surely need help to make sense of today’s mountains of collected data.

After disrupting our distribution channels, digitalization reaches out for our production, but this time not just the creative industries will be affected. The disruption caused by finding and creating stories from data will add to the one that shook up traditional ways of distributing music, film, books, or television, but it will quickly reach beyond the creative industries. Business reports, health records, all kinds of summaries can be automated. The technology has the potential to disrupt everything that implies coherent information, and this means: our expertise. Consequently, the social impact of digitalization will be similarly profound to the impact of industrialization, for at present we are a society of experts.

Sociologist Anthony Giddens made the observation some years ago that expertise has become a ‘pervasive phenomenon’ for now ‘an expert is any individual who can successfully lay claim to either specific skills or types of knowledge which the layperson does not possess’ (Giddens 1994, 84). More recent studies agree that specialization by field has become the dominant paradigm in education (Amirault and Branson 2006). We are all trained to become experts. Will the effects of digitalization on the middle class be similar to the effects of industrialization on the working class? No matter if you are a lawyer or an accountant, a doctor, a teacher, an engineer, a politician, even a chef de cuisine, an author, a car mechanic, a manager, or a micro-chip designer, parts of your skills will soon be taken over by digitalization – and the next chapter will discuss what precisely is happening to experts in great detail. But to start with we should first capture the digital force.

What are algorithms? How are they defined? And what can they know? The next part will provide a brief overview of academic approaches towards algorithms. Then we will make a short detour back in history to reveal how human knowledge could become a problem in the face of the capabilities of machines. Finally, we will examine what today’s algorithms do in order to ‘know’ – a moment we as experts should better be aware of.

Hidden relationship issues

The text writing algorithms are an indicator that the automation of information has reached its next step. Clearly, the project does mark a turning point, albeit not a beginning – what digitalization might mean for the human race still looms. There will always be more data, different algorithms, new devices, groundbreaking developments, and a next version. In a connected world, things are as challenging as they are complicated. Nonetheless it is high time to start asking questions. As the German philosopher Martin Heidegger once put it, questioning builds a way. This way is needed in order to not simply watch digitalization, but also direct it. So what are algorithms?

In her study of algorithms in architectural and interaction design, the philosopher Luciana Parisi answers this question with a surprising statement. It turns their traditional definition upside down, which goes as follows: algorithms are step-by-step procedures for calculations that consist of instructions and follow a finite set of rules to carry out a computation – a definition dating back to the Persian mathematician Muhammad al’Khwarizmi (pronounced in Latin slang ‘algorism’), who had lived c.780–850. But now Parisi writes: ‘algorithms are no longer seen as a tool to accomplish a task’ (Parisi 2013, XIII ). Instead of simply performing rules, she claims that they ‘have become performing entities: actualities that select, evaluate, transform, and produce data’ (ibid., IX).

Parisi’s new approach to algorithms comes at the dawn of a new era. For sure, their computation still consists in sequences of commands that instruct a machine or result (Cormen 2013). For sure, algorithms are still expressed in programming languages such as Java, C++, Python, or Fortran. For sure, they still rely on protocols to exchange communication with other software, devices, or internet nodes. However, in data processing there is an obvious trend to what is generally referred to as ‘big data’, that is larger data sets, whose technical history have been well described by Kevin Driscoll (2012). The rise of data, which is technically driven, changes the notion of algorithms profoundly: experts agree that without data to process, the algorithm remains inert (Berry 2011, 33; Cheney-Lippold 2011; Manovich 2013). The effectiveness of algorithms is strongly related to the data sets they compute, and this is even resulting in a dispute: computer scientists (Domingos 2012) as well as businessmen (Croll and Yoskovitz 2013) ponder if more data beat better algorithms, or if it is the other way round.

The humanities take on algorithms has surely been influenced by the data-driven shift. Ever since the British computer scientist Alan Turing (1936) wrote about the idea of a ‘Universal Machine’, algorithms have fascinated scientists and thinkers alike. His theoretical device, the ‘Universal Turing Machine’, manipulates symbols, and thereby simulates the logic of a computer algorithm. The young student at King’s College Cambridge found that if it is possible to give a mathematical description of the structure of a machine, every machine can be simulated by manipulating symbols. When running a software that consists of algorithms, his ‘universal’ principle comes to life. And as this life is generally described as ‘virtual’, the status of the algorithm is a complicated and highly interesting issue for philosophers.

The term ‘algorithm’ belongs to one language family with ‘code’, the language of our time, and one could say: code is the language in which an algorithm is written. Both describe the same ‘thing’ from a different perspective: the word ‘algorithm’ – a set of rules to be followed by calculations – marks a mathematical perspective, while the word ‘code’ – a system of words to represent others – takes a linguistic perspective. Similarly to its name, its ontological status – what is its being? – has been approached from various sides:

As inexistent: The German media theorist Friedrich Kittler famously claims ‘There is no software’ (1995). In his description, algorithms are a sheer effect of the hardware they rely on, designed to disguise our technical hardware determination. While this is a radical approach of some beauty, it also offers various problems: from a philosophical perspective, it repeats the gesture of idealistic philosophy to seek truth behind a curtain; from a humanistic perspective, it operates along the lines of technological determinism; from a pragmatic perspective, it lacks an approach to study the evolving field of software further. Against Kittler, Andrew Goffey (2008) has demonstrated the many ways in which an algorithm executes ‘control’.

As an activity: Kittler scholar Wolfgang Ernst focuses on the fact that an algorithm stores information in a different way than writing: breaking it down in 0 and 1, it doesn’t narrate but counts. Therefore, archives in the age of online digital collections become a ‘mathematically defined space’ (cf. Ernst 2012; Parikka 2011). Shintaro Miyazaki (2012) also emphasizes the activity, but insists that in a strict sense it is not even mathematical. In his view, an algorithm formulated in a programming language is not the same as an algebraic formula: it is not ‘recursive’. Alexander Galloway stresses the specific ontological quality of algorithms and code as executable: ‘code is the summation of language plus an executable metalayer that encapsulates that language’ (Galloway 2004, 165). While the ontological status of an algorithm as process has been acknowledged, the approach has also been questioned. Against the tendency to treat source code as an origin from which algorithmic actions emerge, Chun (2008 and 2013) makes the point that interfaces are more than the effect of their source.

As an interaction: Software studies look further into the algorithmic activity in a broader sense and seek to overcome the ‘immateriality’ of software (Fuller 2008, 4). Aspects of design, glitches, interactions, or preferences come into focus (Fuller 2003), as well as the social condition of the algorithm’s production (Berry 2011, 43–51). Software can only be unde...

Table of contents

• Cover
• Title
• 1  When Algorithms Learned How to Write
• 2  How the Automation of Knowledge Changes Skilled Work
• 3  The Second Nature
• 4  On the Production of Crowds
• 5  The Digital Public
• 6  The Silent Revolution
• References
• Index