Where do data scientists come from?

A few years ago, I was preoccupied by this question after an illuminating six months working in the Data Science Lab at Warwick Business School. I co-organised a ground breaking conference in computational social science and it was clear this represented a mode of expertise distinct from what I had previously been familiar with as someone whose academic life had been spent until then in philosophy and sociology departments. It made different claims and responded to different concerns, leading to analyses which were radically extensive but lacking in intensivity. However what interested me more were the human factors associated with it, the distinctive stylings and presentations I could see amongst people who identified as data scientists. It struck me as a fascinating category, doing a great deal of work while remaining under defined. It could be performed in a wide range of ways, while there remained characteristics which most data scientists seemed to share. It placed intellectual issues in the foreground while there nonetheless seemed to be an emerging ethos uniting people who identified as data scientists, even if it might only have been partially articulated.

What did it mean to be a data scientist? How did the identity motivate people? How did these feed back into the occupational category and the role this played within data driven organisations? These are questions which haven’t been on my mind for a while but I was reminded of them today when reading about DJ Patil, the first chief data scientist in the US government. This was the account he gave to Michael Lewis on pg 155 of the Fifth Risk, describing how the category of ‘data scientist’ emerged amidst the data tsunami of the last decade. A range of titles and roles had emerged which packaged up emerging competencies related to data in different ways. According to DJ Patil the impulse to unify these under the master category of ‘data scientist’ was bureaucratic, reflecting a managerial discomfort with an explosion of new categories within their organisations:

Along with much more: in the space of a few years, the interest in data analysis went from curiosity to fad. The fetish for data overran everything from political campaigns to the management of baseball teams. Inside LinkedIn, DJ presided over an explosion of job titles that described similar tasks: analyst, business analyst, data analyst, research sci. The people in human resources complained to him that the company had too many data-related job titles. The company was about to go public, and they wanted to clean up the organization chart. To that end DJ sat down with his counterpart at Facebook, who was dealing with the same problem. What could they call all these data people? “Data scientist,” his Facebook friend suggested. “We weren’t trying to create a new field or anything, just trying to get HR off our backs,” said DJ. He replaced the job titles for some openings with “data scientist.” To his surprise, the number of applicants for the jobs skyrocketed. “Data scientists” were what people wanted to be. In the fall of 2014 someone from the White House called him. Obama was coming to San Francisco and wanted to meet with him.

Exploring the interplay between organisations creating roles for ‘data scientists’ and individuals coming to identify as such would make a fascinating project. How have the role of the “analyst, business analyst, data analyst, research sci” and the people performing them changed in the process? For instance consider the description on Pg 172-174 of the intellectual project which the creation of chief data scientist in the US government sought to advance:

DJ Patil had gone to Washington in 2014 to help people find that gold. He was the human expression of an executive order Obama had signed the year before, insisting that all unclassified government data be made publicly available and that it be machine-readable. DJ assumed he’d need to leave when the man who hired him left office, so that gave him just two years. “We did not have time to collect new data,” he said. “We were just trying to open up what we had.”

How do these organisational projects shape how the data scientists involved see their purpose? How do these projects contribute to a sense of data science as a vocation beyond the organisation in question? How should we revise our conception of organisations if all those beyond a certain size will have data scientists within their ranks? There are many questions which can be asked here and investigating the role itself would be an interesting route into them.

Social listening and reclaiming the future

The data warriors and the electoral wars they wage

One of the most interesting issues raised by the rise of data science in party politics is how to untangle corporate rhetoric from social reality. I have much time for the argument that we risk taking the claims of a company like Cambridge Analytica too seriously, accepting at face value what are simply marketing exercises. But the parallel risk is that we fail to take them seriously enough, dismissing important changes in how elections are fought as marketing hype propounded by digital charlatans.

Perhaps we need to focus more on the data scientists themselves. As much as there is something of the Bond villain about Alexander Nix, CEO of Cambridge Analytica, it’s important that we don’t become preoccupied with corporate leaders. Who are the rank-and-file data scientists working on campaigns? What motivates them? How do they conceive of the work they do? There were interesting hints about this in the recent book Shattered, looking at Hilary Clinton’s failed election campaign. Much as was the case with Jeb Bush’s near entirely stalled campaign, there had been much investment in data analytics, with buy-in right from the top of the campaign. From pg 228-229:

These young data warriors, most of whom had grown up in politics during the Obama era, behaved as though the Democratic Party had come up with an inviolable formula for winning presidential elections. It started with the “blue wall”—eighteen states, plus the District of Columbia, that had voted for the Democratic presidential nominee in every election since 1992. They accounted for 242 of the 270 electoral votes needed to win the presidency. From there, you expanded the playing field of battleground states to provide as many “paths” as possible to get the remaining 28 electoral votes. Adding to their perceived advantage, Democrats believed they’d demonstrated in Obama’s two elections that they were much more sophisticated in bringing data to bear to get their voters to the polls. For all the talk of models and algorithms, the basic thrust of campaign analytics was pretty straightforward when it came to figuring out how to move voters to the polls. The data team would collect as much information as possible about potential voters, including age, race, ethnicity, voting history, and magazine subscriptions, among other things. Each person was given a score, ranging from zero to one hundred, in each of three categories: probability of voting, probability of voting for Hillary, and probability, if they were undecided, that they could be persuaded to vote for her. These scores determined which voters got contacted by the campaign and in which manner—a television spot, an ad on their favorite website, a knock on their door, or a piece of direct mail. “It’s a grayscale,” said a campaign aide familiar with the operation. “You start with the people who are the best targets and go down until you run out of resources.”

Understanding these ‘data warriors’ and the data practices they engage in is crucial to understanding how data science is changing party politics. Perhaps it’s even more important than understanding high profile consultancies and the presentations of their corporate leaders.

Devushkin as the Eviscerated Human

I just came across the following passage in this paper by Anna Mary Cooper and Jenna Condie:

Bakhtin’s (1984a) literary analysis of Dostoevsky’s novel ‘Poor Folk’ shows how the character Devushkin, who in recognising himself in another story, did not wish to be represented as ‘something totally quantified, measured, and defined to the last detail: all of you is here, there is nothing more in you, and nothing more to be said about you’ (p. 58).

http://socialtheoryapplied.com/journal/jast/article/view/31/7

This is exactly what I’m trying to get at with the notion of eviscerating the human: analysing human beings in ways which empty out their thoughts, feelings, dreams and aspirations in order to leave only the transparently measurable aspects that lie beneath their recalcitrant minds. What Mark Andrejevic describes in InfoGlut as corporeal literacy: “the attempt to bypass the vagaries of speech in order to get directly at the true underlying emotions that speakers all too often attempt to mask” (pg. 81). This involves constructing what I’ve come to think of as evisceration devices: tools and techniques, in reality dependent upon conceptual proxies, enabling what is ‘inner’ to either be dispensed with or reduced to a corporeal manifestation of it.

If we can identify this as an intellectual project which wins the commitment of many powerful individuals and groups, we can begin to ask sociological questions about their interests and investments in it. The slightly grandiose and perhaps vague terminology of ‘eviscerating the human’ serves a methodological purpose because this is a project that cuts across multiple social domains and does not involve an overlapping awareness of being involved in a shared project. It is attempt to conceptualise the shared characteristics of something that can be seen across a multiplicity of activities, rather than a single endeavour undertaken by a collective with a shared commitment.

The Evisceration of the Human

In a paper I’m writing for the first volume of the next Centre for Social Ontology project, I’m offering an analysis of what I call the evisceration of the human. I understand this as an intellectual project which seeks to get beyond self-understanding, hollowing out the phenomenological froth which characterises the interpretative human and getting to the underlying behavioural reality beneath it. It’s a project which, as Mark Andrejevic puts it in Infoglut, seeks to “sidestep self-understanding and self-representation to get at these recalcitrant minds directly” (p. 86). Interiority is reduced to empirical proxies, proliferating ad hoc hypothesis which explain away the apparent reality of the first-person perspective and reduce it to measurable and testable behavioural factors.

This operational abstraction constitutes a kind of ‘hollowing out’ of the human, seeking to reduce the category to its underlying behavioural reality rather than trying to cope with it in its bewildering wholeness. The instruments change, the precise formulation of the ad hocery changes but the underlying direction of explanatory travel remains the same. That at least is my hunch, though there’s a huge amount of work in the history of ideas which I’d have to undertake to justify it properly.

My focus however is on the present enthusiasm for eviscerating the human we see associated with digital technology and digital data. I really like this formulation from Audrey Watters, on loc 1245 of her The Monsters of Educational Technology:

We are now creating data at an unprecedented scale, with unprecedented velocity and increasing complexity. The temptation is to believe that if we can just collect all the data from our students – all their clicks –run it through an algorithm, do a little pattern-matching, and we’ll solve everything, we’ll unlock the secrets of the human brain, we’ll unlock the potential of each child.

The allure of our new instruments plays a crucial role in this iteration of the evisceration project. We believe that if only we were to collect enough data, the inner truth would be revealed. The secrets are there, waiting to be revealed, if only we can mine down into the human with enough precision and accuracy.

The ontology of data, the ideology of data

In a post yesterday, I expressed my discomfort with how Nick Srnicek invokes the notion of data as a raw material in his Platform Capitalism. In a footnote on loc 1102-1121, he offers a Marxist justification for this use:

I draw here upon Marx’s definition of raw material: ‘The land (and this, economically speaking, includes water) in the virgin state in which it supplies man with necessaries or the means of subsistence ready to hand, exists independently of him, and is the universal subject of human labour. All those things which labour merely separates from immediate connexion with their environment, are subjects of labour spontaneously provided by Nature. Such are fish which we catch and take from their element, water, timber which we fell in the virgin forest, and ores which we extract from their veins. If, on the other hand, the subject of labour has, so to say, been filtered through previous labour, we call it raw material; such is ore already extracted and ready for washing’ (Marx, 1990: 284–5, emphasis added).

The distinction between natural and processed helps address the charge that data-as-raw-material naturalises data but it doesn’t avoid it all together. It recognises that usable data is always processed and cleaned prior to use, ‘filtered through previous labour’ both directly (wrangling etc) and indirectly (socio-technical systems of generation, capture, storage and analysis).

But does it obscure the character of its applications? The original sources of this ‘raw material’ are intensely social, unlike that extracted from nature, arising from the digitalisation of action and transaction within a market economy. The applications of this ‘raw material’ are intensely variable, with data sometimes being a direct source of value that can be realised through a linear activity, other times being an informational good that can be used to coordinate and calibrate and perhaps more often being a source of confusion and speculation.

It would take constructionism too far to deny the ontology of data. Indeed, I think a concern for the ontology of data is a crucial vector through which we can understand the ideology of data. But we nonetheless need to carefully unpick the way ‘data’ is represented by social actors, the interests served by this and the discursive architecture through which these interests are advanced in different organisational settings.

There’s a fascinating example of this ideology which Srnicek quotes on loc 1296:

If collecting and analysing this raw material is the primary revenue source for these companies and gives them competitive advantages, there is an imperative to collect more and more. As one report notes, echoing colonialist ventures: ‘From a data-production perspective, activities are like lands waiting to be discovered. Whoever gets there first and holds them gets their resources –in this case, their data riches.’ For many of these platforms, the quality of the data is of less interest than their quantity and diversity. Every action performed by a user, no matter how minute, is useful for reconfiguring algorithms and optimising processes. Such is the importance of data that many companies could make all of their software open-source and still maintain their dominant position due to their data. Unsurprisingly, then, these companies have been prolific purchasers and developers of assets that enable them to expand their capacity for gaining information. Mergers relating to big data, for instance, have doubled between 2008 and 2013.

How seriously should we take this? My instinct is to say ‘not very’. The notion that all data is of potential value is licensing massive investment and expansion, entrenching the position of precisely those who are speaking most authoritatively in relation to this transition. To be sceptical of a gold rush doesn’t mean you’re denying the existence of gold, it just means you’re sceptical about the exuberance surrounding the search and how it is being deployed by some to benefit their own self-interest. The distinction between the ontology of data and the ideology of data matters.

Interesting material on data science ethics

Credentials, recruitment and big data

A really interesting section in Ross Perlin’s Intern Nation discusses the role credentials play in recruitment as proxies for sought qualities in potential employees. Once this process of credentialisation-as-filtering is embedded in organisations, surely data science (‘human analytics’) is the logical next step? From loc 2482:

The way job markets actually work, argue some economists, is that graduating from college or completing an internship, regardless of the skills learned, sends a signal to employers, most of whom have little ability to gauge the human capital of a prospective hire. “We have an information problem in the society and the problem of allocating the right people to the right jobs,” writes economist Michael Spence, whose early work on “signaling” has been as influential in its own way as Becker’s theory of human capital. At first, employers may have to guess whether any given signal—a high GPA, for instance—will make for a useful predictor of future success, but Spence posits “a feedback loop” in which employer expectations are revised over time, as the productivity of different types of workers reveals itself.

Data Fetishism

I find this argument from loc 270 of Rob Kitchin’s The Data Revolution extremely compelling. It reminds me of Roy Bhaskar’s argument about the fetishisation of facts from his Reclaiming Reality. This is what Kitchin says:

Moreover, just as we think of bricks and mortar as simple building blocks rather than elements that are made within factories by companies bound within logistical, financial, legal and market concerns, and are distributed, stored and traded, so we largely do with data. Consequently, when data are the focus of enquiry it is usually to consider, in a largely technical sense, how they should be generated and analysed, or how they can be leveraged into insights and value, rather than to consider the nature of data from a more conceptual and philosophical perspective.

This “conceptual and philosophical perspective” is a necessary, though insufficient, condition for a critical one. Only if we consider the social ontology of digital data does it become possible to reflexively incorporate critique of the originating platforms into the digital social science that those platforms facilitate.

Data science meets social sciences: Understanding origins, means and consequences of social computing

I wish I could take part in this – it looks fantastic:

Call for Submissions and Registration

We would like to invite submissions for the 2016 Social Study of IT Open Research Forum (SSIT ORF) Workshop at the London School of Economics on the 19th of May 2016. Aiming at PhD and junior researchers, this year’s workshop will investigate the variety of data centric research in the social sciences.

Arguably, we live in the age of the data revolution. The wave of data is redefining what we study, how we study it and how we theorise. Already acclaimed by some as a paradigmatic shift in science, the data revolution and data science receive considerable attention within the social sciences community. As such, this year’s Social Study of IT Open Research Forum of the Department of Management, LSE, aims at presenting research in progress in this area. If you are researching big data and analytics, this is the place for you to talk about your research, discuss potential stumbling blocks and hear from other participants who may be approaching similar research questions from different theoretical or methodological perspectives.

A significant feature of the Open Research Forum is its informality. While we discourage the use of slides, we will focus on panel presentations with plenty of time to focus on emerging questions and issues. You do not need to bring polished research, but rather to briefly present your topic, research question, theory, methodology and tentative or expected findings and engage in discussion with fellow researchers.

Registrations for the Open Research Forum are currently taking place. If you are interested in presenting, please submit a summary (500-1000 words) of the work or research issues you would like to present and discuss to SSIT-ORF@lse.ac.uk and register here. If you are interested in participating without a presentation, you can register to attend here. Deadline for submissions is the 14th of April 2016. Successful participants will be notified by the 22nd of April 2016.

https://orflse.wordpress.com/

Contempt for the social sciences and the new positivism

I’m a bit preoccupied by this paragraph from Philip Moriarty’s recent post about the invocation of quantum physics in the social sciences. I’d assumed this was true, inferring from things I’d seen others say and things I’d seen them not say, but I’ve never seen it acknowledged so bluntly before:

Social science is important – it provides key insights into human behaviour and addresses questions that are beyond the scope of the physical sciences. I enjoy interacting and collaborating with my colleagues in social science both at Nottingham and elsewhere and gain a great deal from our discussions. But I’ll be brutally honest. I know for a fact that there are many in the “hard” sciences (and elsewhere) who would argue that the funding of social science is a waste of money and that it could be much better spent elsewhere. Misappropriating ideas from quantum mechanics in an attempt to ride on the coat-tails of the (highly successful) intellectual framework underpinning physics does social science no favours at all.

https://muircheart.wordpress.com/2016/02/06/were-flattered-but-enough-of-the-physics-envy-its-embarrassing-us-all/

To what extent should the imperialistic tendencies of contemporary data-driven behavioural science be read as, at least in part, expressions of this attitude? Can you be motivated to use the computational tools of the contemporary natural sciences to ‘fix’ the social sciences unless you quite straight forwardly regard them as broken?

the demand for data science

An interesting article about the increasing demand for data scientists and the centrality which corporate imperatives should have on data science syllabi. How will data science change as it scales? It seems inevitable that undergraduate programmes will play a crucial role in meeting this demand:

o what’s the problem? We have a global talent shortage, and the demand for data scientists continues to grow rapidly, far outpacing the anemic growth in supply. A McKinsey study predicts that by 2018 the number of data science jobs in the United States alone will exceed 490,000, but there will be fewer than 200,000 available data scientists to fill these positions. Globally, demand for data scientists is projected to exceed supply by more than 50 percent by 2018.

The need for experts trained in extracting insights from data is more important than ever.

The global talent drought may be partially caused by the dearth of universities that offer data science programs dedicated to preparing the next generation of data scientists. Fewer than one-third of U.S. News & World Report’s Top 100 Global Universities offer degrees in data science. Of the 29 universities that offer data science programs, a mere six make them available to undergraduates (the rest are reserved for graduate students).

Moreover, the average cohort size for one of these data science programs is just 23 students. The small cohort size may be partially attributed to the fact that most of these data science programs tend to be offered to graduate students, and sizes of graduate programs tend to be smaller. Still, at 23 students per cohort, we are unlikely to make a meaningful dent in closing the global data science talent gap.

An analysis of the curricula for these data science programs also suggests cause for concern. Most are focused almost exclusively on the computer engineering aspects of data science, with course titles such as Software Design, Parallel Computing and Software Development.

Largely missing from the data science curricula at many of these universities are courses in statistical analysis, insights and strategy. This oversight may have serious consequences for graduating students and their future employers.

Without training in these other areas, data scientists may be capable of designing an algorithm that is mathematically elegant, but doesn’t make strategic sense for the business. They also may not know how to design an experiment to determine whether the algorithm is effective. In other words, computer science skills alone are insufficient for success as a data scientist in today’s marketplace.

Furthermore, many data science programs lack courses that help students apply their technical data science skills to fields such as marketing, operations, product development and supply chain, and industries such as energy, bioinformatics, transportation and healthcare.

http://techcrunch.com/2015/12/31/how-to-stem-the-global-shortage-of-data-scientists/

the new frontier of computational politics

A really interesting report on the Guardian about the data scientists working within Ted Cruz’s campaign. As Zeynap Tufekci argues, the potential implications of computational politics are remarkably worrying. We urgently need ethical regulation of data science because otherwise this will prove to be the thin end of the wedge:

Ted Cruz’s presidential campaign is using psychological data based on research spanning tens of millions of Facebook users, harvested largely without their permission, to boost his surging White House run and gain an edge over Donald Trump and other Republican rivals, the Guardian can reveal.

A little-known data company, now embedded within Cruz’s campaign and indirectly financed by his primary billionaire benefactor, paid researchers at Cambridge University to gather detailed psychological profiles about the US electorate using a massive pool of mainly unwitting US Facebook users built with an online survey.

As part of an aggressive new voter-targeting operation, Cambridge Analytica – financially supported by reclusive hedge fund magnate and leading Republican donor Robert Mercer – is now using so-called “psychographic profiles” of US citizens in order to help win Cruz votes, despite earlier concerns and red flags from potential survey-takers.

Documents seen by the Guardian have uncovered longstanding ethical and privacy issues about the way academics hoovered up personal data by accessing a vast set of US Facebook profiles, in order to build sophisticated models of users’ personalities without their knowledge.

It means you have a wild west, where the campaigns can do whatever they want and get away with it

Christopher Soghoian, ACLU

In the race to advance data-driven electioneering strategies pioneered by successive Obama campaigns, Cruz has turned to Cambridge Analytica for its unparalleled offering of psychological data based on a treasure trove of Facebook “likes”, allowing it to match individuals’ traits with existing voter datasets, such as who owned a gun.

http://firstmonday.org/article/view/4901/4097

human centred data science

I’m really interested in this call for human-centred data science and the rationale offered for it:

CALL FOR PAPERS/PARTICIPATION: One-day Workshop on
*Developing a Research Agenda for Human-Centered Data Science*

in conjunction with CSCW 2016
http://cscw.acm.org/2016/

Sunday, February 28th, 2016
San Francisco, CA, USA

Workshop Website:
https://cscw2016hcds.wordpress.com/

———————————-
Important dates:
– EXTENDED DEADLINE: 18th December 2015: Submission of Position Papers
– 18th January 2016: Notification of acceptance
– 25th January 2016: Camera-ready version
– 28th February 2016: Workshop at CSCW 2016
———————————-

TOPIC:
The study and analysis of large and complex data sets offer a wealth of
insights in a variety of applications. Computational approaches provide
researchers access to broad assemblages of data, but the insights extracted
may lack the rich detail that qualitative approaches have brought to the
understanding of sociotechnical phenomena. How do we preserve the richness
associated with traditional qualitative methods while utilizing the power
of large data sets? How do we uncover social nuances or consider ethics and
values in data use?

These and other questions are explored by human-centered data science, an
emerging field at the intersection of human-computer interaction (HCI),
computer-supported cooperative work (CSCW), human computation, and the
statistical and computational techniques of data science. This workshop,
the first of its kind at CSCW, seeks to bring together researchers
interested inhuman-centered approaches to data science to collaborate,
define a research agenda, and form a community.

PARTICIPATION:
This workshop provides a venue for attendees to discuss a variety of topics
in human-centered data science. We welcome researchers interested in
exploring how data-driven and qualitative research can be integrated to
address complex questions in a diverse range of areas, including but not
limited to social computing, urban, health, or crisis informatics,
scientific, business, policy, technical, and other fields. Researchers and
practitioners working with large data sets (“big data”) and/or qualitative
data sets looking to expand their methodological toolbox are invited to
participate and share their experiences while learning from the broader
community.

Topics and themes of interest include, but are not limited to:

– Deep ethnographic methods: How do we preserve the richness of traditional
qualitative techniques in data science?
– Scaling up qualitative data analysis: How do we deal with ever growing
qualitative datasets?
– Quantitative and behavioral methods: How are quantitative and behavioral
methods related to data mining, machine learning, and qualitative methods?
– Connecting across levels of analysis: How can we integrate the analysis
of personal data with large-scale data?
– Ethics and values of data use: What ethical questions should we raise in
using large-scale online data?
– Privacy of data use: How can we preserve anonymity and privacy
within data ecosystems
that can easily expose users?
– Human-centered algorithm design: How do we design machine learning
algorithms tailored for human use and understanding?
– Understanding community data: How can we integrate knowledge gained about
communities from their aggregate social data as well as their personal
experiences?
– Health and well-being at micro and macro scales: What understandings can
be exposed or occluded by aggregate or granular perspectives on health and
well-being?

SUBMISSION:
Please submit a position paper (from 2 to 4 pages in the CSCW ACM sigCHI EA
format – see
http://www.sigchi.org/publications/chipubform/sigchi-extended-abstracts-format-
2016/view) by 18th December 2015 to hcds.cscw2016@gmail.com

The submissions will be reviewed by the organizers with support of other
researchers in a dedicated program committee and selected according to
their potential to contribute to the workshop topic and to foster
discussion.

All accepted contributions (notifications will be sent out by 18 January
2016) will be made available on the website to allow participants to
prepare for the workshop. The organizers may consider the publication of
revised versions of accepted papers as part of a special issue in a CSCW
related journal.

ORGANIZERS:
Cecilia Aragon, University of Washington.
CJ Hutto, Georgia Institute of Technology.
Yun Huang, Syracuse University.
Wanli Xing, University of Missouri.
Gina Neff, University of Washington.
Jinyoung Kim, University of Maryland, College Park.
Andy Echenique, University of California, San Diego and San Diego
Supercomputer Center.
Joseph Bayer, University of Michigan.
Brittany Fiore-Gartland, University of Washington.

CONTACT:
For more details, check the workshop website:
https://cscw2016hcds.wordpress.com/

For any further information on the workshop please contact
hcds.cscw2016@gmail.com

images of the human in digital social science

Notes for a talk I’m doing at this event tomorrow

An obvious place to start is with what I take ‘images of the human’ to be and what I take ‘digital social science’ to be. Images of the human are fairly straight forward. I mean everything from the most explicit claims about human properties and powers to implicit assumptions that go unrecognised by those who can be claimed to hold them.

What do I mean by digital social science? There is digital social science being conducted within the boundaries of the academy, both in terms of an explicit tendency towards a digital focus (digital sociology, digital geography, digital anthropology and the digital humanities) and that which is digital in its methods and/or objects but does not designate itself as such. There are also newer entrants to the intellectual scene, such as the computational social sciences and data science, often though by no means always being lead by physicists and mathematicians seeking to understand the social as a complex system, fundamentally no different to any other found in the natural world. Crucially, much digital social science takes place outside the academy and it is through this that the images of the human explicit or implicit within it can exercise an influence upon the world (and the humans) which they represent.

One of the most provocative offerings comes in Dataclysm, a book by Christian Rudder, a mathematician by training who co-founded the popular online dating site OkCupid. Now with one million to thirty million users, depending on how one defines the level of activity that constitutes a ‘user’, it is the 425th highest ranked site on the internet. The subtitle of Dataclysm is Who We Are (When We Think No One’s Looking) and this captures the impulse of the book: as Rudder puts it, “instead of asking people survey questions or contribing small-scale experiments, which was how social science was often done in the past, I can actually go and look at what actually happens when, say, 100,000 white men and 100,000 black women interact in private” (loc 63). The new technological possibilities offered by transactional data are seen to both call into question ‘who we are’ and offer us ways of looking behind this veil to the underlying reality.

In this sense, I think we misunderstand such an approach if we describe it simply as empiricism. There’s a critical theory here, implicitly claiming to disrupt what Hubert Dreyfus and Charles Taylor call “our ordinary pre-critical view of ourselves and the world”. The impulse here is to reveal the reality of human behaviour at scale: the conditions of both our everyday and scientific knowledge claims are implicitly understood to offer only a limited perspective on what it is humans actually do. Through the data we can reveal what they really do, as opposed to what they tell social surveys and interviewers they do. The reality of behaviour can be read from the numbers: ‘let the data speak for themselves”, at least if we are talking about the right kind of data. I’m at a very early stage of this work and I don’t want to impose a uniformity on the objects of my (meta)-critique, but I’ve chosen Rudder’s example because it’s the most explicit articulation of something which I believe is much more pervasive: an image of the human in which reflection and articulation are seen as mere forth, concealing the reality of what lies beneath.

Mark Andrejevic does a superb job in his book InfoGlut of unpicking the variety of ways in which those beholden to this image struggle with the “recalcitrant mind” which disguises the reality of who we are and why we do what we do. Interiority is rendered as a problem, one struggled with through the identification of ever more diverse proxies and ever more sophisticated measuring instruments, but one which continually returns as a frustrating opacity that eludes conceptualisation. It’s hard to make sense of what is being studied if we deny reflexivity: why did people choose OkCupid other competing sites, why are they using online dating rather than other alternatives, why are some demographics so over-represented amongst those who use online data and others aren’t? Each of these questions can be met by ad-hoc hypotheses, propping up the assumed transparency of the self but doing so provides us with a truncated and flattened social world, devoid of reflexivity. Reasons for action are tacitly acknowledged but representations of those reasons is deemed methodologically illicit: the real things people do (and as a corollary the real reasons) are in fact the focus of inquiry but to seek to represent these leaves us tangled up in the thickets of representation that the project of the transparent self is trying so hard to avoid.

What fascinates me is the reflexivity of those so committed to denying reflexivity. Why do they choose these topics? Why do they write these papers? Why do they talk at these conferences? Their trajectories of social action become inexplicable without reference to precisely the human capacity which their intellectual projects so thoroughly empty out from the social world. We need what Nick Couldry calls a hermeneutic of the anti-hermenutic: what drives this impulse towards the dissolution of reasons and evaluations? How does it differ from and what does it share with similar projects to be found in postmodernism and behavioural science? Or indeed in something like phrenology, as Mark Andrejevic, Dan McQuillan and others have suggested. This investigation will be philosophical, recovering the implicit images of the human and engaging with the explicit ones, but it will also be sociological: looking at the designers, engineers and managers and data scientists and the intra-organisational and inter-organisation contexts within which they are embedded, as well as how digital capitalism as a whole is taking shape through their activity.

These questions matter at a political, as well as philosophical level, because as the legal theorist Frank Pasquale suggests, the causal power of the algorithms that now govern the dynamics of reputation and visibility increasingly lead to calls for us to self-reflect in these terms in order to negotiate a social world governed through them: leading to what Pasquale describes as an “algorithmic self”. But crucially the operation of the algorithms remains opaque, subject to review and revision in a way that is always stubbornly outside our purview. Pasquale cites the example of “the self-promoter whose status updates on Facebook or LinkedIn gradually tip from informative to annoying” or “the search engine−optimizing website whose tactics become a bit too aggressive, thereby causing it to run afoul of Google’s web spam team and consequently sink into obscurity”.

This isn’t a case of people being moulded into a new shape by technology, such that we eventually become what the algorithm assumes that we already are. Reflexivity doesn’t go away. In fact it becomes more imperative than ever, but the risk is that it’s a new form of reflexivity, anticipatory in its focus but orientated towards ubiqutuous algorithmic evaluation rather than our own projects and passions. A communicative reflexivity disciplined by the algorithmic environment rather than the evaluations of trusted interlocutors within a secure micro-context.

This is a peculiar kind of hyper reflexivity which becomes necessary under circumstances that are at least for now isolated within particular tracts of daily life for most (e.g. social media use): a hyper reflexivity which ensues when people are “constantly assessing how each word or deed will affect permanent reputational profiles”. Under such conditions, reflexivity becomes a matter of modulation, adjusting to a context within which the rules are both given and yet fluctuating. Crucially, as Pasquale observes, “we all sense that certain activities win the approval of assorted watchers and others do not”. Measurement is not new by any means but what’s new is its ubiquity, driven it should be noted by near entirely commerical concerns, facilitated by the sheer efficiency with which this is possible when digitally mediated action generates transactional data subject to scruinty.

Pasquale invokes the novel Super Sad True Love Story in which credit scores from 400 to 1600 are displayed publically on ‘credit poles’ at shops and on the streets, ‘personality’ and ‘sexiness’ are automatically ranked within any social space and ‘mood + stress indicators’ provide public heirarchical rankings within any work places. There are similar themes explored in The Circle by Dave Eggars, as well as Whiskey Tango Foxtrot by David Shafer. This increasingly rich vein of fiction, beginning to find parallel representation in television, reflects an incipient unease about the place of measurement within social life. As Pasquale describes Super Sad True Love Story, “In an anomic world where social mores are adrift, the characters in the novel scramble to ‘find their place’ in the social pecking order by desperately comparing themselves with each other”. It seems to me that this could equally be a description of digital capitalism in late 2015. The ‘near future’ world these novels explore seems very near indeed.

The risk here is that the a-reflexive human subject becomes something close to a self-fulfilling prophecy. Reflexivity is not dissolved but truncated, restricted in its scope to orientating oneself within hierarchies that become axiomatic: a metric driven neo-traditionalism in which even the most reflexive projects go no further than creative projects of self-optimization through the quantified self? When considered at the level of law and politics, we can see how reflexivity itself can become an object of suspicion in a way likely to become ever more common.

Predictive analytics function by inferring from past behaviour, licensing interventions in concrete fields (policing, education, health care) based on parameters of ‘normal behaviour’ elaborated in a profoundly opaque way: one’s own past behaviour is the most knowable thing about these emerging standards and retreat into what we know we have done in the past is likely to become a safety zone, sought in order to avoid what Kate Crawford has called ‘predictive privacy harms’. Elsewhere she identifies what might be a forbearer of the coming conservatism in the tendency towards ‘norm core’: a style of dress that celebrates bland ubiquity, an aesthetic affirmation of the value of not standing out in any way whatsoever. Predictive privacy harms are by their nature unpredictable and, which is worse, we’re often unlikely to realise we have been harmed in this way. As Frank Pasquale writes, “we risk freezing into place a future that rigidly reenacts the past, as individuals find that replicating the captured patterns of past behavior is the only safe way to avoid future suspicion, stigma, and disadvantage.” We retain our reflexivity but become something less than human, enthusiastically objectifying ourselves – in a paradoxically reflexive way – while becoming objects to which things happen in wider social life.

Developing a Research Agenda for Human-Centered Data Science

I really wish I could go to this:

CALL FOR PAPERS/PARTICIPATION: One-day Workshop on
*Developing a Research Agenda for Human-Centered Data Science*

in conjunction with CSCW 2016
http://cscw.acm.org/2016/

Sunday, February 28th, 2016
San Francisco, CA, USA

Workshop Website:
https://cscw2016hcds.wordpress.com/

———————————-
Important dates:
– 11th December 2015: Submission of Position Papers
– 18th January 2016: Notification of acceptance
– 25th January 2016: Camera-ready version
– 28th February 2016: Workshop at CSCW 2016
———————————-

TOPIC:
The study and analysis of large and complex data sets offer a wealth of
insights in a variety of applications. Computational approaches provide
researchers access to broad assemblages ofdata, but the insights extracted
may lack the rich detail that qualitative approaches have brought to the
understanding of sociotechnical phenomena. How do we preserve the richness
associated with traditional qualitative methods while utilizing the power
of large data sets? How do we uncover social nuances or consider ethics and
values in data use?

These and other questions are explored by human-centered data science, an
emerging field at the intersection of human-computer interaction (HCI),
computer-supported cooperative work (CSCW), human computation, and the
statistical and computational techniques of data science. This workshop,
the first of its kind at CSCW, seeks to bring together researchers
interested inhuman-centered approaches to data science to collaborate,
define a research agenda, and form a community.

PARTICIPATION:
This workshop provides a venue for attendees to discuss a variety of topics
in human-centereddata science. We welcome researchers interested in
exploring how data-driven and qualitative research can be integrated to
address complex questions in a diverse range of areas, including but not
limited to social computing, urban, health, or crisis informatics,
scientific, business, policy, technical, and other fields. Researchers and
practitioners working with large data sets (“big data”) and/or qualitative
data sets looking to expand their methodological toolbox are invited to
participate and share their experiences while learning from the broader
community.

Topics and themes of interest include, but are not limited to:

– Deep ethnographic methods: How do we preserve the richness of traditional
qualitative techniques in data science?
– Scaling up qualitative data analysis: How do we deal with ever growing
qualitative datasets?
– Quantitative and behavioral methods: How are quantitative and behavioral
methods related todata mining, machine learning, and qualitative methods?
– Connecting across levels of analysis: How can we integrate the analysis
of personal data with large-scale data?
– Ethics and values of data use: What ethical questions should we raise in
using large-scale online data?
– Privacy of data use: How can we preserve anonymity and privacy
within data ecosystems
that can easily expose users?
– Human-centered algorithm design: How do we design machine learning
algorithms tailored forhuman use and understanding?
– Understanding community data: How can we integrate knowledge gained about
communities from their aggregate social data as well as their personal
experiences?
– Health and well-being at micro and macro scales: What understandings can
be exposed or occluded by aggregate or granular perspectives on health and
well-being?

SUBMISSION:
Please submit a position paper (from 2 to 4 pages in the CSCW ACM sigCHI EA
format – see
http://www.sigchi.org/publications/chipubform/sigchi-extended-abstracts-format-2016/view)
by 11th December 2015 to hcds.cscw2016@gmail.com

The submissions will be reviewed by the organizers with support of other
researchers in a dedicated program committee and selected according to
their potential to contribute to the workshop topic and to foster
discussion.

All accepted contributions (notifications will be sent out by 18 January
2016) will be made available on the website to allow participants to
prepare for the workshop. The organizers may consider the publication of
revised versions of accepted papers as part of a special issue in a CSCW
related journal.

ORGANIZERS:
Cecilia Aragon, University of Washington.
CJ Hutto, Georgia Institute of Technology.
Yun Huang, Syracuse University.
Wanli Xing, University of Missouri.
Gina Neff, University of Washington.
Jinyoung Kim, University of Maryland, College Park.
Andy Echenique, University of California, San Diego and San Diego
Supercomputer Center.
Joseph Bayer, University of Michigan.
Brittany Fiore-Gartland, University of Washington.

CONTACT:
For more details, check the workshop website:
https://cscw2016hcds.wordpress.com/

For any further information on the workshop please contact
hcds.cscw2016@gmail.com

big data, post-democracy and boardroom liberalism

Another interesting bit of the Andrejevic book. A lot of these ideas are extremely familiar but he puts them together in extremely creative ways, as well as writing with an astonishing degree of lucidity. From loc 524:

Lurking in these two questions is an assumption about the character of knowledge in the digital era: the notion that the only limit on our predictive power is the ability to effectively organize all the available information. If this were indeed the case, then the development of technological information storage and processing technology might compensate for the shortcomings of the human brain by ushering in new forms of aggregate “knowledge” and predictive power.

And from loc 572:

The apparent obstacle to such a resolution is the limit of human perceptions, analytic ability, and time. The ability to overcome such limits is relegated to the realm of the superhuman. As Laplace, the pioneer of mathematical probability, put it, “Given for one instant an intelligence which could comprehend all the forces by which nature is animated and the respective situation of the beings who compose it – an intelligence sufficiently vast to submit these data to analysis … for it, nothing would be uncertain and the future, as the past, would be present to its eyes.” 8 The name for that intelligence, viewed through one historical lens, would be God. In the digital era, it is the computer and the database.

There’s a profoundly post-democratic slide underway here, building upon longer term trends which have led to a hollowing out of representative democracy. What I find concerning is the frequent assumption, in many cases a self-serving one, that social and political problems will melt away through the sustained accumulation of data. The well-reported wonkishness of Obama goes some way to explaining what has been called his ‘boardroom liberalism’. I suspect his sponsorship of the government data science service, using his power to institutionalise data science throughout the federal government, might very well turn out to be one of the most significant actions of his presidency.

Mark Carrigan

Tag: data science

The data warriors and the electoral wars they wage

Developing a Research Agenda for Human-Centered Data Science