"Data and the Future of Critical Social Research" (ICA 2017 Pre-Conference)
Organisers:Nick Couldry (London School of Economics), Lina Dencik (Cardiff University), Andreas Hepp (University of Bremen), Karin van Es (Data School Utrecht) with support of Pat Aufderheide (American University)
Division/Interest Group Affiliation(s): Communication and Technology Division and Philosophy Theory and Critique Division
The growth of automated data collection, processing and analysis, and its installation within contemporary social, economic and political orders has created a number of huge challenges: for protecting fundamental rights and values such as freedom and autonomy, for understanding the connections between communications and social order, for sustaining key institutional processes such as the law, and for maintaining the very legitimacy and authority of decision-making by legal, political and social institutions. At the core of these challenges is a more basic question: what happens to society when communications (in the novel form of datafication and, underlying that, automated symbolic categorizations within database structures) begin to play a historically new role in the organization of life? This question requires communication researchers to be in dialogue with researchers in law and policy to address this fundamental question of communication beyond borders. Specific topics to be addressed include: 1. What ethical, legal and normative concepts are most helpful in building appropriate regulatory frameworks that manage the consequences of datafication? 2. How can we best theorise how economic legal and social institutions are being transformed by datafication? 3. What forms of social order and social governance are emerging through datafication and algorithmic processes, and are they consistent with existing democratic models of social governance? What if they are not consistent? 4. What are the specific implications of datafication processes for the authority of legal institutions and processes of legal decision-making? 5. What can be learned, and what, if anything, must be unlearned from the European GDPR and EU legal proceedings against the abuse of market power by large technology companies? 6. What distinctive perspectives does the Global South on data and datafication and how can those perspectives be effectively integrated into discussions in the Global North? 7. Generally, what are the common agendas and common questions that need to be formulated so that scholars in the fields of communications and law (and, more broadly, management and society) can come together around transdisciplinary solutions to the problematic implications of the datafied society? 8. What practical resources and pathways are needed to help the voices of critical researchers in this area be heard better and more widely? Committed speakers so far are: Keynotes: Julie Cohen, Georgetown School of Law, author of Configuring the Networked Self and (forthcoming) Between Truth and Power; Mark Andrejevic, Monash University, Melbourne and author of Infoglut (2013) and iSpy (2007).Panel speakers: [communications] Payal Arora, Erasmus, Rotterdam; Alison Hearn, Western Ontario; Anna Lauren Hoffman, U of Washington; Thomas Poell, U of Amsterdam; Joseph Turow, UPenn[law] Ellen Goodman, Rutgers; Natali Helberger, U of Amsterdam; Frank Pasquale, U of Maryland; Andrew Selbst, Data and Society, New York; Wolfgang Schulz, Hans-Bredow-Institut, Hamburg
Keynote 1: Mark Andrejevic (Monash University, Melbourne) / Chair: Lina Dencik
PANEL 1: Autonomy and voice in the datafied society / Chair: Andreas Hepp
Joseph Turow (University of Pennsylvania): Voice Marketing and Habituation to the Biometric Age
Usha Raman (University of Hyderabad, India): Does data obscure presence? Considering patient autonomy and ethical practice in the networked clinic
Ellen Goodman (Rutgers University, New Jersey): Cognitive Friction by Regulation
PANEL 2: The normative bases of data justice / Chair: Karin van Es
Solon Barocas (Cornell University): The Problem with Bias: Allocative versus Representational Harms in Machine Learning
Anna Lauren Hoffmann (University of Washington): Data, Violence, and the Limits of Antidiscrimination Discourse
Wolfgang Schulz (Hans-Bredow-Institut, Hamburg)
Keynote 2: Julie Cohen (Georgetown School of Law) / Chair: Nick Couldry
PANEL 3: The Moral order of Datafied Publics / Chair: Patricia Aufderheide
Payal Arora (Erasmus, Rotterdam): Benign dataveillance as the new democratic order?
Thomas Poell (University of Amsterdam): Governing Platforms
Frank Pasquale (University of Maryland): Preserving Well-Ordered Societies: Toward a Thick Theory of Media Regulation
PANEL 4: The datafication of modernity’s institutions / Chair: Andrew IIiadis
Natali Helberger (University of Amsterdam): The hyperresponsive press: how the trend towards newspersonalisation changes the media, law and policy
Alison Hearn (Western Ontario): Counting Heads: Data and privacy in the outsourced ‘academy’
JOSEPH TUROW: Voice Marketing and Habituation to the Biometric Age
Voice is a “gateway drug” to the social acceptance of prejudicial biometrics, and commerce is the gateway territory. Shopping and marketing, the most widely shared public spheres, are entering a biometric age powered by artificial-intelligence, and the analysis of individuals’ voices to exploit their identities, emotions, and even physical well-being mark its leading edge. This mining of voices through “smart speakers,” car information systems, customer service calls, and other tools signals a new era of discrimination where analysts can use actual bodily processes as data for commerce, politics, and governance. The hyper-personalized understanding of individuals provide them with differential opportunities based on profiles driven by machine learning and predictive analytics. At a time when the voice-based world of biometrics is still being built, we need to open it to the values of cultural democracy before socially corrosive processes linked to it become too entrenched to change.
USHA RAMAN: Does data obscure presence? Considering patient autonomy and ethical practice in the networked clinic
In a recent article in the New Yorker, Atul Gawande cited a 2016 study that found that physicians spend more time in the examination room looking at their computer screens than at the patient. Increasingly, patient data, drawn from a variety of sources and integrated temporally and spatially, becomes an important—and undeniably useful—mediator between the clinician and patient. But this layer can also obscure and potentially diminish the value of embodied presence and the subjective cues that the patient may communicate and the doctor may discern. What happens to the subjective narration of experience—and its reception--in such a clinical encounter? What new inequalities might emerge in the datafied clinic—already a space of unequal power relations? How might we imagine and articulate patient rights and clinical responsibility in such a space? This paper explores the implications of data-framed clinical communication, particularly in relation to the integrity and control of patient information, identity, and self-presentation.
ELLEN GOODMAN: Cognitive Friction by Regulation
Government regulation of media has traditionally imposed source disclosure requirements in order to increase autonomy. These “right to know” regimes are particularly common in the contexts of advertising and political speech. Self-regulatory norms around disclosing conflicts of interest, opinion, and authorship do the same kind of work. These source disclosure regimes operate by increasing the “cognitive friction” of communications and attempt to balance the free speech benefits and costs of a more frictional communications environment.
Digital information platforms have been largely free from source-disclosure requirements and have also tended to strip out or diminish source-signifiers where they exist. In doing so, the information platforms share the design feature of digital platforms generally, which is to promote frictionless exchange. In the case of information, more friction may be required to enhance autonomy. I will examine various disclosure approaches being proposed or adopted in democratic states from the perspective of re-introducing cognitive friction.
- Apply existing media law source-disclosure requirements to digital platforms
- Adopt new forms of source-disclosure for new kinds of automated speech
- Adopt new forms of audience-disclosure to address “reach opacity” borne of audience fragmentation/microtargeting
These interventions could help to re-introduce cognitive friction and encourage norm-development around disclosure. Taken too far, they have the risk of increasing friction to levels that impinge on the free exchange of communication.
SOLON BAROCAS: The Problem with Bias: Allocative versus Representational Harms in Machine Learning
Computer scientists have recently begun to explore the many ways that machine learning can reproduce and reinforce troubling histories of discrimination in the kinds of high-stakes decision-making that automation had promised to purge of bias. Machine learning runs the very serious risk of inheriting human biases because it is fundamentally a process of learning from recorded examples of past human behavior. And the examples that “train” machine learning models might encode past discrimination and injustice. Computer scientists have thus begun to worry about the harms that machine learning might visit on already disadvantaged and marginalized populations, particularly in consequential decision-making like hiring, lending, and housing, among others. Now a community has emerged expressly concerned with questions of fairness in machine learning, with the goal of integrating such concerns directly into the process of training models and automating decision-making.
In this paper, we examine how work on fairness in machine learning has imagined the problem of bias and why researchers have struggled to fully account for the harms associated with such bias. In particular, we show that bias has been understood overwhelmingly as a problem of unfair allocation: systematically denying desirable opportunities to deserving members of legally protected classes (e.g., women, racial minorities, etc.). Understanding fairness in these terms has limited the field’s ability to recognize and address harms other than those involving economic opportunities withheld. We survey a number of examples where the harms of bias in machine learning are representational in nature—where applications of machine learning propagate derogatory messages about specific social groups, reinforce cultural stereotypes, fail to recognize members of marginalized and minority communities, or under- or over-represent them in depictions of valorized or censured activities.
Using a series of real-world examples of machine vision and natural language processing, we assess how bias in machine learning can have the effect of privileging certain identities and denigrating others—and we contrast these with the more common problems of allocation addressed by the existing literature. While allocative harms spring from discrete moments of decision-making (e.g., whether or not someone gets a job), representational harms tend to be much more diffuse (e.g., cultivating general beliefs about the people most suited to certain careers). And while the long term effect of these kinds of representational harms might be uneven access to societal resources, we argue that representational harms need to be addressed in and of themselves. To that end, we consider a number of approaches from the critical humanities that can be used to identify, interpret, and respond to different types of representational harms.
ANNA LAUREN HOFFMANN: Data, Violence, and the Limits of Antidiscrimination Discourse
Problems of bias and fairness speak directly to the threat that “big data” and algorithmic decision-making stand to worsen already unjust distributions of rights, opportunities, and wealth. In the United States, grappling with these issues has found clearest expression through discourses of rights, due process, and antidiscrimination. The current focus on these discourses, however, overlooks issues in how ideals of antidiscrimination have been interpreted and applied, resulting in a tendency for work on data and discrimination to place “beyond” consideration important structural and social concerns integral to the realization of data justice. Ultimately, a failure to establish critical distance from antidiscrimination discourse’s most problematic tendencies risks reproducing—rather than upending—hierarchical systems of violence and exploitation at the root of social injustice.
Normative structures can be described as being produced in figurations of actors operating within a thematic frame. The field of internet governance is of great interest for scholars as right now we can observe severe transformations of the respective governance structure (e.g. growing importance of private ordering and of governance by “code”). Therefore internet governance can serve a test field for theories on the production of normative structures.
Against this background, the followings questions arise:
1. What is the explanatory value of the communicative aspect (in a narrow sense, not symbolically generalized communication of social systems as defined by Luhmann) of relations in those figurations in contrast to other non-communicative aspects?
2. To what extend can transformation of the normative structures been explained by the mediatisation of communicative figurations producing those structures?
3. Can the normative reading of technological structures (“code”) be integrated a concept of figurations producing governance structures?
PAYAL ARORA: Benign dataveillance as the new democratic order?
By looking at democracy’s three core dimensions - the socio-political, the economic and the legal, and how big data intersects with them, this study analyzes the following questions in the context of India and China: does digitization of citizen data reduce or entrench inequality in representation and participation? Are these computing systems enabling a more level playing field by helping citizens leapfrog their economic status and circumvent traditional intermediaries of power? To what degree does this datafication ensure citizen’s rights? This paper reveals a complex narrative that goes beyond the Western conceptualization of democracy, and provokes a re-examining of this concept as governance becomes increasingly impacted by the automating of decision-making. By juxtaposing these emerging databased systems against the normative Western model, this study reveals the crossroads we are at when it comes to what constitutes as social inclusion, free market participation, and citizen rights in this digital and global age.
THOMAS POELL: Governing Platforms
Digital platforms enable user-driven forms of organization and collective action (Benkler 2006; Bennett and Segerberg 2013, Shirky 2008). Yet, platform-based activity is simultaneously centrally monitored and shaped through ubiquitous techno-commercial infrastructures (Couldry 2015; Fuchs 2017; van Dijck 2013). As platforms penetrate every sphere of life, this combination of distributed user activity and top-down techno-commercial steering undermines public institutions and destabilizes social relations, enhancing the precarity of labor, unsettling urban communities and disrupting democratic public debate (van Dijck, Poell & de Waal 2018). In the light of these problems, this paper considers how the platformization of society can be governed in correspondence with vital public values. It argues that due to the nature of platform-based activity, effective governing arrangements need to be organized through a framework of ‘cooperative responsibility’ (Helberger, Pierson & Poell 2018). However, a major obstacle in developing such arrangements are the progressively entangled economic interests of the involved actors. In the name of optimization and cutting back public expenditure, governments actively contribute to platformization by deregulating markets and privatizing public infrastructures, while citizens increasingly dependent on asset-based welfare schemes revolving around platforms. Hence, future governing arrangements will need to be based on a new political pact informed by key public values and geared towards reducing dependence on corporate platforms.
FRANK PASQUALE: Preserving Well-Ordered Societies: Toward a Thick Theory of Media Regulation
In past work, I have tried to balance the the type of intermediary responsibility I call for in "The Automated Public Sphere," with my characterization of large internet platforms as common carriers in "Platform Neutrality" and "Internet Nondiscrimination Principles." Critics of thick theories of media regulation have characterized responsibility and neutrality as irreconcilable goals. However, there is another way of looking at the problem, which sees neutrality as most positively meaningful in the context of a robust conception of the good embedded in an egalitarian, pluralist, and social democratic order.
I will first describe the documented, negative effects of online propagandists’ interventions (and platforms’ neglect) in both electoral politics and the broader public sphere. I will then discuss legal tactics to mitigate platforms’ power, or to encourage or require them to exercise it responsibly. As I conceded in "The Automated Public Sphere," some regimes are already too authoritarian or unreliable to be trusted with extensive powers of regulation over media (whether old or new), or intermediaries. However, the inadvisability of extensive media regulation in disordered societies only makes this agenda more urgent in well-ordered societies, lest predictable pathologies of the automated public sphere degrade their processes of democratic will formation.
This paper will lay a foundation for a broader account of what a government and society well-ordered enough to pursue aggressive, substantive media regulation would look like. Harking back to the Rawls-Habermas debates of the 1990s, the idea here is to ground democratic order in a robust theory of the purposes (and limits) of politics. I will attempt to extend Danielle Allen's theory of civic education from educational institutions to the public sphere as a whole, while reflecting on Fred Turner's fascinating account of the Committee for National Morale and other Popular Front models. An entrenchment of such ideals may seem anti-democratic, but is far preferable to the alternative: an increasingly Schmittian public sphere, made all the more dangerous by the behavioristic tactics of psychological manipulation enabled by contemporary tactics of personalized propaganda.
NATALI HELBERGER: The hyperresponsive press: how the trend towards newspersonalisation changes the media, law and policy
In my contribution, I will first trace the origins of the trend towards providing users with more personalised services that are responsive to users' indivdual interests and preferences (which is arguably at least as old as the coin-operated Telemeter payTV services in the sixties), and how datafication is a logical continuation, but also disruption of this trend. I will then reflect on the genesis of media law and policy in the light of the increasing responsiveness of the media, and dare some predictions about the future of media law and policy in the age of the hyperresponsive press.
ALISON HEARN: Counting Heads: Data and privacy in the outsourced ‘academy’
This paper will examine the ways the current logics of platform capitalism and its obsession with marketing and the accumulation of data have infiltrated and re-oriented higher education in North America and Europe. As universities face demands to reduce spending, increase enrolments and rationalize the learning process, they are increasingly turning to privately owned online program managers (OPMs), such as Pearson, Wiley, Blackboard, to help them push their services and operations online. These OPMs are now directly involved in curriculum design, course delivery, student recruitment, counseling and retention, and, of course, marketing. OPMs collect their payment in the form of a revenue sharing model expressed as a percentage of tuition fees, a standard licensing fee, or through a pay per use structure. Another, far less publicized source of revenue for these companies comes in the form of the data they collect from students as a result of their recruitment and course delivery work. In OPM discourse, students are figured as “leads”, and information about them is the most valuable asset these companies generate. Depending on the agreement, the pool of data collected on ‘leads’ can then be used to enhance the OPM’s marketing efforts on behalf of other universities. Under these conditions, students are reduced to target markets, their privacy inconsistently protected at best (Mattes, 2016).
Even as the boundaries between private and public collapse behind the scenes on university campuses, students and faculty generally remain unaware of the degree to which their education and work have been outsourced to big tech companies. But, there can be no doubt that the for-profit motives of the OPMs are pushing the values of perpetual student growth and increasing tuition ahead of educational quality. As a result, issues concerning university governance, labour relations, and government regulation are thrown into crisis as well. To whom are universities responsible? Is education just another way to generate data for private companies? Who are students and colleagues working for? Who benefits when higher education itself is ‘platformized’?