Agents of the General and the Particular

Abstracts

Data reuse and the politics of repair

by Kjersti Aalbu, University of Oslo/DNV; with Susanne Bauer, University of Oslo

This article examines data infrastructures as an ‘agent of the general and the particular’. We specifically analyze data reuse, exploring how collections of data are used to translate from a particular to a different general than what was imagined at the inception of the infrastructure. We analyze two different cases on greenhouse gas emissions from international shipping and global air pollution and health monitoring, focusing on data infrastructures developed by two international organizations—the International Maritime Organization and the World Health Organization. We trace how the data produced by these infrastructures travel through different usages, and specifically explore secondary data use in the finance industry and in the continuously evolving regulations and compliance regimes related to emissions from international shipping and global air pollution and health. We use these two cases to then explore how experiences with data reuse feed into the politics of repair and shape temporalities within and beyond these data infrastructures. How does data collection and reuse shape the existing politics of repair within international organizations? How are data infrastructures reshaped by experiences with primary and secondary data use, and how does this shape the infrastructures’ further function as a translation agent? What are consequences for the role of these data infrastructures in the governance of ship emissions and global health? Drawing on two different cases, this paper is an attempt to pursue questions about processes of translation, generalization, and modification that not only inform regulatory decisions, but also shape what environmental regulation is about.

Making Examples: From Particulars to Generalization in Machine Learning

by Alexander Campolo, Durham University; with Katia Schwerzmann

Machine learning as an ensemble of modeling practices, data infrastructures, and knowledges has a particular orientation toward the general. Its core logic involves an imperative to “generalize,” to produce a model that can apply to data it has never seen before. Our paper examines machine learning's emerging inductive logic from particular instances of data, called “examples,” in large-language models. These generative models take generalization a step further than previous models, drawing on huge text corpuses to produce new texts by predicting sentences. Against naturalistic claims that these capabilities simply emerge as particular examples scale, we investigate a set of three concrete, norm-giving processes that make data exemplary: the engineering of linguistic datasets to optimize for particular types of results, the use of contract labor in the global south to exclude the most harmful examples, and so-called “reinforcement learning with human feedback (RLHF).” We conclude by considering how this emerging logic of exemplarity transforms existing relationships of generals and particulars, drawing on Lorraine Daston’s recent work on rules. The scale of examples in machine learning differs from the abstract generality and systematicity of rational rules. Whereas examples were long thought to be singular, concrete instances capable of pointing toward more general essences, in machine learning, data is made exemplary through norm-giving processes where features can be algorithmically identified at scale. This gives rise to ethical questions regarding a possible transition from highly explicit and calculative rules to a more immanent and implicit form of “exemplary” norms.

Don’t we know enough to act? :
How to tackle challenges of the unidirectional translations in science-society relations

by Selen Eren, University of Groningen; with Anne Beaulieu

Collections of observations and measurements of particular endangered birds and of fields in Friesland (in the Northern Netherlands) are continuously translated by the ecologists with whom we collaborate. These specific instances are turned into survival trends and models that can be cited and used in global academic settings. The general expectation is that these models can be translated by local stakeholders into transformative actions suited to the particularities of Friesland. Frustratingly, neither the expectations of ecologists aiming to produce societally relevant scientific knowledge nor the expectations of the stakeholders motivated to contribute to transformative actions are met.

In the literature on science-society relations, such translations of knowledge about specific places into general models serve to make the issues governable (Miller 2004; Hulme 2010; Turnhout et al. 2016). However, the shortcoming of such global knowledge in supporting local actions on the ground is also problematized, especially due to the detachment of knowledge-making from local meaning-making and acting (Lawrence and Turnhout 2010; Krzywoszynska 2020). A possible alternative is the acknowledgment of knowledge in the plural, and co-existence of different types of partial knowledge, without the need to integrate or reduce them to another (Turnhout et al. 2019).
In this presentation, we will first analyze why certain generalities cannot always be translated back into particularities in our study case. Then, we will discuss what is needed to enable bidirectional translations between particularities and generalities in environmental knowledge production processes, by engaging with the concept of global knowledge and alternative approaches, especially place-based knowledges.

Digital phantom objects as knowledge producers: Known truths, ground truths and medical AI algorithms

by Charlotte Högberg Department of Technology and Society, Faculty of Engineering, Lund University; with Peter Winter, School of Sociology, Politics and International Studies (SPAIS), University of Bristol

Within medical imaging, ‘digital phantoms’ - virtual representations of the human body – are used to either test technological equipment or train humans in diagnostic practice. For example, digital phantoms can be used for simulation studies in assessing the precision of anatomical detection, or for validation studies, in comparing the performance of segmentation algorithms. By and large, however, these efforts focus primarily on ‘synthetic data’ (i.e. an artificial set of data) and insights generated from synthetic data. Such phantoms serve as knowledge producing agents, through simulating and generating data, and as “ground truths” to validate the accuracy of medical AI algorithms.

This paper aims to map how knowledge is produced through and with the use of digital phantom objects. The objects are found in the midst of increased use of data to train algorithms, and demands for synthetic medical data, and validation of AI algorithms’ clinical performance. Arguments in favor of digital phantoms are that they are not constrained by the messy materiality of real physical objects. Further, the digital phantoms are framed as a way to increase control, due to the ability to adapt and know all parameters, thereby working with what the scientists call a “known truth” closer to a real “ground truth”. This also provides the opportunity to create more representative data sets. While acting as generalizations, the digital phantom objects are still argued to be making the presence of particularities possible, by their digital generative capabilities.

Technical review in the upgrade of a particle detector: zooming in and zooming out as evaluative and translative strategies

by Daria Jadreškić, University of Klagenfurt

Technical review is a form of peer review in product development and project management usually encountered in industry and engineering. This contribution focuses on technical review and reviewing as a form of evaluative practice in a high-energy physics collaboration during an upgrade of a particle detector. The physics requirements of a detector upgrade need to be satisfied by the detector design, which brings the relation between the hardware components and the anticipated physics goals to the fore.

This relation is stabilized in a series of technical reviews in which designer teams, reviewers, and management representatives come together to evaluate design specifications, models, and prototypes against a myriad of constraints: epistemic – does it advance the physics? temporal – will it last? spatial – will it fit? material – is it radiation hard? financial – is it affordable? I argue that this is done by interactively and interchangeably zooming in on the particular and out to the general, and I show how these strategies shed light on the relation between evaluation and translation. What is evaluated is not only how parts assemble into whole, but rather how ideas translate into objects, objects into tools, and tools into experiments.

Directions and modes of translation are manifold. For example, standardized (general) electric circuit boards get particularized for the detector environment in which they assemble chains of boards that ‘talk’ to each other in electric impulses. Reconstruction of their energies then ‘talks’ to physicists about the particular particles’ decay channels which get analyzed with the hope of general claims.

Seeing what the forest cannot see: The generalized value of bio

by Oscar Krüger, Lund University; with Alexander Paulsson

Forests cannot think, but many are those who think about forests. After all, forests are increasingly being positioned as a key to biodiversity and as carbon sinks in the fight against climate change. However, industrialized forestry also provides a lucrative product: biomass. Understood by the forest industry as a renewable source, biomass could be used more broadly to replace non-renewable materials, such as plastics.

Building on this insight, we explore the forest politics in the EU through documents and press releases, and how forests have emerged as a matter of concern and a source of conflict in EU's Green New Deal. By using value as a sensitizing concept, we identify two movements in the discourses on value. First, value is derived from exchange within Europe, rather than through trading with the cheaper Brazilian biomass system. Second, value is translated from securing biodiversity to substituting non-renewable materials. The two modes of valuation are connected in so far their common imperative is efficiency, which is underpinning many of the proposals in EU's Green New Deal.

Our conclusion suggests that the value of forests in EU’s Green New Deal should be understood as more than a processes of commodification. Instead, the value of biomass emerges when pitted against biomass exchanged outside EU. We thereby contribute to ongoing debates at the intersections between social studies of forests and the anthropology of value.

Transforming scientific computer models into water management software 

by Catharina Landström; Chalmers University

Water scientists write computer code to test scientific models, i.e., they create executable computer programmes that calculate quantitative mechanisms of processes in nature. By running the computer model with data from a typical locality the scientist will find out if their understanding of the process is correct and, if it is, explore possible scenarios.

It is widely acknowledged that computer code written by scientists is far from what software experts consider to be good code. That does not matter as scientists wite code in order to learn about something else, their object of study (Kelly, 2015). Scientists use computer models, i.e., code, as tools to think with.

However, some scientific computer programmes are developed into software for use by actors in society, for example, water management professionals. Such software is evaluated against different criteria, such as usability. It must also be tested and approved by a wide range of actors in the relevant environmental governance space (Landström, 2023). But which scientific code gets transformed into such professional tolls? Why? How?

This presentation reflects on questions raised by the transformation of some scientific computer codes into software packages used in water management. The reflections are illustrated with examples from scientific and management water modelling in the UK. The aim of the paper is to outline a future research agenda.

Demystifying the Politics of Machine Learning: From Magical Generalities to Mundane Particularities

by Francis Lee, Chalmers University of Technology

Today, machine learners are often proclaimed to be black boxes, whose inscribed logics are impossible to scrutinize and understand. This proclamation and understanding of AI and machine learning runs counter to much of the methodologies that have been developed in Science and Technology studies.
 
The proclamation of the opacity of the black boxed machine learners, and the insistence of STS to open the black boxes of science and technology studies offer a productive ground to think about how we can study machine learners in practice. And especially how we study and analyze the practices and politics of machine learning.
  
Drawing on a deep well of practice oriented studies of science and technology, the paper aims to outline some empirical situations where the politics of machine learning become possible to study in practice, as well as some strategies for analyzing and understanding the practical politics of these situations. It attempts to grapple with the question: How can ethnographers of science and technology study and understand the black boxed politics of machine learning in practice?
 
The broader aim of the paper is to outline some analytical strategies for understanding how the machineries of classification of our time are constructed, and open up a space for studying the politics of classification in machine learning.

Politics of stabilisation and destabilisation

by Allan Lidström Gothenburg University, Department of Theory of Science

STS-analyses of patient organisations and citizen science projects are expressions of the field’s longstanding interest in the relationship between lay-persons, science, and politics.  A core theme in such accounts are explanations of how situated knowledges of concerned groups are mobilised to enact social change, of how individual experiences are transformed into common concerns. In this paper a case study of a disability rights organisation for persons with intellectual disability and a group of Disability Studies scholars is used to deepen the analysis of such movements by introducing the concepts of “politics of stabilisation” and “politics of de-stabilisation”. On the one hand, the study shows that stabilisation of an identity allows for production of generalisable knowledge that can be used to seek political impact within the network of relations the identity is stabilised in. On the other hand, it shows that political impact also can be sought by de-stabilisation of the same identity.  This impact is however of the revolutionary kind as its aim is to contest the network that serves as the foundation for constructing the identity in the first place (thereby also the knowledge produced about it). While these strategies could, on a theoretical level, be understood as mutually exclusive, the different actors employing them are bound together both by their common concern and by the fact that they use the same particularities (experiences of persons with intellectual disability) as their starting point. The difference between them is found in their political ontologies and the respective relations these afford.

The agents and movements of HPV vaccination ‘for everyone’: frictions in the realization of a new HPV vaccine policy

by Lisa Lindén, Division of Science, Technology and Society, Chalmers University of Technology; with Ylva Odenbring, University of Gothenburg

This paper explores the movements between the general and the particular in human papilloma virus (HPV) vaccine policy practice. It builds upon an ethnographic project focused on the introduction of a new HPV policy including all children in the fifth grade of primary schools in Sweden. In the project, we observe HPV vaccine appointments in schools, and we conduct interviews with school nurses, children, parents, and diverse policy actors. This presentation looks at the relation between the policy work of governmental and municipal officials, the advocacy for (further) policy change from LGBTQI rights actors, and the realization of the policy in school health, when children are given the vaccine shot. What particular tools and devices – such as simulation models, research articles, interdisciplinary expert workshops and vaccine shots – are enrolled to assemble, debate and realize the new HPV vaccine policy, and with what consequences? What issues, data and actors get included, marginalized, and/or excluded in the movement from the general HPV vaccine policy to the particular vaccine shot in schools? We take a particular interest in the issue of so-called risk groups (especially, men who have sex with men) and how this specific issue creates action, friction and resistance. This, for example, includes how LGBTQI right advocates enroll and translate particular epidemiological data to assemble evidence for the further protection from HPV-related cancers for men who have sex with men.

Ideas in-between the general and the particular

by Jakob Lundgren; University of Gothenburg

The question of how the particular is translated into generalities is a question that has seen recent attention in STS. However, the opposite phenomenon also deserves theoretical attention. Actors that produce knowledge commonly reference general and abstract ideas as informing their practices – ideas such as objectivity or societal relevance. Meanwhile, empirically oriented work has shown how localized practices embody localized conceptions of general ideas, their meaning attuned to the particular case (Panofsky 2010). However, little theoretical work has been done to understand how actors manage the translation between ideas in their general form and instantiation in local cases. Crucially, the local instantiation must both reflect the needs and interests of the particular situation, yet also be recognized by actors as an instance of the same general idea. A framework is needed to understand how actors make such ideas both local and general.

I suggest revisiting an institutionalist framework to address this question. Proceeding from Frickel’s and Moore’s definition of institutions as “relatively durable sets of practices and ideas” I argue that this perspective lets us understand how actors interpret and reinterpret general ideas in a way that shapes and can reshape practices on local levels. I illustrate with a case of transdisciplinary peer review in two review panels at a Swedish research funder, where the practice of reviewing is shaped by ideas about methods, relevance and epistemic peerage. I show how reviewers are in some cases able to reinterpret and re-institutionalize these general ideas to effect durable changes in practices.

Give them a testbed and they will (try to) raise the municipality

by Doris Lydahl, University of Gothenburg

In this presentation I discuss the translation of social care policy to particular practice(s). However rather than focusing on how individual care workers translate generalities into action (c.f. Lydahl, 2021), this presentation takes as its point of departure in what sociologist would call the meso-level.

I study the rollout of welfare technology in Swedish eldercare. The Swedish government regards welfare technology as a prerequisite for meeting the demands of future eldercare, and hundreds of millions of Swedish Krona have been allocated to stimulate the digital development in eldercare services. While a broader introduction of welfare technology has been slow, countless pilot projects have been carried out throughout the country, and much of the technology are available in pilot tests.

In this presentation, I draw on fieldwork of a regional infrastructure, called a testbed, which aims to coordinate welfare technology pilot projects. The testbed is run by a co-operative organisation uniting 13 Swedish municipalities, and have since its initiation organised several pilot tests, conferences, network meetings and educational activities. All their activities are directed towards other organizations, rather than individuals, and in this sense, they are acting at a meso-level. Building on the classical article “Give me a laboratory and I will raise the world” by Latour (1983) I will explore this infrastructure as a key location in which the activities of the actors connected to it are vital in the translation from the general to the particular.

On Speaking in General Being Particular: Destabilizing the General/Particular Binary By Holding On To The Specificities of Our Ethnographic Cases

by Roos Metselaar, Wageningen University & Research; with René Nissen; Sam van der Lugt

‘General’ and ‘particular’ seem intuitive. But as we go through our fieldnotes, wherever we try to find a ‘general’ to think with, we find only particulars with high aspirations. And so we wonder: might this division - researching things as ‘generals’ drawn from ‘particulars’ - not conceal different relations between specificities? Interlacing ethnographic examples from fieldwork on yeast, resemblance, and soap, we explore what we lose out on if we narrate things as divided between generals and particulars. (1) Ecological certification and ecotoxicology define soap as 'generally' benevolent or harmful. But this 'general' turns out to be a value judgment for some knowledges over others that hides translation errors when household cleaners 'know' their soap in different ways. (2) In the andrology department of a hospital, where staff-members try to ‘match’ sperm donors to intended parents, ‘particular’ individuals are put into ‘general’ - supposedly ‘more objective’ - categories, to make comparison possible. But the distinction between ‘general’ and ‘particular’ dissolves quickly when the two are - literally - placed next to each other on a computer screen, and forced to interact. (3) And “Saccharomyces cerevisiae”, a taxonomic ‘particular’ to the category of “yeast”, is a ‘general’ species term for both “Brewer’s Yeast” and “Baker’s Yeast”. But only for some taxonomists, and certainly not for brewers and bakers, whose differing concerns – taste, health, convenience – bring about different ways of grouping. In all cases, ‘general’ and ‘particular’ are not particularly useful distinctions, covering up the complexities of and relationalities between particular-generals/general-particulars.

Red, Yellow, Green or Grey: Which Colour Does This Pupil’s Knowledge Have?

by Freja Morris, Department of Sociology,  Lund University

This paper aims to make sense of teachers’ parallel document systems for monitoring their pupils’ academic development. Generally, schools require teachers to document their pupils’ knowledge advancement on a digital learning platform. These digital learning platforms are advertised as ideal tools for teachers in order to monitor, document and share information regarding the pupils’ academic development. They are tailored to the learning outcomes and guidelines set by the Swedish National Agency for Education and the Education Act. Materially, they look somewhat different but they all contain color schemes, grid systems and comment sections linked to the learning outcomes and grading system. However, most teachers also develop their own idiosyncratic document system to monitor their pupils’ knowledge advancement. These idiosyncratic document systems can be digital or on paper based on the teacher’s preference. They very often also contain grid systems, spreadsheets, references to the grading scales, systems for commenting and color schemes. On the surface then, although idiosyncratic to the teacher, these document systems do not contain very different elements from the document system on the learning platform. Through fieldwork and interviews, I have explored teachers’ document practices, both idiosyncratic and formal, and I suggest that the parallel document system that most teachers keep is a good site to explore some of the tensions in translating between the particular and the general. The tension that arises when that translation is mediated by standardised tools, such as the digital learning platforms, is of particular interest in this paper.

A particular cheese?

by Kasper Ostrowski, STS center, Aarhus University

A cheese is just one small piece of the world – one lump of microbe-riddled milk curds - but each is an end point of centuries of tradition, collections of instances, rules, laws, regulations, treaties and ongoing metrologic work (Callon et al., 2002: 198-199). This paper aims to unfold a set of mundane cheese names and show how cheese particularities and cheese types might enter the world. Prompted by an extension of European food legislation In 1952 the Danish government introduced a set of new Danish names for renowned foreign cheese types produced in Denmark. The new set aimed to avoid false indications of origin by discarding semantic kinship and affiliations to foreign - now regulated - cheese types and names. The need for establishing Danish cheese particularities grew out of a conglomerate of regulation promulgated through multiple international conventions and treaties (E.g. Paris Convention 1883, Madrid Agreement 1891, Brussels 1900). Departing from the ‘invention’ of Danish cheese types this paper traces how demarcations, delimitations, classifications and bordering schemes within the food industry create particularities. Specifically it investigates how and to what effect juridical infrastructures establish and circulate geographic particularities in order to erase local varieties, establish grouping and transform what used to be dissimilar under the same category (Berg & Timmermans, 2000: 33).)

What is a fuel tank? Law, engineering and the historicity of aircraft

by Will Rollason; Brunel University, London

On July 17th 1996, the B747-100 serving TWA flight 800 exploded and crashed into the Atlantic with the loss of 230 lives. The investigation into the crash showed that the explosion was caused by faulty electronics in the fuel tanks. In this paper, I examine the investigatory and regulatory aftermath of the crash. First, I follow the investigatory report, which demonstrated that regulation of aircraft fuel tanks was importantly to blame for the explosion. Second, I explore the resulting controversy, in which the nature of fuel tanks and the engineering solutions to the risks they pose were systematically called into question. Thirdly, I examine the reconstitution of fuel tanks as stablised technico-legal entities, now equipped with fire-suppressant equipment. This train of events demonstrates, 1) how airline regulation constitutes particular designs, components, and machines as instances of general rules; 2) as they are employed in actual situations, how those generalities are inherently unstable, endlessly creating new particularities, and 3) how new engineering solutions, mandated by new rules are required to domesticate these novel happenings. In the process, law and regulation are constantly called upon to shore up engineering, and engineers are endlessly required to fix the legality of aircraft. Neither law nor technical know-how will work alone as an agent of generalization; rather they constantly alternate between the poles of the general and the particular. This alternation, which is embodied in the materiality of actual machines and their day-to-day operation, constitutes the historicity of aircraft and perhaps other complex, risky systems.)

STS scholarship as an agent of the general and particular

by Morten Sager, Department of philosophy, linguistics and theory of science, University of Gothenburg

The ”evidence movement” consists in the provision of a number of institutional arrangements that serve the purpose of explicitly linking types of particular situations of inquiry and decision-making with specific versions of validity, cumulativity and causality that thence guide professionals’ search for, assessment and use of scientific results.

Often, the role of STS has contributed to and commented on such efforts by descriptively highlighting the innocuous ubiquity of contingency and uncertainty in science. The field has been less prone to lend normative guidance to actors grappling with how to positively define how general epistemic values can be interpreted in particular situations of inquiry and decision-making.

In this sense, there may reside a certain particularist bias in STS. How do we meet actors’ needs to understand science in general, not only as a cornucopia of particular cases? Perhaps Nordic theory of science (vetenskapsteori) can offer a path. This disciplinary tradition has cultivated the ”service mission”, supporting professionals’ need to understand how science is done: How can different versions of validity, cumulativity or causality be established in various disciplines? Just as STS has added empirical details to Nordic theory of science, it may now be time for an increased attention to general epistemological notions in STS. This will benefit professionals that need to understand particular situations in light of general epistemological notions – without resorting to narrow definitions offered through the evidence movement.

STS scholarship can, then, not only study agents of the general and the particular but become such an agent itself.

Making Examples: From Particulars to Generalization in Machine Learning

by Katia Schwerzmann, Ruhr-Universität Bochum; with Alexander Campolo

Machine learning as an ensemble of modeling practices, data infrastructures, and knowledges has a particular orientation toward the general. Its core logic involves an imperative to “generalize,” to produce a model that can apply to data it has never seen before. Our paper examines machine learning's emerging inductive logic from particular instances of data, called “examples,” in large-language models. These generative models take generalization a step further than previous models, drawing on huge text corpuses to produce new texts by predicting sentences. Against naturalistic claims that these capabilities simply emerge as particular examples scale, we investigate a set of three concrete, norm-giving processes that make data exemplary: the engineering of linguistic datasets to optimize for particular types of results, the use of contract labor in the global south to exclude the most harmful examples, and so-called “reinforcement learning with human feedback (RLHF).” We conclude by considering how this emerging logic of exemplarity transforms existing relationships of generals and particulars, drawing on Lorraine Daston’s recent work on rules. The scale of examples in machine learning differs from the abstract generality and systematicity of rational rules. Whereas examples were long thought to be singular, concrete instances capable of pointing toward more general essences, in machine learning, data is made exemplary through norm-giving processes where features can be algorithmically identified at scale. This gives rise to ethical questions regarding a possible transition from highly explicit and calculative rules to a more immanent and implicit form of “exemplary” norms.

The supervising veterinarian as a boundary-spanning agent? Examining animal protection cases in Finnish supreme administrative court

by Jaakko Taipale, University of Helsinki; with Tomi Lehtimäki

Animal protection cases at the Finnish supreme administrative court typically feature ex post facto challenges to supervising veterinarians’ decision making and their powers to inspect suspected  animal welfare violations. The court’s examination of the characteristics and boundaries of the vets’ professional practice in a particular case allow us to analyse one type of professional expertise in an entanglement of scientific and legal institutional rationales and authority. While veterinarians are required to mobilize their scientific expertise in their inspections of local conditions of animal welfare, they also need to take account of legally defined competence, economic aspects and interests as well as individual rights and personal relations.

In this presentation we describe our court case type and data, and reflect on the methodological aspects of using legal court decisions for making observations about agency. How do the court decisions speak about and relate the different domains of knowledge/practice and experience? Veterinarians can be seen as stradling or inhabiting several domains of knowledge/practice, especially with regard to their dual role as experts in veterinary medicine and as law enforcement officials. How are the veterinarians’ decisions contested in the selected court cases by either drawing on the particular or the general, and with what consequences? We present empirical themes that exemplify translations across the domain boundaries and across the particular/general divide.

Can someone translate? How did experts exert agency in the making of the Water Framework Directive

by Nin Valin, Vrije Universiteit Amsterdam (NEWAVE); with Dave Huitema

The sociology of translation has known more than one ‘’turn’’. In its most recent cultural and power trends, the role of translating agents and their agency on translation have been put forward. However, political dimensions, domination and power relations embedded in translation processes are not yet fully understood. In the study of environmental policy change in particular, the power of expert practices, the politics of their success or failures in advising and proposing policy solutions, cannot be ignored. We argue that both sociology of translation and studies of policy change need to be equipped with an analytical framework that takes into account political interests and institutional constraints. We suggest that the concept of transcoding could offer such tools. Transcoding describes the processes and arrangements that bring scientific knowledge to become ‘’governable’’, actionable policy measures. In other terms, translating agents bridge general scientific statements with specific policy proposals, while overcoming competing interests. This paper empirically implements this concept in the negotiations context of the European Water Framework Directive, and expands its analytical traction by proposing four fields where translating actors have agency: ambiguity, complexity, implementability and credibility. The paper zooms in on the critical phase of development of two key policy principles embedded in the Directive: the cost-recovery principle and the river basin scale approach. Our analysis sheds light on the combined agency of policymakers and experts who took up the task of pushing through these principles and ‘’transcoded’’ their proposed ideas into applicable policy decisions.

Organizers

Lisa Linden and Francis Lee, Chalmers University of Technology; David Ribes, University of Washington