Publication by DTH-Lab & Data Solidarity Project team University of Vienna: Data solidarity operationalising public value through a digital tool

Introduction

Researchers and others are responding to the increased use of artificial intelligence (AI) in more and more areas of public and personal life (see e.g. Berryhill et al., Citation2019; Farah, Citation2023) with criticism. Sharon and Gellert (Citation2023) criticise the ability of big tech companies to access and use public data to develop proprietary products and services without providing fair compensation to the public sector. In health and healthcare, for instance, large datasets held by government actors (e.g. NHS in the United Kingdom) have been shared with private companies to build large models to increase the speed and accuracy of disease diagnosis (see Powles & Hodson, Citation2017) leading to criticisms around the conclusion of the data-sharing agreement with insufficient transparency, a lack of safeguards to ensure private companies do not use such data to improve other unrelated products and no consideration of whether this project would create public value (see definition below; see also Kickbusch et al., Citation2021; Shilton et al., Citation2021). Researchers also criticise aspects that extend beyond data-sharing. Academics and civil society actors have noted that the development of large language models, often trained using large amounts of scraped data from the internet, primarily benefits US – and UK English speakers and generally persons inhabiting the Global North (see Nicholas & Bhatia, Citation2023). Underlying both problems are questions related to data governance. These questions include: Whose data is shared with whom? For which purpose and under which condition is this data being shared? Who do benefits accrue to and who is bearing the costs?

As these questions gain importance, this article critically examines the underlying data governance regimes enabling and regulating data practices. We understand data governance in the broad sense of the word as the laws, policies, and informal rules that shape how data are created, used, and protected. We begin by exploring the core logic of existing regimes and highlighting their shortcomings, particularly in generating equitable outcomes. To address these limitations, we introduce the concept of data solidarity, a framework proposed by Prainsack et al. (Citation2022). Data solidarity revolves around public value as its guiding principle. Discussing how public value can be conceptualised and operationalised, we present PLUTO (Public VaLUe Assessment TOol), a tool designed to measure the public value of specific instances of data use. Overall, this article provides a critical analysis of how current data governance regimes perpetuate injustices and inequalities while offering data solidarity as a practical and equitable alternative. By outlining the development process of PLUTO, this paper aims to invite critical scrutiny, encourage collaborative refinement, and enable others to adapt or replicate the tool for diverse contexts and data governance challenges. Our focus, here, lies on data governance’s evaluative dimension: assessing the public value of specific data uses by weighing benefits and risks, and ensuring equitable outcomes. This means considering not only individual rights but broader societal impacts, asking: Does this specific data use generate public value? If so, how do we support it; if not, how do we mitigate harms and redistribute benefits?

Rather than giving an ultimate and definitive answer of what public value is and how different aspects of public value should be weighed, the PLUTO tool is meant to represent a starting point for a wide-ranging discussion about these questions, and give organisations (companies, public bodies, etc.) a prototype that they can adjust and adapt to meet their own specific needs in terms of the questions they focus on and the weights that are given to different responses.

Background: Data governance regimes and their shortcomings

As ever more aspects of people’s lives, environments, and bodies become datafied, how we govern data has become a pressing concern. Categorising data based on data types (personal v. non-personal, human v. environmental, etc.) has been the fundamental logic underpinning data governance in both the United States (US) and the European Union (EU) (see Fazlioglu, Citation2019 for discussion). This model derives from the belief that data type directly correlates with the potential risks posed by its usage. For instance, personal or sensitive information, such as religious beliefs or political affiliations, is deemed to carry particularly high risks if exploited by malicious actors (see e.g. Beldad et al., Citation2011).

Fazlioglu (Citation2019) points to a number of primary challenges that the existing categorization-based data governance model encounters:

1. Emergence of New Data Types: The proliferation of digital data collection has led to the generation of novel data types that do not fit neatly within predefined categories. Consequently, the current framework struggles to address data with evolving attributes (see also Einav & Levin, Citation2014).

2. Triangulation and Context Sensitivity: Convergence of data sources and advanced analytics techniques often transform seemingly innocuous data into sensitive information capable of identifying and potentially harming individuals. The boundary between sensitive and non-sensitive data has become increasingly tenuous (see e.g. Office of the President, Citation2015).

To address these shortcomings, scholars advocate for a paradigm shift towards context-centric data governance (see also Gratton, Citation2013; Nissenbaum, Citation2010; Winter & Davidson, Citation2019). This approach acknowledges that each data point, regardless of its category, can potentially contribute to identifying an individual when combined with other publicly available information (Fazlioglu, Citation2019). Fazlioglu suggests focusing on assessing the likelihood of data linkage, the potential disclosure of sensitive information to unauthorised parties, and the context of data usage when determining necessary legal safeguards. Beyond that, some of us have previously argued that a data-types-based approach is blind to the risks that emerge from the linkage of different (types of) datasets that, cumulatively, determine the risks to people. Finally, the data-types-based approach cannot take into account whether the use of such data generates public value (Prainsack et al., Citation2022).

As discussed above, both the EU and the US data governance regimes focus on the distinction of data by type, but there is a key difference between these regimes. A few exceptions aside, the EU’s General Data Protection Regulation (GDPR) applies universally throughout all sectors, meaning that whether or not a point of data is viewed as personal data does not depend on the context in which it is being used (Solove & Schwartz, Citation2011). Meanwhile, the US follows a strategy where sectoral laws regulate the use of personal data in ‘specific industries or specific contexts’ (Schwartz & Solove, Citation2014, p. 881; see also Flaherty, Citation1992). Thus, while the US approach at least acknowledges the problem of precisely delineating the sensitivity of information, it still does not focus on data use and the risks and benefits associated with specific instances of data use.

The concept of data solidarity emerges as a response to the challenges posed by conventional categories and approaches to data governance. Drawing inspiration from context-centric principles, data solidarity emphasises evaluating data use rather than its category (Gratton, Citation2013). A context-centric perspective categorises data uses as positive, negative, or neutral based on their impact on data subjects (Gratton, Citation2013). Data solidarity goes beyond Gratton’s model to propose that the evaluation of data use should play a more important role in data governance. European policymakers have started acknowledging the significance of contextual analysis in data governance (Article Citation29 Data Protection Working Party, Citation2014). They recognise the need to consider the context and purpose of data processing to effectively address sensitivity and emerging technological trends (Article Citation29 Data Protection Working Party, Citation2014).

Specifically, the proposed data solidarity perspective emphasises evaluating data uses based on their potential and likelihood to generate public value, which we operationalise as the ‘net result’ of weighing risks and benefits to people and communities (we will say more about this in the next section). We now turn to explaining data solidarity in detail.

How data solidarity foregrounds data use over data type

The way in which data solidarity focuses on data use instead of data type becomes most apparent when we examine the three fundamental pillars that form the core of data solidarity (see Table 1). Pillar I states that data uses that create significant public value should be actively facilitated. Pillar II mandates that harms must be proactively prevented wherever possible – such as by prohibiting data uses that are unlikely to yield benefits for people while entailing significant risks. Where prevention is not feasible or not successful, harms must be effectively mitigated. Pillar III argues that where corporate data use has no discernible public value but is conducive to profit generation, the profiteer should be required to share the benefits (including financial profits) with the public.

In the following sections, we will explore each of these pillars and the concept of public value in more detail, on Pillars I and III. In addition, we will highlight the central role that public value plays within these pillars.

The content of PLUTO

PLUTO tool comprises 25 questions organised into four categories. The four categories are:

1. Information about the Data User: This section investigates the identity, motivations, and intentions of the data user to ensure that users who tend to use data in ways that benefit public value (e.g. a climate NGO) are treated differently from actors that do not (e.g. a large advertising company).

2. Benefits of Data Use: Within this category, we explore the positive impacts that data can be used for and who these benefits manifest for within and across societies.

3. Risks of Data Use: PLUTO takes a detailed look at potential challenges and pitfalls associated with data usage, and who these risks are highest for within and across society.

4. Institutional Safeguards: We also examine the measures data users put in place to minimise risks, maximise benefits, monitor and react to potential emergent harms.

For an in-depth exploration of each category and access to the complete set of questions, please refer to the full questionnaire in the appendix on the PLUTO website. Below we outline our reasoning behind the inclusion of each category along with a number of example questions.

Limitations of PLUTO and data solidarity

While Data Solidarity and the PLUTO tool offer valuable insights and opportunities for enhanced data governance, they also have limitations. These limitations stem from the complexity of the data landscape, the evolving nature of the concept of public value, and the normative choices made in their development.

One of the primary limitations of the PLUTO tool is the reductionism inherent in any tool that transforms complex qualitative assessments into quantified inputs. The tool may not fully capture the intricate nuances and multifaceted aspects of real-world data uses, which are often influenced by factors beyond the four categories prioritised in PLUTO. We acknowledge that a tool focusing on four categories cannot encapsulate the full complexity of data use. At the same time, we believe this focused approach provides a valuable starting point for assessing public value. For instance, an argument could be made that a large company using data for targeted advertising creates public value if personalised ads lead people to consume more sustainable products. PLUTO’s categories may not adequately account for such complex and context-specific scenarios. This highlights the need for ongoing refinement and adaptation of the tool.

There is a great diversity of data uses in contemporary societies. The effects of data use are often influenced by factors that extend well beyond the categories employed by PLUTO. While PLUTO does not encompass every possible data use context, its current categories represent a much-needed foundational framework. We anticipate that future iterations of the tool will address a broader range of applications as the tool evolves. In the current data-driven world, the range of applications and their implications are constantly expanding. Rather than attempting to capture every single use case immediately, we have prioritised creating a flexible tool that can adapt to these changes. This is an area where further development and user feedback will be crucial. As a result, the tool may struggle to comprehensively evaluate data uses outside the predefined categories. Having said this, the tool’s nature as a prototype that can be adjusted and amended to suit changing data practices, or the particular concerns and needs of specific organisations or communities (e.g. an organisation working on the green transition could place specific emphasis on climate-related and other environmental impacts of data use), mitigates this concern.

To facilitate the concept of data solidarity, it is essential to attach numerical or relative values to different data uses. While there are inherent challenges in quantifying something as complex as public value, this is a common challenge within the field of data governance and one that PLUTO, in its current form, does not fully resolve. However, we believe that the structured approach offered by PLUTO provides a more consistent basis for evaluating public value than relying solely on qualitative judgements. Furthermore, answering the questions in the tool can be influenced by subjective judgments, which may result in disputes over the assigned scores. We acknowledge this subjectivity and are exploring ways to enhance objectivity and transparency in future versions.

Public value is a dynamic and evolving concept, which means its interpretation may vary across different contexts, stakeholders, and time periods. PLUTO’s normative choices in defining public value may not align with every perspective, and disagreements over the very meaning of public value persist. The evolving nature of this concept introduces a level of uncertainty into the tool’s assessments. Public value, as a concept, is not static; it evolves in response to societal changes, technological advancements, and shifting values. As a result, PLUTO will need to adapt its criteria and definitions over time to remain relevant. Also for this reason we decided to keep PLUTO at a general level (i.e. the questions asked are applicable to different types of uses in different contexts). This widely applicable version of PLUTO can hence serve as a basis for context-specific versions of the tool that reflect contemporary or particular understandings of public value.

The question of how PLUTO can be institutionalized and whether its use can be made mandatory raises further complexities. Implementing PLUTO as a mandatory framework may require changes in existing laws and regulations, and this process could be met with resistance or difficulties in enforcement. The potential legal implications and the need for institutional support are important aspects to consider when contemplating the widespread adoption of PLUTO. We believe that public value assessments with a tool based on the PLUTO concept have a place in various fields of data use (e.g. public bodies could decide to prioritise data access requests by applicants whose suggested data use is likely to generate a lot of public value; regulators could decide not to grant licenses to data users who propose data use with negative public value, etc). While we advocate for regulators and policy makers to pay more attention to the public value of data use, we continue to improve the PLUTO prototype. The data gathered from real-world applications will inform further refinements, ensuring the tool’s effectiveness and reliability before widespread implementation. PLUTO’s flexible design allows adaptation to diverse contexts and priorities, while its focus on transparency and stakeholder engagement fosters a more inclusive decision-making process. These features, we believe, make PLUTO a valuable tool for navigating the complex data landscape. To operationalise data solidarity, it is essential to define and distribute responsibilities clearly. Policy interventions, such as implementing data taxes or establishing Harm Mitigation Bodies, require action from policymakers at national and supranational levels. Conversely, the practical implementation and daily operation of harm mitigation mechanisms is the responsibility of the data-using organisations themselves. Collaborative partnerships between public bodies (e.g. Findata), research institutions, and nonprofit organisations are often the most effective way to facilitate beneficial data uses. The development and application of assessment tools like PLUTO, which are crucial for evaluating public value, can be spearheaded by academic or civil society groups, potentially with public funding support. This distribution of labour aligns responsibilities with the relevant expertise and capacities of diverse actors, fostering a collaborative approach to realising data solidarity.

PLUTO is envisioned as an evolving tool. We recognise its current limitations but believe that releasing it in its current form, and actively soliciting feedback, will enable faster development and refinement compared to waiting for a ‘perfect’ version. Domain-specific versions of the tool may have different weightings and adapt to unique contexts and interpretations of public value. This iterative approach to development allows us to respond to user needs and evolving understandings of public value. The tool’s creators anticipate ongoing refinement and expect the broader research community to contribute to its development, thereby addressing some of the challenges identified. We are actively seeking input and collaboration to enhance the tool’s functionality and address the identified limitations. In response to these challenges, several measures have been taken to enhance the transparency and adaptability of PLUTO. At the website where the tool is available, an appendix is provided for each new iteration of the tool (on the publicly available website), offering detailed explanations of the weightings assigned to various factors, promoting transparency and allowing for weightings to be contested. Additionally, the content management system (CMS) feature enables easy adjustments to be made to weights, questions, answers, and additional information by authorised people, even if they have only limited computer or programming knowledge. It is also important to treat PLUTO as an aide in making a structured assessment – and supporting a structured debate – about public value is, and not as a tick-box exercise yielding an infallible outcome. Moreover, the use of PLUTO, and the results that it yields, should be seen in the context of the values that underpin the data solidarity approach.

In summary, while data solidarity and the PLUTO tool hold promise for advancing data governance, they are not without limitations. Their reductionist nature, challenges in quantifying data value, evolving interpretations of public value, and potential legal and institutional hurdles all require careful consideration. As data governance continues to evolve, addressing these limitations is essential to develop a more comprehensive and adaptable framework for evaluating the public value of data uses. To see how PLUTO could apply in a real-world scenario, we return to the example of Google’s Sidewalk Labs project mentioned earlier. This ‘smart city’ project, intended to use comprehensive data collection to improve urban infrastructure and services, raised significant concerns about privacy, data governance, and public accountability despite its potential for public value. By using PLUTO’s structured approach – which assesses data user information, benefit distribution, risk evaluation, and necessary safeguards – stakeholders could systematically evaluate the advantages and limitations of such data practices. PLUTO, for instance, would allow stakeholders to clarify how benefits from data collection might be equitably distributed across the population and highlight high-risk practices, such as continuous geolocation tracking, that would benefit from mitigative measures. We aim to expand on this with comprehensive case studies in future work.

Conclusion and outlook

In this paper, we have highlighted the inadequacy of existing data governance regimes, which predominantly rely on categorising data by type. These conventional frameworks have become increasingly obsolete in the era of digital data proliferation and the potential for data linkage. Such developments have rendered the traditional boundaries used to segregate different data types brittle, prompting the need to adopt a more context-driven approach.

Our proposed solution, data solidarity, represents a significant paradigm shift. It emphasises the significance of data uses over data types in determining the appropriate policy responses. In essence, our framework advocates that data uses generating high public value, characterised by high benefits and low risks, should be actively facilitated. Conversely, those data uses involving high risks and high benefits should have their risks mitigated to maximise public value. Data uses with high risks and low benefits must be prohibited, while those with low risks and low benefits should be subject to taxation through financial levies or benefit-sharing agreements.

At the core of our approach lies the notion of public value, which is instrumental in guiding our actions. Public value considerations, in our view, encompass both environmental sustainability and the well-being of future generations. It is essential to recognise that the concept of public value transcends short-term gains and encompasses the long-term welfare of society and the planet. Realising data solidarity requires being able to assess the public value of data use. For this reason, PLUTO, a tool that can be employed by individuals and organisations to evaluate the public value of specific instances of data use, was developed. We are keen on providing insight into the process of developing this tool, with the aim of enabling others to either replicate our approach or learn from our experiences and challenges. We understand that the concept of data solidarity and the PLUTO tool may have limitations. Collaborating with diverse stakeholders can help us identify and address these limitations effectively, ensuring that our approach remains robust and adaptable to changing circumstances.

Looking ahead, we have outlined three key plans for the future. First, we are committed to refining the PLUTO prototype to make it as user-friendly and transparent as possible. This entails fine-tuning the weightings and response mechanisms to ensure that the tool can be easily understood and applied by a broad range of users. Second, it is key to consider versions of PLUTO that are tailored to the needs of specific communities, institutions, or use contexts. We aim to identify specific domains in which specialised versions of PLUTO can be most effectively employed. PLUTO will first be tested and refined in the health domain. This choice is based on two primary reasons. Firstly, this approach has found support within the global health community (see e.g. DTH Lab, Citation2023). Secondly, there are concerns about the inequitable use of health data and the high potential benefits and risks associated with its use. PLUTO aims to address these concerns, collaborating with experts and stakeholders to ensure responsible and ethical deployment in the realm of healthcare. Relatedly, we are looking to explore opportunities to develop PLUTO certificates (similar to Secure Sockets Layer certificates) that indicate the public value of a number of standard data uses. Additionally, we will explore the idea of assigning custodial responsibilities for PLUTO in these sectors and support organisations in its development. We are also committed to maintaining control over the basic PLUTO tool while ensuring that the tool is not misappropriated. Lastly, it is important to study the legal changes necessary to institutionalize data solidarity. We acknowledge the need for legal changes to enhance the impact of data solidarity on current data governance approaches. This includes advocating for legal reforms and normative frameworks that align with the principles of data solidarity and support its implementation in various contexts.

In conclusion, our paper underscores the significance of data solidarity in the evolving landscape of data governance. It emphasises a shift from data types to data uses, with a keen focus on public value and sustainability. The development of the PLUTO tool represents a critical step toward practical implementation, and we are committed to refining and expanding its utility. As we move forward, we invite collaboration, feedback, and the collective effort of diverse communities to advance the cause of data solidarity and its vital role in shaping a more equitable and sustainable data-driven future.

Transparency

While one of the authors is currently employed in industry, the development of the PLUTO tool and the insights presented in this paper were strictly within the academic context.