A good data governance framework is essential to ensure that data is consistent, trustworthy, and doesn't get misused. Therefore, the European Parliament is looking for policy options that align with a data justice perspective for artificial intelligence (AI). Such a perspective focuses on equity, recognition, and representation of plural interests 1 . The Panel for the Future of Science and Technology conducted a study for the European Parliament to identify and examine different policy options and methods for the EU's data governance framework, looking at various contexts across Europe and around the world and their potential contribution to European digital strategies. Their aim is to answer the question: “How to foster a positive vision of AI as contributing to public goods and creating public value?” Well, only through governance models that distribute power and the data ecosystems that they rely on.
How to foster a positive vision of AI as contributing to public goods and creating public value?
Let's start with defining governance. Governance provides an institutional architecture from which different systems are organised, it offers a framework and regulation for participation by different actors, it allows for negotiations between competing authorities by providing policies and procedures and provides ways in which conflicts can be mediated 2 . Regulation can be seen as a large subset of governance that steers the flow of events and behaviour. Within the EU there are different forms of established and emerging data regulation at state, private, and global levels. Even though each is presented separately, in practice they interact with each other extensively. Laws and regulations can help with creating a data just governance. New legislations can offer a way of limiting or giving power and holding accountable of both public and private actors regarding the use of data.
The currently dominant data governance approach aims for a worldwide free market for data and embraces data as an asset and promotes its appropriation by and through a range of artificial intelligence (AI) applications. However, with this approach, overarching principles potentially conflict and imbalances between users and platforms can be created, especially in terms of access to the data and control over its use. This means that there is a question of prioritising and balancing. Through data governance, these principles can be balanced to determine how different objectives should interact, which public values should be incorporated in frameworks, and how people's interests should be represented in governance. Alternative models of data governance are possible and are already practiced in various countries around the world.
- Public data trusts
The public data trust model has a public institution in charge of managing data. The public institution accesses, aggregates, and uses the data collected from different sources. The agreement and relationship between data subjects and public institutions is based on trust and depends on public engagement like consultations, strong accountability mechanisms, and collective benefits. Public interest is important in this model, which gives the possibility to use public data for policy making, social innovation and to address social challenges.
- Data collaboratives
The data collaborative is more an alternative model rather than a model that goes against the current model. The concept of a data collaborative is that private data collected by companies is pooled with public data through an independent third party. A third party governs the data. Data access, sharing and use is restricted to the members only. In this model, data is defined as a public interest and public good.
- Data (semi) commons
This model interacts the most with the free data market. The commons model is a model where a system is based on making data available through shared infrastructures in a way that minimises the ability of economic interests to exclude it. The idea of data commons is mainly focused on how to keep public goods available to the public.
- Data cooperatives
In this alternative data governance model data subjects are the main actors that organise the use, sharing and access to data through the collective organisation of groups or communities 3 . The main value of this model is the public interest, data subjects voluntarily provide data to the pool, they also participate in the construction of the governance model.
- Indigenous data sovereignty
Data sovereignty is more of a principle than it is a model. The idea is that data is an infrastructure that should be managed according to the norms and values of data subjects. The aim is to promote and encourage self-determination and rebalancing the current power relationship between data subjects and data controllers.
- Personal data sovereignty
This model defines data subjects as market agents that aim to control the access, use and sharing of their data. In this model data subjects have more choice over their data and receive more benefits. This model is dependent on trust from individuals to share and transfer personal data. Each model has its pros and cons, understanding the differences and similarities between the models pushes to focus more on the principles that we need to protect and balance which ones are of most important.
It is important that people have control over their own visibility. People should be represented through their data when it is in their interests. People should have autonomy over whether they adopt technologies or not. Most importantly, people should not be made responsible for protecting themselves from exploitation through systems. It should be the responsibility of the government to ensure that the public is protected, both individually and collectively.
Democratising the process of oversight and enforcement regarding data and AI could be a solution to ensure that these social concerns will be addressed. However, it is important to ensure that oversight and enforcement structures have public-facing component to have some accountability. In conclusion, the EU still has work to do in conceptualising what kind of public good data should be and how it should be governed.
- 1 https://www.europarl.europa.eu/stoa/en/document/EPRS_STU(2022)729533
- 2 Gunnar F. Schuppert, 'Governance,' in International Encyclopedia of the Social & Behavioral Sciences (Second Edition), ed. James D. Wright (Oxford: Elsevier, 2015), 292–300, https://doi.org/10.1016/B978-0-08-097086-8.75020-3
- 3 Mozilla Insights, Jonathan van Geuns, and Ana Brandusescu, 'Shifting Power Through Data Governance' (Mozilla, September 2020), https://assets.mofoprod.net/network/documents/ShiftingPower.pdf