Data Governance Deep Dives - Energy Data Coops
Notes from our third data governance deep dive session
Every month Digital Commoners together with the Aapti Institute hosts Data Governance Deep Dives. An informal session that connects practitioners and theorists, while we together explore real-world data governance models.
This month the discussion was centred on how data cooperative structures can enable data sharing, especially in the energy sector. Key insights from research conducted by an energy data cooperative were shared. The goal was to create a model for energy co-operatives that could enable value creation from the data they hold and formulate a mechanism for this value to be returned to communities.
In the energy sector, it was found that companies often use energy metres that can determine the type of appliances that people are utilising, by analysing behavioural data. However, the potential value from this data fails to reach individuals and the community, which is where energy cooperatives play an important role. Energy Cooperatives engage in pooling resources to enable local energy micro generation, retro-fitting, energy profiling and allied efficiencies - with an aim to reduce carbon emissions, while enabling communities to benefit economically from the efficient use of energy data.
The discussion involved key points on data flow, consent management, and cooperative structures, which are relevant across sectors.
Types of data flows involved in Co-operative Structures
A vital aspect of structuring data-sharing in Co-operatives is Data Flow. One key finding of the research was 5 Principle data flows:
Flow 1, Member to Co-operative: A lot of existing data of members of co-operatives, is siloed and underutilised. One way to combat this is through creating a data infrastructure to which data can flow, leading to data pooling.
Flow 2, Intra Co-op Flow: Often, members of Co-operatives are unable to access information that is specific to them. Enabling Intra- Co-operative flow would ensure that data requesters can benefit from the pool of data. In the energy sector, this would mean that people living in similar houses can avail information on energy efficient appliances specific to their homes. This leads to environmental benefits for the community, but also economic benefits for individuals.
Flow 3, Inter-Co-op Flow: This is an attempt to connect like minded organisations/ individuals without commercial middlemen, through federating data. In the energy sector, this means that Co-operatives can pool data from various micro-generation facilities to efficiently utilise excess capacities from specific resources, by matching buyer co-operatives with seller co-operatives.
Flow 4, Co-operative to Open Data Commons Flow: This involves data flow with third parties – which may include commercial entities engaged in licensing agreements with appliance providers or energy providers; or academia/ knowledge generating institutions.
Flow 5, Co-operative to Data Commons: This can involve sharing altruistically for common good and open access/use.
How can consent be managed in Data Co-operatives?
While data co-operatives are built for the purpose of data sharing, and therefore are legally permitted to share data (in the UK/EU), it is also pertinent to consider that data co-operatives are built on trust and confidence. Hence, consent in the context of data co-operatives is not from a purely legal perspective, but involves ethical concerns. The research yielded three possible types of Consent Mechanisms: Granular consent: A mechanism where every type of data is given explicit consent to be shared, and choices for sharing specific types of data is highly customisable. However, under varying circumstances of data sharing/ collection- seeking consent for every new type of data to be collected could be burdensome on both individuals and organisations.
Persona/ Archetype-based Permissions: This is a proxy based system where consents are based on the characteristics of a fictional persona or archetype. Users identify with the persona/ archetype that matches their own and accept their recommendations. It was found that the creation/designing of profiles/ archetypes is an extremely complex process as it must address and reflect values, choices, preferences etc of the member, which may also be dynamic.
Traffic light consent: This is a simple process where members choose what the nature of their consent would be: Green- share data with anyone; Amber- Happy to share but choices made on a case by case basis; Red- data is used for the co-operative’s own purpose. How should third party access be mediated ? Co-operative models are built on the basis of trust, confidence, representation and empowerment. Therefore, governing third party data access by matching consent to data requests becomes challenging. The judgement on granting access involves a complex network of ethical decisions by the Co-operative. While considering purposes of data use and values of the data user, the Co-operative must also factor in features like transparency, ethos, geographic impact, industry type etc. A preference table where members of the Co-operative list their choices on the values they prioritize, could simplify determining access. This would also enable individual members to exercise autonomy in tailoring choices according to their own specifications. While there is scope for automating this decision making process, the trust basis for Co-operative models would require that the algorithmic process behind the decision making be transparent and legible for members, in order to truly preserve their agency. Simplified process of data sharing in a Data Cooperative
The discussion yielded a simplified explanation of the process that facilitates data sharing in a Co-operative, in a straightforward 6-step format:
Step 1: Someone joins a data cooperative- the purpose is to pool and share data- which is a legitimate interest- the decisions of the organisation are based on ethics and trust rather than legal mandates.
Step 2: Members share preferences through a consent portal & those attributes get assigned to the data.
Step 3: New members select their sharing preferences through a consent portal
Step 4: A data user approaches the data cooperative for members data and submits details regarding themselves and purpose of data request
Step 5: Details are verified and matched with the conditions of sharing that have been determined through the consent portal
Step 6: ID details are verified and matched with the conditions of consent, then the data is released to the requesting data user
Questions that remain
Governing Third Party Access: The governance mechanism for third party access must enable recourse to members for enforcing standards of protected data use. For members to be confident in sharing data, it becomes important to create an accountability mechanism through which data users can be held up to ethical criteria and member preferences. While licensing/contractual arrangements allow for members to seek redress under law, heavy expenses may preclude co-operatives from choosing this route. A community audit or participatory budgeting process could potentially solve this problem, but its feasibility remains unclear. Algorithmic Accountability While transparency is an important facet to granting data access in the first place, it also becomes relevant in the context of analysis. It is possible that even if a member opts out of sharing data with a specific process- the analysis might arrive at conclusions based on averages that may somehow affect that member. Therefore, the subsequent processes in respect of the data is an important criteria for Co-operatives to assess, prior to granting access. This is also closely tied to future revocation of consent and the dynamics of governing consent overtime. How can data users deal with consent revocation once the data is aggregated? Ethical alignment between like minded organisations could potentially enhance trust in the overall processing, analysing and utilization of data shared by members, but is this sufficient to safeguard member rights? The Co-operative could additionally act as a mediator in harnessing collective action for deletion of data, to enable the effective exercise of members’ rights. Whether the appointment of ethics boards in Co-operatives could effectively enhance algorithmic legibility and accountability? Through this measure, individuals could donate data by surrendering control to the Ethics Board, which would oversee data use, access and subsequent analysis. Sandboxing data could also preserve algorithmic accountability, but is implementing sandboxing viable under the co-operative model? This process would require data users to carry out calculations/ analysis within the sandboxed space and may also allow for the algorithms to be audited, but data co-operatives may not have the infrastructural capabilities to enable these processes.
Agency and Scalability: Individual agency and empowerment through collective representation, is central to the data co-operative model. This is also what sets it apart from data trusts. This leads to challenges in maintaining a sustainable data infrastructure that is scalable, without compromising on agency of the members or breaking trust connections. In economic terms, the value that the cooperative actually creates through external transactions, must be sufficient to sustain the organisation. Another pertinent concern is on who could be entrusted with the decision making capabilities on behalf of the co-operative? One solution is to appoint a nominee, who has the confidence of the co-operative, to ultimately control its decisions. If so, what would be the process and method of selection, and the role of this nominee? While consent formulation and algorithmic accountability were recurrent themes, the session yielded valuable insights on possible mechanisms for consent management. However, it also brought to light key concerns on empowering individual members to ensure third party adherence to ethical and agreed standards for data use and analysis.