AI and Energy Justice
Tilburg Institute for Law, Technology and Society (Tilburg University)
Artificial Intelligence (AI) techniques are increasingly used to address some of the challenges of the energy transition. Such challenges include the integration of renewable and more volatile energy sources as well as the increased demands on the limited network capacity due to the growing electrification. AI techniques have been used to, among other things, forecast and predict energy consumption and supply, control behind-the-meter appliances and devices, optimize energy storage, schedule battery charging, and monitor system health. These techniques promise to enable new coordination mechanisms, new scales of flexibility and complexity, as well as new relationships, roles, and responsibilities. As such, they could support the shift from centralized control to decentralized control of electricity systems.
The use of AI technologies in electricity networks offers many opportunities but also raises multiple concerns. One key concern is how AI will affect energy justice, that is, the fair distribution of the benefits and burdens of energy systems, as well as equitable access to decision-making processes and adequate recognition of different stakeholder groups. The delegation of more and more decision-making tasks in the energy sector to opaque and complex AI systems could lead to energy injustices when these systems disproportionality disadvantages certain groups in society. What are the energy justice implications of the use of AI in smart electricity systems and what does this mean for the design and regulation of these technologies? We address this crucial question in our paper ‘AI and Energy Justice’ published in Energies.
Energy justice implications of AI techniques: the example of PV Curtailment
‘Energy justice’ is a concept that has emerged predominantly in social science research to highlight that energy related decisions, particularly as part of the energy transition, should produce just outcomes. One predominant framework to understand what energy justice encompasses three dimensions: distributive justice (the fair distribution of burdens and benefits), procedural justice (equitable access to decision-making processes) and recognition justice (the acknowledgment and respect of diverse stakeholder groups). 
In our paper, we use these three dimensions of energy justice to evaluate how AI technologies promote energy justice or create risks of injustice. To make this more tangible, we looked at AI technologies applied to the curtailment of solar photovoltaic (PV) systems in low-voltage networks. Curtailment involves decreasing or stopping the generation of electricity from particular solar panels when there is too much energy available and too little demand in the grid. There are various ways of doing this. For instance, curtailing those panels where the voltage is highest or – with the help of smart devices – spreading curtailment over multiple connections. AI technologies can be used here to find a solution that achieves certain goals, such as minimization of overall curtailment or a far distribution of curtailment. This provides an example of the application of AI techniques to address grid management issues that directly impact citizens.
From our analysis, we found that design choices, such as which type of AI techniques, which data or which fairness metrics to use, affect energy justice to the extent that a particular design choice leads to a particular distribution of burdens and benefits across stakeholders. It can either increase or inhibit transparency and participation in decision-making processes, and to the inclusion or exclusion of certain stakeholder groups.
With regard to distributive justice, design choices can affect how costs and benefits are distributed across relevant stakeholders, including economic costs, but also cost and benefits in terms of power or knowledge. A curtailment algorithm may disproportionately benefit some prosumers, in particular those close to the transformer, as compared to those furthest away. ML algorithms can correct this existing inequality, but there are multiple ways of translating fairness into the design of the algorithm. Different approaches will thus have different effects on the costs and benefits of curtailment strategies.
With regard to procedural justice, the way that AI technologies are designed and implemented is directly related to how much and in what way stakeholders, including prosumers and non-users, can be involved in the choices that directly affect them. The opaqueness, complexity, or their enabling functions, for example, affect the extent to which accountability is possible or whether interventions can be made at any time.
Finally, in modeling contingent environments, ML techniques are based on choices on how and what to represent about the world that include or exclude aspects of the world, such as how to represent the variety of households with solar panels. Such choices have implications for recognition justice, as this may lead to the misrecognition of certain groups and their particular needs. Using more fine-grained data, for instance about households, could mitigate this to some extent, but this, too, is not without trade-offs, as it may, for example, lead to privacy and/or personal data protection concerns.
Moreover, we observed that operationalizations of energy justice in terms of legal norms or design guidelines applied to AI technologies in the electricity sector are lacking, in particular in the European Union (EU). Existing ethics guidelines and the proposal to regulate AI at EU level are too general and don’t reflect the particularities of the electricity sector. Current electricity sector legislation does not fully flesh out energy justice as a normative principle and does not provide guidance regarding AI technologies.
A better understanding of the interaction between AI technologies and energy justice is necessary, as these technologies are still in the early stage of development but may have significant disruptive effects on the energy sector. This is particularly important considering that policymakers in the EU increasingly see the digital and energy transitions as two interrelated developments, whereby digital technologies are seen as an enabling force to realize the energy transition. Our research is a first step towards understanding the relationship between AI and energy justice identifying what kinds of (legal and ethical) frameworks are needed to ensure that the digital and energy transitions will leave no one behind.
To dive deeper into the energy justice implications of AI technologies in the electricity sector, read our paper ‘AI and Energy Justice’ (open access).
This research is part of the research project MegaMind (Measuring, Gathering, Mining and Integrating Data for Self-management in the Edge of the Electricity System), (partly) funded by NWO (Dutch Research Council) through the Perspectief program under number P19–25.
 McCauley, D.; Heffron, R.; Stephan, H.; Jenkins, K. Advancing Energy Justice: The Triumvirate of Tenets. Int. Energy Law Rev. 2013, 32, 107–110