JED-AIs: Justice, Energy, Demand flexibility and AI for Sustainability
Lead Research Organisation:
University of Strathclyde
Department Name: Electronic and Electrical Engineering
Abstract
JED-AI brings together electrical engineering and social science co-leads across the socio-technical research spectrum to develop an inclusive methodology for learning about and strengthening households' capabilities to participate in energy demand flexibility services through interdisciplinary co-designed household segmentation and interventions. The aim is to help realise a more just and sustainable net-zero transition. JED-AI will provide clear answers to the following research questions: 1. How do different household- and community- scale attributes and capabilities shape, enable and constrain household participation and engagement in energy demand flexibility? 2. How can AI analysis be combined with social science insights to generate more just and sustainable interventions in energy demand flexibility? 3. What are the key productivities, effects, challenges and benefits of interdisciplinary approaches to stimulating equity and justice in the net zero transition?
JED-AI builds on an integrated capabilities framework for fair and inclusive participation in demand response that defines flexibility capabilities as the ability to shift energy use in time and space, or through changes in intensity or vector (e.g., fuel to electricity). JED-AI will design novel household segmentation approaches, integrating social-science smart energy capabilities attributes at both household and community scales with engineering quantification of flexibilities, through automated learning from historical smart meter data analysis of usage patterns across hundreds of households with various indicators of capabilities, to generate clusters of household attributes. This will generate a spectrum of capabilities that comprehensively characterise potential for engaging in demand flexibility services. The segmentation will inform recruitment screening and selection of living labs, from both ends of the "capability spectrum", to design interventions. An interdisciplinary methodology for intervention, in the form of accessible prompts co-designed with households, tailored to particular needs and routines, will be co-created and trialled by the research team based on social science capability assessment and principles of trustworthy AI, by placing social science in the centre of the AI design loop. The progress, outcomes and interdisciplinary working will be regularly assessed through "checkpoints" and the development of a novel multi-criteria interdisciplinary evaluation strategy to assess if and to what extent the designed interventions have strengthened household capabilities to engage with demand flexibility. The evaluation strategy will include engineering as well as social science metrics evaluating accuracy, environmental impact, grid friendliness and adoption, but also qualitative measures that indicate fairness and inclusivity.
Through the aforementioned levels of mixed methods analysis, JED-AI will explore how social science data and methodologies can enhance the trustworthiness of AI-based recommendations as well as how engineering-designed AI recommendations can provide an evidence-base to scale up social science intervention methods. The lessons learnt will have much wider implications in the way AI-informed energy demand research is designed and deployed for trials, taking advantage of the detail from social science processes to reduce risk of harmful decisions and removing human agency, and embedding inclusiveness and fairness. On the other hand, social sciences will benefit from the scalability and low cost of large-scale analysis of low carbon technology usage patterns leading to improved understanding of how integrated capabilities framework is reflected into the low carbon technology adoption, and how inclusion and justice measures affect and are affected by engineering technology.
JED-AI builds on an integrated capabilities framework for fair and inclusive participation in demand response that defines flexibility capabilities as the ability to shift energy use in time and space, or through changes in intensity or vector (e.g., fuel to electricity). JED-AI will design novel household segmentation approaches, integrating social-science smart energy capabilities attributes at both household and community scales with engineering quantification of flexibilities, through automated learning from historical smart meter data analysis of usage patterns across hundreds of households with various indicators of capabilities, to generate clusters of household attributes. This will generate a spectrum of capabilities that comprehensively characterise potential for engaging in demand flexibility services. The segmentation will inform recruitment screening and selection of living labs, from both ends of the "capability spectrum", to design interventions. An interdisciplinary methodology for intervention, in the form of accessible prompts co-designed with households, tailored to particular needs and routines, will be co-created and trialled by the research team based on social science capability assessment and principles of trustworthy AI, by placing social science in the centre of the AI design loop. The progress, outcomes and interdisciplinary working will be regularly assessed through "checkpoints" and the development of a novel multi-criteria interdisciplinary evaluation strategy to assess if and to what extent the designed interventions have strengthened household capabilities to engage with demand flexibility. The evaluation strategy will include engineering as well as social science metrics evaluating accuracy, environmental impact, grid friendliness and adoption, but also qualitative measures that indicate fairness and inclusivity.
Through the aforementioned levels of mixed methods analysis, JED-AI will explore how social science data and methodologies can enhance the trustworthiness of AI-based recommendations as well as how engineering-designed AI recommendations can provide an evidence-base to scale up social science intervention methods. The lessons learnt will have much wider implications in the way AI-informed energy demand research is designed and deployed for trials, taking advantage of the detail from social science processes to reduce risk of harmful decisions and removing human agency, and embedding inclusiveness and fairness. On the other hand, social sciences will benefit from the scalability and low cost of large-scale analysis of low carbon technology usage patterns leading to improved understanding of how integrated capabilities framework is reflected into the low carbon technology adoption, and how inclusion and justice measures affect and are affected by engineering technology.
