Datasheets for Energy Datasets: An Ethically-Minded Approach to Documentation

Abstract

This work presents an argument for the use of specific documentation for the ethical development, use, and sharing of energy datasets, and an evaluation of current practice in the energy AI community. Drawing on a recently developed resource from the broader machine learning community and applying it to the specific context of energy AI research, opportunities for more transparent collection and distribution of energy datasets are revealed. To help elucidate the utility of the datasheets and the energy community’s current level of documentation, two publicly available energy datasets are chosen for analysis. One has published documentation covering 66% of the datasheet questionnaire, while the second covers 42% of the suggested information. Two additional questions are recommended for energy-relevant datasheets that will promote ethical AI practices in the energy domain. A new resource for exploring and aligning energy datasets with demographic data is provided.