Ethical Sensitivity in Machine Learning Development

Despite a great deal of attention to developing ethical mitigations for Machine Learning (ML) training data and models, we don't yet know how these interventions will be adopted by those who curate data and use them to train ML models. Will they help ML engineers find and address ethical concerns in their work? My proposed dissertation seeks to understand ML engineers? ethical sensitivity? their propensity to notice, analyze, and act on socially impactful aspects of their work-while curating training data and describe the effects of context documents and ethical guides as practice-based ethics interventions in this early stage of ML development. It asks how ML engineers recognize,particularize, and judge ethical questions while exploring new training data; introduces Ethical Sensitivity to the study of social computing; and will describe how Datasheets intervene in perception and particularization; and will develop a document that can help engineers move from particularization to judgment. It will accomplish these goals using a think aloud experiment with engineers working with unfamiliar training data (with or without a Datasheet), a Value Sensitive Design study that aims to fit an ethical mitigation guide to engineers? work practices, and a systematic review of ethical sensitivity.