Data Quality, Data Diversity and Data Provenance: An Ethical Perspective

Predictive modelsPredictive model are increasingly pervasive in many areas of daily life, from engineering to the social sciences. Their use is gradually automating and replacing tasks that were typically done manually and are no longer sustainable. Recently, the increasing amount of data and developments in data science facilitated the deployment of predictive models by creating them through data-driven approaches. The first wave of data science aimed to improve predictive models in terms of accuracy and efficiency. However, the ethical implications and the careless use of these models have been overlooked. After several ethical problems arose during the second wave of data science, the focus shifted from what could be done with data to what should or should not be done with them and how. As a result of this new ethical attention to the use of data, the entire life cycle of data was then analyzed with newly proposed methodologies. Data QualityData quality is one of the main factors often under the spotlight. It is fundamental to build an accurate and ethical predictive modelPredictive model since it directly affects the model’s outcomes. However, other data-related elements are equally important, such as diversity and provenance. This chapter claims that such aspects are also essential and should be regularly and equally present in the data science ethical-technical debate. Thus, starting from practical examples, this work first presents Data Quality, Data DiversityData diversity, and Data ProvenanceData provenance problems. Then, it discusses the corresponding trade-off and mitigations and how they coexist and cooperates to address some ethical issues from a technical perspective.