Mitigating Dataset Harms Requires Stewardship: Lessons from 1000 Papers
This work addresses ethical issues in machine learning datasets for researchers and practitioners, offering insights into mitigating harms, though it is incremental in building on existing calls for higher standards.
The study analyzed 1000 papers citing three problematic face/person recognition datasets to understand ethical harms, finding that derivative datasets, unclear licenses, and management practices introduce concerns, and proposed a distributed approach for harm mitigation across the dataset lifecycle.
Machine learning datasets have elicited concerns about privacy, bias, and unethical applications, leading to the retraction of prominent datasets such as DukeMTMC, MS-Celeb-1M, and Tiny Images. In response, the machine learning community has called for higher ethical standards in dataset creation. To help inform these efforts, we studied three influential but ethically problematic face and person recognition datasets -- Labeled Faces in the Wild (LFW), MS-Celeb-1M, and DukeMTM -- by analyzing nearly 1000 papers that cite them. We found that the creation of derivative datasets and models, broader technological and social change, the lack of clarity of licenses, and dataset management practices can introduce a wide range of ethical concerns. We conclude by suggesting a distributed approach to harm mitigation that considers the entire life cycle of a dataset.