Observing and Recommending from a Social Web with Biases
This addresses fairness and legal compliance in AI systems for developers and policymakers, but it appears incremental as it focuses on existing concerns without new solutions.
The paper investigates how developers can ensure that black box algorithms do not unlawfully discriminate against individuals with protected characteristics like gender or race, but it does not provide specific results or numbers.
The research question this report addresses is: how, and to what extent, those directly involved with the design, development and employment of a specific black box algorithm can be certain that it is not unlawfully discriminating (directly and/or indirectly) against particular persons with protected characteristics (e.g. gender, race and ethnicity)?