Reciprocity in Machine Learning
This work addresses fairness and equity concerns for individuals contributing data to ML systems, but it is incremental as it builds on existing influence measures.
The paper tackles the problem of assessing whether contributions and benefits in machine learning systems are reciprocal, proposing measures for outflows, inflows, and reciprocity based on training data influence, with initial results indicating approximate reciprocity under certain distributional assumptions for some model classes.
Machine learning is pervasive. It powers recommender systems such as Spotify, Instagram and YouTube, and health-care systems via models that predict sleep patterns, or the risk of disease. Individuals contribute data to these models and benefit from them. Are these contributions (outflows of influence) and benefits (inflows of influence) reciprocal? We propose measures of outflows, inflows and reciprocity building on previously proposed measures of training data influence. Our initial theoretical and empirical results indicate that under certain distributional assumptions, some classes of models are approximately reciprocal. We conclude with several open directions.