So what is this workshop about?

In real world settings, the data used by a learning algorithm is often supplied by organisations and individuals directly impacted by the decisions said algorithm makes. As a result, decision makers are often engaged in strategic interactions with data providing agents. Examples of such settings include, but are not limited to, spam filtering, loan approvals, poll aggregation, and college admissions. Such scenarios do not neatly fit into classical learning theory, where it is typically assumed that algorithms have the ability to sample directly from a distribution of interest, without relying on data providing agents. On the other hand, the strategic relationship between data providers and learners can be substantially more nuanced and less pessimistic than the worst-case assumptions implicitly made by algorithms for adversarial learning. As a result, such settings present a novel problem, which lies at the intersection of machine learning, mechanism design, and game theory. The aim of this workshop is to bring together the broad community working on strategic learning problems of this form.