Title: Economic mechanisms for crowdsourcing markets
Authors: Alexey Drutsa - Yandex (Russia) [presenting]
Abstract: The current level of automation significantly affects the global economy, and the power of this influence continues to spread fast. On the one hand, this leads to a reduction in the need for labor and, thus, increases the unemployment rate. On the other hand, since machine learning technologies play a key role in automation, the need for collected and processed data is growing. In fact, data are now being transformed into a new kind of ``oil'' consumed by new ``machines'' (AI). Surprisingly, this need for data may be a clue to solve the problem of people unemployment and to balance the negative tendency in the global economy between humans and machines. In order to produce large amounts of data (for example, text entity recognition, object segmentation on a photo, etc.), machine learning-based products actively use crowdsourcing platforms (e.g., MTurk and Toloka) -- platforms of a two-sided market between task requesters and their performers. The main feature of this market is that tasks are short, while performers can freely choose which tasks to execute. In particular, this is why this market is not similar to the classic labor market. We discuss how economic mechanisms can help crowdsourcing markets by overviewing main open research questions that arise while building a crowdsourcing platform. In particular, we consider the problem of ``weak'' matching: how to match performers to a ranking list of tasks taking into account the incentives of both sides of the market.