Adaptive and Interoperable Crowdsourcing

Abstract

Crowd-based computing is an increasingly popular paradigm for building Web applications, which uses the collective strength of human actors for performing tasks that are more suited to humans than computers. Interaction with the crowds was originally confined to specifically designed crowdsourcing platforms, such as Amazon Mechanical Turk. More recently, crowd-based computing has been reconsidered and extended, targeting social networks such as Facebook, Twitter, or LinkedIn, or including basic and direct interaction mechanisms, such as routing personal emails or tweets. Crowdsearcher, the system presented here, fosters interoperability and adaptation in crowd-based applications -- for example, the ability of supporting multiplatform applications and adapting them in reaction to events. This approach specifically supports dynamic interoperability (that is, the ability to modify the execution platforms while the application is ongoing) as a reaction to crowd behavior, which is hardly predictable. The authors show how to specify interoperability control at a high, declarative level and then implement it using active rules, thereby obtaining answers from crowds engaged in different communities. They also show the approach's effect on precision, delay, and cost.