Online collective ideation platforms, such as OpenIDEO or Quirky, have demonstrated the potential of large-scale collective innovation in various domains. However, the users of these platforms face new challenges of leveraging collective contributions. The large number of collected ideas prevents users from making full use of these ideas. Finding inspirations from the ideas involves wading through a sea of possibly mundane and redundant ideas. Synthesizing a few solutions from these ideas takes a lot of time and e ort. I argue that leaving users to explore ideas in a haphazard manner is ine ective and can decrease the quality of people's creative output. Prior work in cognitive science and creativity research has also suggested that deliberate exploration of the solution space can improve users' creative output and experience.
I introduce the concept of an idea map, a computational model of the emerging solution space that enables deliberate exploration interactions: 1) presenting a set of ideas with a controlled level of diversity appropriate to the stage of the creative process and 2) presenting a summary view of the solution space. I describe two scalable crowdsourced methods for generating this computational model. The first method computes the model from responses from small micro-task questions. The second method takes an “integrated crowdsourcing” approach that computes the model from users' natural activities during idea generation. The evaluation of the derived models show that the idea maps from both approaches agree with human judgments of similarities among ideas. I show the application of the idea map concept through experiments and a system called IdeaHound. IdeaHound derives an idea map using the integrated crowdsourcing approach and uses the derived model to guide users' exploration of the solution space. The results of the experiments show that an idea map can inspire people to generate diverse ideas. The integrated activities that enable IdeaHound to collect similarity judgments do not deter users from generating ideas and provide enough information to generate a reliable idea map. I also present a study on the e ects of di erent timings of delivering example ideas on an individual's idea generation. The results demonstrate that an intelligent system can provide inspiration at the right moment by using a computational model that is aware of semantic relationships between ideas. Finally, I demonstrate how to use an idea map to support sensemaking during the solution synthesis and present an empirical study of the e ect of presenting a summary view of ideas on people's solution synthesis.