People are more creative at solving difficult design problems when they use relevant examples from outside of the problem's domain as inspirations. However, finding such “outside-the-box” inspirations is difficult, particularly in large idea repositories such as the web, because without guidance people select domains to search based on surface similarity to the problem's domain. In this paper, we demonstrate an approach in which non-experts identify domains that have the potential to yield useful and non-obvious inspirations for solutions. We report an empirical study demonstrating how crowds can generate domains of expertise and that showing people an abstract representation rather than the original problem helps them identify more distant domains. Crowd workers drawing inspirations from the distant domains produced more creative solutions to the original problem than did those who sought inspiration on their own, or drew inspiration from domains closer to or not sharing structural correspondence with the original problem.

Previous work has shown the promise of crowdsourcing analogical idea generation, where distributing the stages of analogical processing across many people can reduce fixation, identify inspirations from more diverse domains, and lead to more creative ideas. However, prior work has only considered problems with a single constraint, while many real-world problems involve multiple constraints. This paper contributes a systematic crowdsourcing approach for eliciting multiple constraints inherent in a problem and using those constraints to find inspirations useful in solving it. To do so we identify methods to elicit useful constraints at different levels of abstraction, and empirical results that identify how the level of abstraction influences creative idea generation. Our results show that crowds find the most useful inspirations when the problem domain is represented abstractly and constraints are represented more concretely.

Seeking solutions from one domain to solve problems in another is an effective process of innovation. This process of analogy searching is difficult for both humans and machines. In this paper, we present a novel approach for re-presenting a problem in terms of its abstract structure, and then allowing people to use this structural representation to find analogies. We propose a crowdsourcing process that helps people navigate a large dataset to find analogies. Through two experiments, we show the benefits of using abstract structural representations to search for ideas that are analogous to a source problem, and that these analogies result in better solutions than alternative approaches. This work provides a useful method for finding analogies, and can streamline innovation for both novices and professional designers.

Harnessing crowds can be a powerful mechanism for increasing innovation. However, current approaches to crowd innovation rely on large numbers of contributors generating ideas independently in an unstructured way. We introduce a new approach called distributed analogical idea generation, which aims to make idea generation more effective and less reliant on chance. Drawing from the literature in cognitive science on analogy and schema induction, our approach decomposes the creative process in a structured way amenable to using crowds. In three experiments we show that distributed analogical idea generation leads to better ideas than example-based approaches, and investigate the conditions under which crowds generate good schemas and ideas. Our results have implications for improving creativity and building systems for distributed crowd innovation.

A significant challenge for crowdsourcing has been increasing worker engagement and output quality. We explore the effects of social, learning, and financial strategies, and their combinations, on increasing worker retention across tasks and change in the quality of worker output. Through three experiments, we show that 1) using these strategies together increased workers' engagement and the quality of their work; 2) a social strategy was most effective for increasing engagement; 3) a learning strategy was most effective in improving quality. The findings of this paper provide strategies for harnessing the crowd to perform complex tasks, as well as insight into crowd workers' motivation.

A method of organizing the crowd to generate ideas is described. It integrates crowds using evolutionary algorithms. The method increases the creativity of ideas across generations, and it works better than greenfield idea generation. Specifically, a design space of internet-scale idea generation systems is defined, and one instance is tested: a crowd idea generation system that uses combination to improve previous designs. The key process of the system is the following: A crowd generates designs, then another crowd combines the designs of the previous crowd. In an experiment with 540 participants, the combined designs were compared to the initial designs and to the designs produced by a greenfield idea generation system. The results show that the sequential combination system produced more creative ideas in the last generation and outperformed the greenfield idea generation system. The design space of crowdsourced idea generation developed here may be used to instantiate systems that can be applied to a wide range of design problems. The work has both pragmatic and theoretical implications: New forms of coordination are now possible, and, using the crowd, it is possible to test existing and emerging theories of coordination and participatory design. Moreover, it may be possible for human designers, organized as a crowd, to codesign with each other and with automated algorithms.

The goal of this research is to perform large-scale experiments to see if the crowd, mediated by technology, produces creative designs by combining designs. To fulfill this goal, a sequential combination system is built. The system is a variant on a human based genetic algorithm, through which the crowd participates in an iterative process of design, evaluation, and combination. The study will provide a way of creatively solving problems in a number of different domains. It will also shed light on the mechanisms of social creativity: how individuals can build on each other's work and how technology can facilitate design by encouraging collaboration through shared designs.

Combining concept maps aids creative problem solving. In two studies, participants were presented with a story describing a problem and its solution. They were then asked to read an analogous problem and provide a solution. In order to facilitate transfer from the first story to the second, concept maps of the stories were elicited and participants instructed to combine them. The results show that participants were more likely to solve the problem when they produced and combined concept maps, as compared to a condition in which concept maps were used without combination, and compared to a condition in which no maps were used, but instead summaries were created and combined.

A sketch combination system is introduced and tested: a crowd of 1047 participated in an iterative process of design, evaluation and combination. Specifically, participants in a crowdsourcing marketplace sketched chairs for children. One crowd created a first generation of chairs, and then successive crowds created new generations by combining the chairs made by previous crowds. Other participants evaluated the chairs. The crowd judged the chairs from the third generation more creative than those from the first generation. An analysis of the design evolution shows that participants inherited and modified presented features, and also added new features. These findings suggest that crowd based design processes may be effective, and point the way toward computer-human interactions that might further encourage crowd creativity.