| Problem solving groups are often composed of members from different knowledge backgrounds. Previous research has demonstrated mixed results with some studies showing advantages for more cognitively diverse groups, and others showing disadvantages for diversity. Two theories were developed and tested. The first, group synergy theory, was based on the idea that more cognitively diverse groups have a broader knowledge base that will inspire the generation of more problem solutions, and therefore perform at a higher level than less diverse groups. The second theory, common ground, was based on the idea that cognitively homogeneous groups establish better common problem representations and therefore perform better. Two experiments were conducted to investigate these competing theories. The first experiment compared the performance of homogeneous and diverse dyadic problem solving groups, while solving binomial probability problems. Homogeneous dyads were composed of members who both received the same type of training lesson, so that each member received either a procedurally-based or conceptually-based training lesson. The diverse dyads were composed of members who each received different training lessons. In order to better control for demographic variables, and to more easily record dialogs, the dyad members collaboratively worked on problems while located in physically separate workspaces, using a virtual chat environment as a medium to jointly develop problem solutions. The homogeneous groups performed more accurately on standard problems, solved the problem sets faster, and reported a better working relationship. The diverse groups performed more accurately on transfer problems, and spent proportionately more communications on generating problem solutions. The results suggest that cognitively homogeneous groups perform better on routine problems, while cognitively diverse groups perform better on unusual or non-routine problems. The second experiment was a replication of the first, except that participants solved the problems as individuals. Participants who received both training lessons performed more accurately on standard problems, while participants who received only one training lesson performed more accurately on transfer problems. The responses of the individual problem solvers were then used to create a combined with other individuals in order to create a nominal dyad model by which the actual dyad performance was then compared. |
