An integrated assessment (IA) model combines knowledge from very different disciplines in view of a practical problem. Most models developed so far are rather monolithic in the sense that it is difficult to combine components from different models for purposes of new assessments. We propose to develop a modular approach to IA based on advances in knowledge management as well as in object oriented software engineering. The incentive structure of modular IA is based on turning the knowledge produced neither into public nor into private, but rather into club goods. Competition amongst modelers becomes a process of discovery at the level of module design and module coupling, with strong synergies between competing teams. Together they develop a community pool of IAM-modules, along with software and know-how for running them in varying combinations.