To solve complex problems brainstorming with a Group Support System can be a helpful tool. Social comparison and association effects are factors that stimulate the generation of creative and high quality ideas or concepts. Also, the GSS’ capabilities that allow a group to work in parallel contributes to the swiftness with which a fast growing set of ideas is generated. Groups of 10 to 15 are able to produce 100 to 150 ideas in as little as 15 minutes. The set of generated ideas however also has some limitations. Typically the set of ideas is characterized by redundancy, ambiguity, off-topic ideas and a lack of shared understanding. Extracting the key ideas from such a large set of ideas is time consuming and easily cause the facilitator and participants to suffer from cognitive overload. Extracting the key ideas is a process in which the group uses a combination of selecting and summarizing ideas and uses clarification techniques to create shared understanding. To address these challenges, groups and facilitators can benefit from methods and techniques to effectively extract the key ideas from the brainstormed list of ideas, without losing any promising ideas. Such techniques and methods are typically referred to as convergence methods. In interviews with professional facilitators we indeed found that they considered convergence to be difficult and time consuming. Also a large body of literature, mainly GSS case studies, was found that describes hurdles to convergence. These hurdles include (1) information overload at the beginning of a convergence task and (2) the cognitive effort required to complete a convergence task. Therefore the main research question of this thesis is: ‘how can convergence processes become more successful and effective?’. Based on an analysis of the current set of methods for convergence we identified four opportunities to improve the successfulness and effectiveness of a convergence process. The methods included in this study originate from the ThinkLet library, the method database of the International Association of Facilitators (IAF) and from a literature review. The identified opportunities are: Removing the task of detecting redundant concepts away from the facilitator to lower his/her workload. Improving the current hurdles that exist when converging in a parallel way, as is done is the FocusBuilder thinkLet among others. The current limitations of this thinkLet include: o Lack of comprehensiveness of the end result o Inability for the facilitator to monitor the process Creating a scalable and fast pre-selection method. Improving support for inexperienced facilitators to manage a convergence process in a large group. Classifying and comparing the methods found was possible by using a classification scheme based on two axes, (1) the output of a method and (2) the way of working implied by a method. By using the example of a creative problem solving workshop, in which we tried to find a match between a method for convergence and a convergence task for different scenarios, we have identified the opportunities for improvement. The differences in the scenarios were in the number of participants and facilitator skill level. In response to these opportunities we have designed three artefacts. One new thinkLet, Divide&Conquer, is developed that enables large groups to quickly make a pre selection of concepts. Secondly we designed a modifier to the FocusBuilder thinkLet. This thinkLet supports the creation of shared understanding, achieves a reduction of the concepts under consideration and removes redundant concepts from a brainstorm artefact in a scalable and fast way. Thirdly a technique for similarity detection, normally used for plagiarism detection and automatic grading of written texts, is adapted and evaluated for use to detect redundant concepts in a brainstorm or convergence artefact. The technique uses normalized vector representations of concepts based on a thesaurus to detect similar concepts. To assess the effectiveness and success of the designed artefacts the following process and result oriented metrics are used; process oriented: acceptance, satisfaction, facilitator dependence, scalability, commitment, productivity and efficiency. Result oriented: speed, redundancy, reduction, refinement, comprehensiveness, shared understanding (ambiguity), satisfaction and commitment. Evaluation in groups of the technique for similarity detection, the new Divide&Conquer thinkLet and the modified FocusBuilder thinkLet revealed that: Even with a moderate detection rate of 50% participants are able to remove redundant concepts faster than participants that did not use the artefact in which concepts were ordered according to the automatic detected redundancies. Evaluation however is limited to one case study. Further evaluation is needed to validate the results and research the use of similarity detection within the new and other thinkLets. The Divide&Conquer thinkLet can be used within groups to quickly make a pre-selection of concepts that the group deems worthy of paying further attention to. The process and results of the thinkLet were accepted by the participants of two workshops, however the process needs thorough explanation before the start to reach agreement on the process. The thinkLet achieves a pre-selection quicker than other pre-selection methods because in principal less votes than the number of participants are collected per concept. Based on the average value and standard deviation it is decided whether more votes per concept are needed. This increases speed and therefore scalability of the pre-selection process. The pre-selections made in the evaluation workshops with this thinkLet contained only on-topic items and reduced the original brainstorm artefact by 50% on average with a standard deviation of 10%. Besides explaining the process and presenting the results no facilitator efforts are required. The modified FocusBuilder thinkLet can be used on a brainstorm artefact directly or after a pre selection has been made. The thinkLet fosters the creation of shared understanding and achieves a (further) reduction in the number of concepts under consideration by removing and summarizing redundant concepts and removing off-topic concepts. The thinkLet uses sub groups of participants that work on sub sets of concepts in parallel and convergence is achieved in three or four rounds. In previous case studies the comprehensiveness of the end result was too low. We removed the first round from the thinkLet, in which the participants work alone, to limit participant bias. Evaluation revealed that the comprehensiveness of the end result increased, without changing any other values that already were positive. Because of the parallel way of working the thinkLet is fast and scalable. Facilitator interventions are needed to explain the process and to present the end result, the real convergence effort is executed by the participants, therefore facilitator dependence of this thinkLet is low. The inability for the facilitator to monitor the process also is an opportunity for improvement of this thinkLet. A design for this is described, but is not evaluated. The outcome of this project is relevant for every professional interested in efficient collaboration within his project team, business unit or organisation. But also for practitioners, facilitators and collaboration engineers, because it proposes solutions for the time consuming step of convergence in GSS supported meetings. Further effort is however needed to evaluate the performance of the two thinkLets within more workshops, but the results indicate that the field of evaluation can be extended to organizations and professionals. Further research is needed to improve the accuracy of detection of redundant concepts and to integrate the detection technique within the two thinkLets mentioned.