Non-Functional Requirements (NFRs) are often managed late, either during design or, more often, just at implementation. This trend results in higher costs and low-quality software. Our work studies mechanisms to support RNFs, based on the collaborative dynamics of a negotiation process, during the requirements construction activities As such, we created a collaborative strategy for the construction of NFRs. An evaluation was conducted using a tailored tool built to implement the proposed Negotiation-Collaboration mechanisms. The choice of requirements by various stakeholders, within the process of eliciting software requirements through consensus, is a complex problem. Representing and structuring group dynamics, which are composed of human actions, is a challenge. During the group decision-making process, a set of ideas is debated, not always clearly expressed and not always understood by everyone in the same way. The debate involves several profilesq of participants, with different and sometimes conflicting points of view. Such differences, however, are fundamental to the quality of the group decision, as the ideas are analyzed from various perspectives. A collaborative strategy for building RNFs and their computational support is proposed. The Negotiation-Collaboration approach is reused to work specifically with RNFs. It takes into account not only the construction of the RNFs themselves, but also the construction of their interdependencies. Such interdependencies between the requirements impact the decision on the RNFs to be built. The strategy evaluation was supported by software developed to support, at the same time, Negotiation-Collaboration mechanisms and RNFs specification activities. The software is asynchronous and geographically distributed, facilitating group communication, even with physically distant members. Three evaluative activities were carried out, and the results showed positive indications for the use of the strategy in the construction of RNFs. For each of the projects used in the evaluations, a list of RNFs was produced, whose consistency was judged by the participants involved to be sufficiently satisfactory. The number of RNFs was duplicated in the three construction activities, revealing greater coverage regarding the quality attributes initially listed for the software.