Trust in Multi-Agent Systems can be used to measure uncertainty, not only of interactions between agents but of a number of sources. For each of these sources, a different trust model might be necessary. The first step to deal with uncertainties with the help of trust is therefore to find the different sources. A source of uncertainty can be any stochastic process in the system or in the environment as well as other agents or components in the system that are not under the control of the system developer.

If data originating from a source of uncertainty is used in the decision-making processes and it is possible, to check the data for correctness, trust values can be used to deal with the uncertain information it provides. As part of this task, the sources that provide uncertain information are identified. Based on the requirements documents and the current system architecture, the analyst and the developer—if necessary in accordance with the relevant stakeholders—check the different processes in the system for their behaviour. If a process can potentially yield uncertain data, it is added to a preliminary list of sources of uncertainty. The behaviour should then be investigated further and the potential for mitigation with a trust model should be evaluated. The main criteria is that trust models are only applicable if the past behaviour of the process is an indication of its future behaviour.

If a source of uncertainty has been identified that warrants a treatment with a trust model, the necessity to create such a model can either be captured in the requirements documents (e.g., by introducing new goals in a goal-driven approach or writing the appropriate user stories) or captured in a separate document that can then be used as the input for the next steps.