Interaction between agents representing real world actors in computational models must be informed by knowledge about interaction processes occurring amongst real world actors. We propose the use of endorsements to implement the cognitive processes underlying the decisions that lead to interactions among agents. The main advantage in applying the idea of endorsements lies in the fact that they allow for combining the efficiency properties of numerical measures with the richness and subtleties of non-numerical measures of interest or belief. We demonstrate the expediency of our approach with two evidence-driven agent-based models. From these case studies we derive suggestions for suitable extensions of the endorsement concept.