Currently, the design of computational systems requires considering both technical aspects and the impact of autonomous decisions made by these systems. That is, given the significant progress of artificial intelligence, the choice of actions of an intelligent system must be appropriately aligned with the system of human values, both individual and social. This paper proposes using natural language processing techniques to extract human values from action descriptions. The model is proposed as an extension of a value-based agent architecture whose decision process is based on the agent’s value system and the value promotion scheme of each available action. For the testing of the model, the decision process on participatory projects has been simulated. Given a set of projects, the proposed model was used to detect the values each promotes, and then the agent’s decision process was simulated. The results, although preliminary, show that the proposed extension of the architecture has a behaviour consistent with the theory of values used as a basis.