Among the symptoms of schizophrenia, deficits in the recognition of intention is one of the most studied. However, there is no cognitive model of intention recognition that takes into account both innate and environmental/developmental factors. This work proposes a developmental model of intention recognition based on a neural network. This model enables us to emulate different types of impairment. Particularly, the dopamine hypothesis of schizophrenia is simulated through an impairment of the visual saliency, and environmental influence of the behavior of the caregiver is evaluated.