In the new conceptual framework of computational psychiatry, psychiatric dysfunctions are seen as problems in information processing of environmental situations. This new framework has three implications which I will explore in this talk. (1) Diagnosis should align to information processing changes (computational failure modes) rather than symptom clusters or physical circuit phenomena. (2) Treatment should require modifications of that information processing, either through modification of the patient themself (via physical changes such as pharmacology or electrical or magnetic stimulation or via learning and training paradigms) or through modification of the environment the patient experiences. (3) Translation across species requires computational parallels such that both species are doing the same information processing in the same manner, which will facilitate animal models for understanding and treatment. I will explore these implications with concrete examples.