Abstract: Based on the expectation value， deviation degree and hesitancy degree of the probabilistic linguistic judgment information， a new multi-granular probability linguistic distance measure was proposed to solve multi-attribute decision-making problems， where the attribute values were demonstrated in multi-granularity probabilistic linguistic information， and the attribute weights were unknown. The new distance measure could effectively overcome the difficulty that the existing distance measure cannot exactly measure the distance in some circumstances. Based on the new distance measure， a generalized TODIM method was proposed by comprehensively considering the quality and quantity of the judgement information and the consensus measurement and was applied in the selection of waste sorting and recycling APP. The comparison results with existing methods demonstrated that the proposed distance measure in the multi-granular probabilistic linguistic environment had good effectiveness， which also verified the feasibility and superiority of the consensus measure-based generalized TODIM method for multi-granular probabilistic linguistic information.