Abstract: During multi-attribute decision making, it is difficult to assign weights directly for all attributes, and decision makers therefore prefer offering the relevant information for part of alternatives. How to use partly the alternative preference information to make decision is studied, of which the basic process is to use G1-method to convert the language information into data information. Then, a programming model is developed to approach the idea about decision and, by solving the model, the optimum weights are obtained. The method is very simple to use and embodies the integration of instinctive decision making with scientific decision making. An example is given to illustrate the effectiveness of the proposed method.