For multiple attribute decision making analysis, general fuzzy measures that only fulfill the boundary conditions and monotonicity can be used to express the grades of subjective importance of attributes, and the human subjective evaluation process can be accurately approximated because of full degrees of freedom. However, it is difficult to collect complete information regarding all attribute subsets in actual applications for fuzzy measure identification. Thus, this study aims at reducing the information demand of general fuzzy measures. We proposed sufficient information to cover all importance information of attribute subsets. Sufficient information can reduce the information demand by at least one half. Based on the framework of sufficient information, a sampling procedure under partial information was further proposed, in the hope that most information could be captured with the least number of samples. The comparisons among complete, sufficient, and partial information indicate that partial information works effectively in practice and significantly reduces information demand. Finally, we presented an implementation procedure for questionnaire investigation under the sampling design of partial information, and gave an example with illustrations.