In this paper, we propose a collaborative decision fusion framework when the participants of the decision making process are human agents. We consider a binary hypothesis testing problem in which a group of n people makes individual decisions on which hypothesis is true based on their own knowledge and perception. The observations at individual agents are assumed to be corrupted by a common as well as independent noise signals. Local decisions made at individual agents are sent to a moderator (fusion center) to make the final decision. We assume that the moderator has imperfect knowledge about the thresholds used by the individual decision makers and model them as random variables. The fusion performance in terms of the probability of error at the moderator is derived when the exact realizations of the individual thresholds of the agents are not available. With two human agents in the group, we derive the performance of the likelihood ratio based decision fusion strategy. For an arbitrary number of human agents n(> 2), we derive performance of decision fusion with majority rule using certain approximations.