We use the synergetics to study how to allocate input probability in the binary symmetric discrete channel in order to maximize average mutual information for memoryless and memory channels with interference. We obtained the result that for the memoryless channel, the maximal average mutual information is in accordance with that the equivalent input probability can give according to the information theory. Moreover, we also studied the binary discrete memory channel with interference (noise), taking into account the input and output symbols having different memory degree and interference factor, and finally obtained the symbol optimized input probability. This a new method for studying information theory is presented.