The critical process of personalized search is to reorder candidate documents of the current query based on the user’s historical behavior sequence. There are many types of information contained in user historical information sequence, such as queries, documents, and clicks. Most existing personalized search approaches concatenate these types of information to get an overall user representation, but they ignore the associations among them. We believe the associations of different information mentioned above are significant to personalized search. Based on a hierarchical transformer as base architecture, we design three auxiliary tasks to capture the associations of different information in user behavior sequence. Under the guidance of mutual information, we adjust the training loss, enabling our PSMIM model to better enhance the information representation in personalized search. Experimental results demonstrate that our proposed method outperforms some personalized search methods.