<p>Community detection in social networks aims to group users into clusters or modules. These clusters are defined in such a way that the internal connections within clusters are dense and strong, while the connections between clusters are weaker. In community detection, influential nodes (seeds) may exhibit a higher number of connections both within communities (clusters) and between communities. Therefore, identifying influential nodes that have the ability to establish more connections than regular users can lead to the formation of clusters with higher modularity. In other words, the presence of influential nodes, which have a greater impact on the structure of communities, creates the necessary conditions for forming clusters with high modularity characteristics. In this paper, a novel method for community detection based on influence maximization using collective intelligence, named SICDIM, is proposed. In this method, the Manta Ray Foraging Optimisation (MRFO) algorithm is first used to solve the influence maximization problem in order to identify influential and independent nodes whose neighbors do not overlap. Then, the Penguin Search Optimisation Algorithm (PeSOA) is utilized for community detection from the clusters of users connected to influential nodes. This method improves the performance of community detection on several well-known datasets, including Karate, Dolphins, PolBooks, and Football, with modularity (Q) values of 0.2570, 0.33, 0.4737, and 0.4692, and normalized mutual information (NMI) values of 0.5589, 0.398, 0.6228, and 0.3904, respectively. Moreover, the SICDIM method outperforms the PeSOA algorithm in various aspects such as modularity, NMI, precision, and recall, particularly in community detection.</p>