Abstract: Three-dimensional facial expression animation has been widely integrated into people's daily lives, but most existing expression generation techniques have not taken into account the end-user preferences. These methods mostly rely on pre-defined psychological archetypes, limiting their potential to generate rich and expressive facial expressions. Aiming at the above issues, this paper proposes a novel encoding scheme for facial expression animation individuals, and uses Interactive Genetic Algorithm (IGA) to achieve automatic evolution of facial expression animations. During decoding, the Dirichlet Free-Form Deformation (DFFD) algorithm is used to control the deformation of the entire facial mesh. By introducing human evaluation and generating the initial population randomly, the aforementioned issues are effectively resolved. Additionally, to maintain the diversity of the facial expression animation population, two mutation operators are designed: Uniform Mutation with Exclusion Zones introduces the concept of exclusion zones, Uniform Mutation with Shared Mutation Points considers the actual meaning of chromosomes and employs the same mutation points for gene segments that control the same parts. Finally, the experimental results show that these two operators effectively reduce the similarity within the population during the evolution process. The actual user participation experiments further validate the feasibility of the proposed approach in generating various facial expression animations that meet user preferences.