How Does the GenAI Usage Patterns of Postgraduates Affect Their Scientific Research Innovation Ability?

Generative artificial intelligence (GenAI) has been deeply embedded in the entire process of postgraduates' scientific research, exerting dual effects of empowerment and alienation on their scientific research innovation ability. However, the existing studies have not yet revealed the differentiated impacts of diverse GenAI usage patterns on scientific research innovation ability. Therefore, this paper used the questionnaires of 1252 postgraduates in the three provinces of Northeast China as samples, identified the potential subtypes of GenAI use via latent profile analysis, compared the differences in scientific research innovation ability among different subtypes by utilizing the BCH method, and examined the relative mediating effect of scientific research self-efficacy in the relationship between GenAI use pattern and scientific research innovation ability based on Bootstrap method. The results showed that the GenAI usage of postgraduates was characterized by high frequency of behavior, emotional moderate dependence, and cognitive prudence, and the influence directions of each dimension on scientific research self-efficacy and scientific research innovation ability were different. Their GenAI usage patterns can be divided into four categories: contradictory, rational, conservative and risky, among which the contradictory type accounted for the highest proportion, followed by the rational type came second, while the conservative and risky types were relatively fewer. There are significant differences in scientific research innovation ability among the four types of GenAI usage patterns, with the risk model scoring the lowest and the rational model the highest. Scientific research self-efficacy played a significant mediating role between GenAI usage pattern and scientific research innovation ability, and the rational type achieved the maximum gain in innovation ability by enhancing scientific research confidence. Based on this, the paper suggested that colleges and universities should work together from four aspects: guiding scientific research behaviors, strengthening emotional management, reconstructing technical cognition and optimizing institutional ecology, so as to help postgraduates achieve the balance between technical instrumental rationality and academic autonomy, and promote the sustainable development of postgraduates' scientific research innovation ability in the era of artificial intelligence.