Cognitive Activation or Design Dependency? The Dual Mechanisms of Large Language Models in Shaping Individual Design Quality Across Tasks of Varying Complexity

Abstract

Abstract

With the widespread integration of large language models (LLMs) in the field of design innovation, the mechanisms through which they influence individual designers’ performance remain unclear. Existing studies primarily focus on the macro-level impact of LLMs on creative diversity, while systematic investigations into their micro-level effects on individual cognitive processes in specific design task contexts are lacking. Based on Cognitive Load Theory and Collaborative Cognitive Load Theory, this study proposes a dual-opposing-mechanism model to explain how LLMs affect individual design quality, introducing task complexity as a moderating variable to systematically examine the differential effects of LLMs across varying task contexts. In Experiment 1 (N = 216), a 2 (assistance type: LLM-assisted vs. human-assisted) × 2 (task complexity: simple vs. complex) between-subjects design was employed. Results indicated that for simple design tasks, LLM assistance significantly enhanced individual design quality through the activation of cognitive associations (cognitive activation pathway); conversely, for complex design tasks, LLM assistance led to a decrease in design quality due to induced design dependency (design dependency pathway). Experiment 2 (N = 216) further examined the effect of constraining LLM outputs, revealing that in complex tasks, constrained LLM responses effectively mitigated design dependency and improved design quality, whereas in simple tasks, such constraints weakened the cognitive activation effect, thereby reducing design quality. These findings elucidate the dual-opposing mechanisms of LLMs in design innovation contexts and provide theoretical foundations and practical guidance for differentiated deployment of LLM tools in design education and practice.

Keywords: large language models; design quality; cognitive activation; design dependency; task complexity; cognitive load theory