Thinking Outside the Box Inside the Chat Box - The Impact of Large Language Models on Innovation

Abstract

Innovation is a key driver in the pursuit of development for both organizations and society. The development of large language models (LLMs) is disrupting industries and redefining various tasks traditionally performed by humans, including how innovation could be conducted. To explore how the use of LLMs could affect innovation, an experiment with business students at a Swedish university was conducted. In the between-groups experiment, participants were tasked with improving a product to increase its customer value. In total, 40 students were divided between using an LLM, ChatGPT (GPT 4), as a creative tool to assist them with product improvement or being allocated to the control group. The results show that while the participants utilizing an LLM generated a higher number of ideas than the participants who did not, there were fewer unique ideas. The participants were then asked to select and refine their three best ideas. These innovations were then categorized into the Henderson and Clark (1990) framework of the four categories of innovation. The results show that the participants utilizing an LLM were less likely to produce architectural or radical innovations in comparison to the control group. The findings indicate that the utilization of an LLM risks leading to a hampered rather than enhanced creative capability, as the user becomes narrow-minded and more likely to suffer from component myopia. Furthermore, the findings indicate that the risks of utilizing an LLM as an innovative tool must be considered and mitigated to successfully implement LLMs in organizations.