Large Language Model Based Brainstorming (LBB) for Idea Generation and Evaluation

主講人 : 孫河山 教授

Brainstorming has become a widely adopted technique for idea generation within organizations. However, traditional in-person brainstorming sessions are subject to several well-documented limitations such as production blocking, social anxiety, social loafing, and bias against nonconforming ideas. Recent advancements in large language models (LLMs) offer promising avenues for reimagining and enhancing the brainstorming process. In this research, I aim to achieve new theoretical understanding of how LLM- d brainstorming (LBB) simulates human brainstorming (in conversational and nominal sessions) for idea generation. Hypotheses were developed  d on the recent breakthroughs in few-shot learning and chain-of-thought prompting. To empirically test these hypotheses, I conducted a series of simulated brainstorming sessions utilizing GPT-4o agents. The output from these sessions was assessed using both an LLM- d evaluator agent and human raters. The findings largely support both hypotheses and reveal several noteworthy insights with implications for both research and practices.

主講人簡(jiǎn)介：

孫河山教授目前任職于美國(guó)俄克拉荷馬大學(xué)普萊斯商學(xué)院信息管理系統(tǒng)系。他的研究聚焦于信息技術(shù)對(duì)個(gè)人、組織和社會(huì)的深刻影響和相互作用，具體包括人機(jī)交互、商業(yè)分析、在線/數(shù)字行為等。他在MIS Quarterly（5篇，其中獨(dú)作發(fā)表2篇）、Information Systems Research（4篇）、Journal of the Association for Information Systems、Decision Support Systems、International Journal of Human-Computer Studies、Journal of the American Society for Information Science and Technology等信息管理領(lǐng)域的國(guó)際頂級(jí)學(xué)術(shù)期刊發(fā)表了多篇學(xué)術(shù)論文，在2018-2022年“Worldwide on the Top IS Researcher List”中排名第45位。他目前是Information Systems Research的Editorial Review Board Member，并擔(dān)任MIS quarterly、Journal of the Association for Information Systems和AIS Transactions on HCI的Senior Editor。