Harnessing Generative AI for Problem Identification and Selection in Science and Innovation

In a rapidly evolving technological landscape, the intersection of artificial intelligence (AI) and problem identification in science and innovation has become increasingly crucial. At this year’s Academy of Management (AOM) conference, we organized a Professional Development Workshop (PDW) to explore this topic. Our session was co-sponsored by several divisions: Technology and Innovation Management (TIM), Strategy (STR), Organization and Management Theory (OMT), and Communication, Digital Technology, and Organization (CTO). We gathered leading experts from different fields to discuss how generative AI can reshape the way we identify and select problems, i.e., how AI may impact the early stages of science and innovation projects. Our distinguished speakers included: Markus Baer, Carolin Häussler, Hyunjin Kim, Dashun Wang, Ola Henfridsson, and Frank Piller.

AI’s Role in Shaping Mental Models

One of the most thought-provoking discussions centered on AI’s ability to challenge and shape the mental models that managers use. These mental models may not only shape how managers identify and solve problems but also whether and how managers employ AI in problem identification. As such, rigid mental models may constrain the opportunities arising from using AI for problem identification and selection. One roundtable discussed how AI could act as a catalyst, disrupting entrenched thinking patterns and enhancing problem-solving capabilities in complex scenarios. Alternatively, participants raised the idea that alignment between managers’ mental models and AI can also be achieved by changing how AI operates or frames the outputs it produces. 

Proactively Detecting Latent Problems

Another critical insight from the PDW was the potential for AI-driven tools to proactively detect latent or emerging problems that might escape human attention. Similar to biomarkers in medicine, these tools could identify issues before they fully materialize, enabling organizations to address them proactively. This raises important questions about how to integrate AI into existing innovation management practices, ensuring that AI-generated insights are translated into timely and effective problem resolution.

Addressing Bias in Problem Identification

Bias in problem identification is a well-known challenge in the innovation process. Experts often fall into the “experience trap” or suffer from local search bias, which may lead them to overlook novel ideas. Our discussions considered whether AI could mitigate these biases. While AI has the potential to reduce human biases, the conversation also acknowledged that AI itself is not free from bias, depending on how it is trained and applied. This duality requires careful consideration in the deployment of AI tools.

AI as a Simulated Expert

The concept of using generative AI to simulate expertise across different fields emerged as another exciting avenue for innovation. This approach, discussed in another roundtable, could enable more comprehensive problem framing, particularly for complex, ill-structured issues. By simulating expert input, AI could help avoid the “jumping-in” bias, where innovators prematurely focus on solutions without fully understanding the problem.

Sustaining AI At Scale

As organizations increasingly adopt AI, the challenge of scaling these technologies becomes evident. One roundtable focused on the organizational hurdles of scaling AI, such as data fragmentation and local integration. Many more hurdles were identified through participant ideation and discussions with GenAI. The discussion underscored the need to balance flexibility with standardization to achieve long-term benefits of using GenAI in organizations.

Looking Ahead: The Future of AI in Science and Innovation

The PDW also offered participants a sneak peek into Dashun Wang’s innovative tool, SciSciGPT, which promises to be a game-changer in the field of Science of Science. This tool, still under wraps, is expected to spark significant interest and discussion among scholars.

Experimenting With AI Tools for Problem Identification During the PDW

Beyond the inspiring pitches from our speakers, it was particularly fun and insightful to experiment with GenAI tools for problem identification and selection in our own research during our roundtable discussions. We discovered that AI’s effectiveness in these tasks is influenced by several factors, including time constraints, varying levels of AI proficiency, the choice of tools, and more. Given these complexities, it was unsurprising that participants found their inter- and transdisciplinary roundtable discussions more effective than relying on AI alone for developing strong problem statements. This experience underscores the importance of blending AI with human expertise to achieve the best results in research and innovation.

As organizers of the PDW, we extend our gratitude to all participants for their enthusiastic engagement. As we move forward, it’s clear that generative AI will play a pivotal role in shaping the future of problem identification and selection in science and innovation. Let’s continue this important conversation!