City Research Online

Abstract

The role of generative Artificial Intelligence (GenAI) in early-stage innovation processes, specifically problem-finding (PF) and idea generation, remains largely under-explored. Historically, AI's involvement in these processes has been constrained by perceived shortcomings in reasoning, creativity, and intuition—qualities deemed essential yet uniquely human. However, the emergence of GenAI—characterized by advanced reasoning, enriched creativity, sophisticated multimodal learning, and human-like conversational capabilities—calls for a re-evaluation. This dissertation comprises three papers that collectively enhance our understanding of GenAI’s role in these innovation stages.

The first paper conceptually examines the role of Large Language Models (LLMs)—a subset of GenAI—in PF for ill-structured, wicked, and multi-stakeholder problems, a crucial yet understudied area. Drawing on Simon’s cognitive-behavioural perspective, it identifies key PF activities and skills, assessing how structured human-LLM collaboration can enhance performance. Despite humans’ innate strengths, cognitive limitations such as bounded rationality, satisficing, and uncertainty avoidance constrain their effectiveness. LLMs potentially address these constraints by broadening the search space, providing alternative problem framings, and augmenting creativity. However, given LLMs’ inherent limitations—including biases, hallucinations, and difficulty handling complex problems—the paper advocates structured human-LLM interactions, illustrating a proposed collaboration framework through two case studies in product development and social innovation.

The second paper empirically investigates how consumer interactions with GenAI shape the novelty and appropriateness of product ideas under constraints, using the creative cognition lens. Three experiments compare GenAI-assisted and human-only settings under input-resource and financial constraints. Findings indicate that merely introducing GenAI does not automatically enhance creativity; instead, outcomes critically depend on task distribution between humans and GenAI. Assigning idea generation primarily to GenAI and exploration to humans significantly improves appropriateness with a marginally significant increase in novelty. Additionally, covertly guiding GenAI toward transformational creativity enhances novelty without sacrificing appropriateness, though at the expense of financial value. Both conditions, however, reduce participants’ own idea generation, raising concerns about over-reliance on AI. A third experiment reveals that GenAI mitigates choice overload among high novelty-seekers and eases constraint-induced difficulties for low novelty-seekers, narrowing the creativity gap between these groups.

Building upon these insights, the third paper argues that creative self-efficacy (CSE) inadequately captures collaborative creativity involving humans and GenAI. It introduces a new concept, co-efficacy, and validates creative co-efficacy (CCE)—a construct representing the belief in joint human–GenAI creative capabilities—emphasising perceived joint efficacy in collaborative creative outcomes. Across three studies, it develops a reliable three-item CCE scale, identifies antecedents (positive AI experiences, prompting skills, cognitive flexibility), and experimentally compares the relative effects of CCE and CSE in constrained creative tasks. Results suggest that higher CCE may surpass CSE in predicting creative interactions, with higher CCE scores prompting consumers to request more ideas from GenAI, thus enhancing creative outcomes. However, effective human-GenAI collaboration structures remain critical.

Collectively, this dissertation synthesizes PF literature, extends the cognitive-behavioral perspective to less structured problems by proposing a structured human-LLM collaboration framework, advances creative cognition research within human-GenAI interaction, and introduces and validates CCE as distinct from CSE, with broad implications for personal and organizational creativity.

Publication Type:	Thesis (Doctoral)
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science Q Science > QA Mathematics > QA76 Computer software
Departments:	Bayes Business School > Bayes Business School Doctoral Theses Bayes Business School > Faculty of Management Doctoral Theses

Export

Downloads