On the Effects of Complementary and Substitutive Use of Generative Artificial Intelligence in Innovation Processes

This dissertation investigates how substitutive (automation) and complementary (augmentation) modes of generative artificial intelligence (AI) integration shape human behavior and performance within the two fundamental building blocks of innovation: learning and ideation. Through field studies and laboratory experiments in educational settings, our research demonstrates that while complementary use - utilizing Large Language Models (LLMs) as personal tutors - improves deep comprehension, substitutive use via automated solution generation increases task volume at the cost of actual understanding, ultimately widening inequality by harming learners with less prior knowledge. In the domain of design ideation, our work reveals that frontloading manual sketching induces design fixation and reduces idea variance, yet paradoxically results in significantly higher design quality by anchoring focus and mitigating AI-induced off-task distractions. Synthesizing these findings, this dissertation reveals that the specific mode of integration is the primary determinant of innovation outcomes, showing that misplaced substitution erodes essential mental effort while strategic complementarity reinforces the user engagement necessary for sustained performance. These insights yield actionable design principles for innovation managers, advocating for a new philosophy of human-AI collaboration that prioritizes safeguarding human cognitive involvement to realize the full potential of generative technology.