The Hidden Cost of Human-First Collaboration: Human-AI Collaboration Sequence and Creativity in the Idea Elaboration Phase

Abstract: As artificial intelligence (AI) becomes increasingly embedded in creative work, organizations face a critical question: how human-AI collaboration should be structured to maximize creativity. Prior research has largely favored a human-first collaboration sequence, arguing that premature exposure to AI-generated content constrains independent thinking and fosters cognitive inertia. However, given the multi-stage nature of the creative process, it remains unclear whether the benefits of a human-first sequence generalizes across different stages. Drawing on the anchoring effect, we argue that in the idea elaboration phase, a human-first collaboration sequence can undermine creativity. Specifically, when individuals elaborate on ideas they initially generated themselves, they develop stronger psychological ownership over those ideas, which increases resistance to subsequent AI-generated structural suggestions. As a result, creators become less willing to incorporate AI input that could improve idea articulateness, ultimately reducing overall creativity. We further propose that trust in AI mitigates this effect by increasing individuals’ receptiveness to AI-generated suggestions.
We tested our hypotheses across three complementary experiments involving diverse creative tasks and samples (N = 741). Study 1 recruited 230 full-time employees to complete a product design elaboration task. Study 2 was a field experiment involving 271 participants who developed marketing plans for newly launched coffee products, including both human-first and AI-first collaboration conditions as well as a no-AI control condition. Study 3, involving 240 full-time employees, replicates the findings while directly testing psychological ownership as the underlying mechanism.
Across the three studies, we find consistent support for our hypotheses. Contrary to the prevailing assumption that human-first collaboration is universally beneficial, we find that in idea elaboration tasks, a human-first collaboration sequence reduces creativity relative to an AI-first sequence. This effect occurs because a human-first sequence heightens psychological ownership over initial ideas, leading individuals to discount or resist AI-generated structural suggestions, which in turn inhibits improvements in idea articulateness and ultimately reduces overall creativity. Moreover, trust in AI attenuates these negative effects by increasing willingness to integrate AI input during idea elaboration. 
This research contributes to the literatures on human-AI collaboration and creativity in three ways. First, we challenge the prevailing view that a human-first collaboration sequence is inherently advantageous by demonstrating that its effectiveness is contingent on the creative phase. Second, our findings uncover a cognitive mechanism explaining why collaboration may remain suboptimal even when AI is incorporated into creative work. Third, by shifting attention from whether organizations should use AI to how and when AI should be integrated into creative work, this research provides new insights into how organizations can more effectively harness the complementary strengths of humans and AI in creative problem solving.