In a dissertation, Ziyi Zhao examines how people and digital technologies can work together to improve problem-solving and innovation. In one study, Zhao combined randomized experiments with detailed screen recordings of programmers using ChatGPT or Stack Overflow. Programmers considered human sources more credible but rated identical GenAI-attributed solutions as higher quality. The strongest results came from programmers who both modified and verified AI-generated code, achieving a 75% success rate, while users who repeatedly revised prompts without testing the results often became trapped in unproductive cycles. A second study analyzed more than 695,000 daily observations from mobile apps and found that expanding from iOS to Android produced stronger and more persistent quality improvements when developers used cross-platform software development kits with shared codebases. Together, the studies show that technology improves knowledge work when people create feedback-rich processes that turn interaction into learning, rather than relying on automation alone.
Collaborative Intelligence and Innovation in Knowledge Work
