HBS Professor Says AI Can Boost But Not Replace Human Creativity in HAA Webinar

Jacqueline N. Lane, a professor at the Harvard Business School, argued that artificial intelligence can meaningfully amplify human creativity, but cautioned overreliance on the technology during a webinar on Thursday evening.

During the event, hosted by the Office of the Vice Provost for Advances in Learning and the Harvard Alumni Association, Lane highlighted how AI has redefined what creativity means.

“AI is forcing us to fundamentally rethink this question of what makes you creative, and this answer has profound implications for how we work, how we innovate, and how we build companies,” she said.

Lane, who leads Harvard’s Laboratory for Innovation Science, stressed that AI works best when it “informs but does not replace our human decisions.” She noted that models are particularly helpful to help users weigh competing ideas, but not create them.

Throughout the hour-long talk, Lane spotlighted three studies of her team at LIS focused on measuring AI’s creative ability. The studies focused on whether AI could match human idea generation, judge human creativity, and help startups grow.

To test AI’s creativity, Lane and her team partnered with a gig-work platform called Freelancer to crowdsource business ideas on the circular economy and had human experts evaluate them alongside ideas generated by ChatGPT. The researchers found that there was “no difference” between the creativity of human and AI outputs, but there was a difference in how innovative and impactful the ideas were.

“Our most innovative idea that came back from our crowd sourcing challenge, and that happened to be a human generated idea,” Lane said.

“However, when it came to looking at the most valuable ideas, in terms of those ideas that were most likely to create a financial impact and an environmental impact, those ideas came from AI,” she added.

Lane noted that if AI is able to generate ideas, the large volume of ideas could create a “bottleneck” for evaluating these ideas. To study AI’s ability to assist humans in evaluating ideas, her team partnered with MIT Solve — an entrepreneurship platform that runs innovation challenges.

Lane’s team gave some of the judges of an MIT Solve competition AI recommendations with detailed explanations, others a “black box” condition where AI gave recommendations with no explanations, and a control sample which received no AI input at all. The researchers found that while AI assistance improved the quality of selected ideas compared to the control sample, the detailed explanations caused overreliance.

“Having these justifications that accompany the recommendations, didn’t necessarily lead to better thinking, but it did lead to more deference,” she said.

“They can actually backfire when we are trying to make choices, because we may use them as these kinds of persuasive shortcuts as opposed to exercising our independent thinking,” she added.

The final study Lane presented during the talk spotlighted Focus Fuel, a caffeine gummy startup founded by three entrepreneurs with “no investors, no employees — just them and ChatGPT.”

AI enabled the company to launch in a crowded category and report $1.5 million in first-year revenue.

“AI helped them understand formulation, supply chain strategy, retail negotiations, regulatory compliance,” Lane said. “AI was able to help them bridge expertise gaps in the way that they were able to think at a level that normally requires years of experience.”

Lane ended the talk by saying that despite advances in AI’s creative abilities, human input remains valuable in the creative space.

“The final judgment must stay human, and we can’t let those narratives, or those confidence sounding narratives, really replace our critical thinking,” she said