In a session during the Cleveland Clinic's A.I. Summit for Healthcare Professionals, Brian Uzzi, PhD, posed an interesting question to the audience: Do you think bots can perform better, the same, or worse than you on a creativity test?

Many attendees of the session noted that they expected the AI to perform better than an individual on the test and were greatly surprised when they were then proven wrong.

"AI, when used in certain ways, dampens our creativity and our innovative thinking. Lots of [research has] shown that the impact of AI is making scientists more productive—they publish more papers, but at the same time, those papers become more and more narrow and copies of one another. So, it expands what you’re publishing, but it doesn't improve the quality. It narrows ideas. It also limits critical thinking," Dr. Uzzi explained.

Dr. Uzzi is the Richard L. Thomas Professor of Leadership and Organizational Change; Co-Director, Northwestern Institute on Complex Systems (NICO); Co-Director, Ryan Institute on Complexity; Professor of Industrial Engineering and Management Sciences, McCormick School (Courtesy); and Professor of Sociology, Weinberg Northwestern University.

 

Dr. Uzzi walked the participants through an exercise to demonstrate the impact of AI on creativity and prove that AI cannot replace human creativity. After the exercise he explained how human–AI partnerships can be optimized to combine the strengths of both.

 

 

Dr. Uzzi introduced the group to a creativity simulation called Beat the Bot where they were asked to come up with 10 words as far apart from each other as possible within 4 minutes. The further apart the words were, the higher the individual’s resulting creativity score was.

 

After the 4 minutes were up, he showed the group their combined average: 80.4, as well as the highest score: 94. He noted that scores between 95 and 100 were representative of creative geniuses such as Leonardo da Vinci.

 

Next, a chosen model (25 options) completes the same test. "In our surveys, about 80% of people say the bot is going to crush my score," he noted. "After all, the bot has got access to all recorded human knowledge, it potentially has a much bigger vocabulary than I do, it can search more quickly, etc."

 

When this phase of the test was completed, the models had achieved an average score of 79. "On average, the bot [achieves about] the average of people, in part because it's trained on the data that we're using as well," Dr. Uzzi commented.

"What the bot does is bring up the floor," he explained further. "One of the regularities we see in bots is bots will raise the floor of human ability, but it will push down the ceiling of human ability. And for society, it's people in the ceiling, it's the geniuses among us, that tend to have a disproportionate effect on moving society forward. They're the ones that come up with the surprising solutions to grand challenges."

 

The same test is then performed in an "augmented pair" with the help of AI, where AI provides 10 baseline words and the individual can then alter as many of those words as desired to try to come up with more divergent answers.

 

Participants again expected the scores to improve with the addition of AI; however, Dr. Uzzi noted that in the majority of cases, the scores go down with the augmented pair. "It goes down for two basic reasons: people get anchored to the words the bot gives them. They scan for common words and change 'car' to 'trolley' and 'ocean' to 'puddle.' The other thing is in real-life, when people struggle for 4 minutes for the first test, then when they get to this test, they look at the words and work for just 1 minute. It creates laziness and the scores go down."

 

Next, the session participants were divided into groups to work as small teams to complete the same test again. Dr. Uzzi noted that scores almost always rose as a result of the teamwork.

 

The final step in the process is to take the test as a team again with the addition of AI. But again, he explained that the teams tend to get anchored to the baseline words suggested by the AI model and rush through the test, making the scores drop again.

 

 

"Don’t ask bots for answers. If you ask for an answer, you short-circuit your own thinking, you don’t add your own secret sauce—the thing that makes you special," Dr. Uzzi emphasized. "And, you’re also less likely to learn. Any answers people get from a bot are forgotten almost immediately."

 

"What is the best way to use a bot? Ask the bot for a process that you can follow that can help you get to the best outcome," he urged.

 

Dr. Uzzi explained that AI can be used to break apart the constraints humans have for truly divergent thinking and tackle process to ultimately think in new ways.

He concluded that AI should be used as a "sparring partner" rather than a collaborator to challenge your thinking and push individuals toward new creative solutions.