Newsroom

Merage Professor Introduces a Groundbreaking ‘Random Walk’ Model to Boost Creativity

October 23, 2024 • By UC Irvine Paul Merage School of Business

In the evolving innovation landscape, creativity stands at the core of progress across disciplines, ranging from science and engineering to the arts. Professors Vidyanand Choudhary of the UCI Merage School of Business and Shai Vardi of Purdue University’s Daniels School of Business have crafted a pioneering approach to understanding and enhancing creativity.

Their new model, “A Random Walk Modeling Framework for Boosting the Creativity of Humans and AI,” draws inspiration from computer science, physics, and neuroscience. This novel framework provides insights into how creative ideas are generated, with the potential to help AI and humans boost their creative output.

 

The Inspiration Behind the Model: Combining Semantic Networks and Brownian Motion

The collaboration between Choudhary and Vardi stemmed from a shared interest in creativity and their backgrounds in graph theory, a mathematical structure used to model relationships between objects. At a conference last year, the two scholars discussed how creativity, often viewed as abstract and difficult to quantify, could be examined through the lens of computer science. This conversation led them to look at existing literature on creativity, particularly about how neuroscience and psychology have studied the human brain as a connected graph of ideas.

“In neuroscience, the brain is thought of as a network of nodes that talk to each other,” Choudhary says. “Similarly, psychologists have mapped creativity through semantic networks, where words and ideas are linked in meaningful ways.” For example, when someone hears the word “phone,” their brain might immediately associate it with the word “communication.” This connection forms a semantic network, illustrating how ideas are related to one another.

The duo realized they could combine this notion of semantic networks with a concept from physics known as Brownian motion, which describes random movements of particles, to model creativity. “Our research talks about creativity as a random walk,” Choudhary says. “By walking randomly through a network of ideas, we can trace paths that lead to novel and potentially creative thoughts.” This idea forms the crux of their model: Creativity arises from moving between a network of nodes—or ideas—in unexpected ways.

 

How Creativity Functions as a Random Walk

The random walk model of creativity emphasizes that while humans may tend to follow familiar patterns of thought, it is often on the less-traveled paths where creativity emerges. For instance, if one starts with the idea of a “phone,” common associations such as “communication” or “charger” may come to mind. However, jumping to less obvious ideas, like “phone” to “paper” or the more abstract “phone” to “relationships,” may cause more creative connections to arise.

According to Choudhary, this leap from idea to idea is probabilistic. Some associations are more likely than others, but those rare, unexpected connections are often where creativity happens. “Creative ideas are just a little rarer,” he says. These rare ideas are not only novel but can also be highly useful.

 

Designing Interventions to Boost Creativity

Choudhary and Vardi’s work goes beyond explaining creativity. Their goal is to design interventions that can actively boost creative thinking. After they developed their abstract model, their research involved identifying techniques to enhance creativity in both humans and AI. “We want employees, artists, plumbers—everyone—to be more creative,” Choudhary says. To achieve this, they explored well-documented creativity-boosting strategies, or interventions, and also applied the random walk concept to design new interventions.

One such method involves constraining the creative process to force deeper exploration of less obvious ideas. For example, when asked to generate a verb for the noun “desk,” an individual may be limited to using verbs that start with the same letter as the noun. This constraint prevents them from defaulting to common associations and encourages them to dig deeper for creative alternatives.

 

Using AI in Creative Testing

AI played a pivotal role in testing these interventions. Choudhary and Vardi implemented a well-known method called the Verb Generation Task (VGT), in which AI was asked to come up with a creative word association. The researchers introduced novel elements, such as asking AI to consider absurd ideas like “space elephant,” while creating responses that significantly boosted the creativity of the AI. “All the answers AI came up with were way more creative,” Choudhary says. Imposing constraints and altering the context of tasks improved creativity not just in humans but also in AI systems like ChatGPT.

 

Testing the Model on Humans

After conceptualizing their model and testing it on AI, Choudhary and Vardi were asked to extend their experiments to humans. They discovered their new interventions and previously known interventions also worked to boost creativity. One of the methods they used was the Alternative Uses Task (AUT), where participants were asked to concoct multiple uses for a given object. Their model not only explained why traditional creativity-boosting interventions worked but also provided a road map for developing new methods.

This four-step process—creating an abstract model, validating existing interventions, proposing new interventions, and testing the model on both AI and humans—set the stage for future research and practical applications.

 

Implications for Creativity in Practice

Choudhary and Vardi’s research carries profound implications for industries that rely on creative problem-solving. By providing a structured model of creativity, their work offers companies a blueprint to develop their employees’ creative skills.

“As academics and researchers, our goal is to build one brick at a time,” Choudhary says. “It’s our hope that this research can be a useful brick that provides others the foundation to facilitate useful and inventive ideas to build upon. That’s my first hope. The second one is more practical in terms of industry: Thinking about creativity in this way helps managers understand what they need to do to nurture and improve creativity beyond hiring creative people. You can look to your employee base and teach them the skills that boost creativity.”

 

Inspiring the Future of Creativity Research

In the broader academic context, the researchers hope their model will serve as a foundation for future studies. “If people had a model, they could try to improve it or apply it in ways we haven’t thought of yet,” says Choudhary. Whether in academia or industry, this model presents a powerful tool to foster creativity, offering new methods to enhance the creative capacities of both AI systems and humans.

Choudhary and Vardi’s random walk model of creativity opens exciting possibilities for future research and practical applications. As creativity continues to play a central role in driving innovation, their groundbreaking work provides a crucial framework to understand and enhance this vital human and technological capability.