Can big data yield big ideas? Blend novel and familiar, new study finds

The study authors, Olivier Toubia and Oded Netzer of the Columbia Business School, based their research on the concept that ideas are considered more creative when they have a good balance between novelty and familiarity. By analyzing and text mining thousands of previously generated ideas, they identified common patterns in ideas that are considered creative.

"You can think about idea generation as creating a recipe where you to choose the appropriate ingredients and mix them together" said Netzer. "By analyzing the text in a large number of ideas across different domains, we were able to link an idea's judged creativity to its set of 'ingredients.' We found that what makes an idea creative as judged by both consumers and firms' executives is a mix of ingredients (words) that includes a balance between words that commonly appear together (familiar combinations) and words that do not (novel combinations)."

Thus, a creative idea is not simply an idea that includes novel ingredients, but combinations of words that are novel when appearing together balanced with combinations of ingredients that are more familiar. For example, if one generates an idea for an app that would help people live healthier lives and includes the words "running," "counting" and "steps," this would not classify the idea as creative, as all three of these words are fairly frequently used together. However, including the word "calendar" as an additional ingredient would take the idea in a more creative direction, for example the creation of a calendar app in which you could track daily movement and accomplishments. Even though the word "calendar" may not be novel with respect to the topic of health apps in and of itself, its combination with "running," "counting," and "steps" is novel.

Using insights gained from this research, the authors have developed a tool that can help people come up with better ideas.

"The tool analyzes in real time word combinations included in the idea and recommends words that help the innovator improve her idea" says Netzer. "If the idea is too familiar the algorithm will offer words that would make it more novel. On the other hand, if the idea is too novel, it will offer words that will make the idea more familiar."

The authors have found that their tool helps people improve their ideas and generate more balanced and creative ideas. The tool appears more helpful in improving the novelty of ideas that have too much familiarity than in improving the familiarity of ideas that have too much novelty.

Toubia noted: "By leveraging tools from the world of Big Data such as text mining and semantic network analysis we were able, for the first time, to scientifically analyze a large set of ideas and identify successful patterns that can help people improve their ideas. In a way, our tool turns the computer into a "sous chef" that helps the user assemble the right ingredients to combine in order to form creative ideas."

While only time will tell if computerized tools will surpass humans in generating creative ideas, the present research pours some systematic analysis and science into a domain -- creativity and ideation -- that has been largely based on intuition and hunch.

To view the public tool developed by the authors, visit: http://newtopic.protoideation.org/