A popular meme suggests that one of the hardest things about being an adult is figuring out what to cook for dinner night after night for the rest of your life. How do you throw together a dish where the ingredients work together while keeping things creative enough that boredom doesn’t set in?
That's the problem Oded Netzer, Arthur J. Samberg Professor of Business at Columbia University — along with Sibel Sozuer, assistant professor of marketing at UNC and a previous doctoral student at Columbia Business School, and Kriste Krstovski, associate research scientist at the Data Science Institute at Columbia — set out to tackle.
The team analyzed 57,000 recipes from Allrecipes.com, looking at ingredient combinations, how often recipes were tried, and how highly they were rated to get a sense of what ingredients typically work together and what happens when unusual ingredients are included. The researchers encountered a paradox: novel recipes were less likely to be tried, but once people tried them, they tended to prefer them — as long as the recipes didn’t deviate too far from more familiar ones. These findings point to a lesson that applies to creativity beyond the kitchen: the most successful ideas tend to strike a balance between novelty and coherence. They’re surprising without being so far outside our frame of reference that we can't understand them.
“Consider that funny jokes work not because they are way out there, but because they are pretty close to reality with a small twist,” Netzer says. That observation backs up the paper’s central finding, suggesting that coherence appears to be a critical component of successful creative products.
The science behind a good recipe
The recipes from Allrecipes.com gave the team two important sets of data: the frequency with which ingredients appear in different types of dishes and the combination of ingredients in each recipe. The team then used a large language model to determine how predictable an ingredient might be given the other ingredients in a recipe. For instance, how likely are tomatoes or basil to appear in a pasta dish, compared to an unusual ingredient like oranges? The more likely ingredients were to appear together, the more coherent the recipe.
The researchers could then compare patterns of coherence with whether people tried and enjoyed the recipes. They found that recipes need to be coherent to receive positive feedback. And they found that unusual ingredients, even if they’re a good culinary fit, can make people less likely to try new recipes.
“This is a unique role for novelty,” Netzer says. “The whole literature on creativity almost always talks about novelty being good, and finding an instance where it isn’t was a surprising result.”
The finding may come down to uncertainty. Before trying something unfamiliar, people don't know whether they'll like it. But after they've tasted the dish, that uncertainty disappears. A chocolate chip cookie with cardamom might sound questionable to someone encountering the combination for the first time. But if they try it and discover that the flavors work together, the novel ingredient suddenly becomes appealing. Accordingly, the team found that once people tried a recipe with a novel ingredient, that novelty appeared to be an advantage, as long as the novel ingredient fit well with the existing ingredients, and these recipes tended to be highly rated.
Creativity beyond the kitchen
The researchers argue that the principles they found governing coherence and novelty can apply to creative products of all kinds — particularly those in which individual elements work together to create a coherent whole, such as paintings, items of clothing, or fragrances. The key is that it’s okay to introduce something unusual if it remains connected to the other elements.
And these concepts can be applied even more broadly, according to Netzer. Consider a health app that combines a running tracker with music that responds to a user's heart rate. The combination could be novel, but it still makes intuitive sense because running, music, and heart rate already have conceptual connections, making it likely that users will accept such an innovation.
These ideas could even shape how companies roll out new products. A meal-kit service, for example, might initially offer recipes with familiar ingredients to encourage customers to try them. Once customers have built trust in the service, it could gradually introduce more novel combinations.
The goal, Netzer says, should be to find the unexpected combination that feels like it was meant to be there all along. “You want people to feel they know what you’re talking about,” Netzer says. “As long as you do that, you can take them in a direction they didn’t expect.”
