I’ve just been reading a fascinating new doctoral dissertation by a recent Ph.D. from Washington University’s Olin School of Business: Chihmao Hsieh, now a professor at the University 0f Missouri. He asks a fascinating question about innovation: what is the structure of successful new ideas? If that sounds abstract, I’ll try to explain by drawing on some of the research from my book GROUP GENIUS.
First, creativity researchers know that all new ideas are combinations of existing ideas. That’s why psychological studies of “conceptual combination” are so important–what happens when people combine words like “pony” and “chair” and what they say a “pony chair” looks like.
Second, innovations today are always combinations of previous ideas, products that are already out there, being used every day.
So the question “What is the structure of good ideas” is really a question about these combinations. If only there were some way to evaluate the combinations that form new ideas–perhaps we could count up the number of previous ideas that the new idea is based on? Perhaps we could measure how those new ideas are related to each other?
Hsieh’s dissertation is exciting because he has figured out a way to measure the relatedness of all of the component ideas in a new idea. He started with a database of all patents granted between 1975 and 1999. For each patent, this database indicates which previous patents are cited, and which later patents cite it. Using this huge database, and applying some fancy mathematical tools, it’s possible to calculate the relatedness of any two patents (essentially, you count how many of the same other patents that both of these patents cite). Then, for each patent, Hsieh figured out the overall relatedness of all of the patents that it cited. The key question is: which level of internal relatedness is most likely to result in a successful patent?
He measured the success of a patent by counting how many times later patents cite it, and here’s what he found: the most successful patents have an intermediate degree of relatedness. And he found something else interesting: patents that cite more other patents are more successful.
I’m excited by this research because it reinforces what psychologists have discovered about creativity: good ideas involve certain kinds of combinations among existing concepts and ideas. But it’s very hard to look inside the brain and observe these connections in action. Hsieh’s clever use of patent data provides us with a window into what successful innovation looks like. The key lesson for all of us is to learn as much as you can about the ideas that are already out there; that’s the only way you’ll be able to come up with that unique new combination that no one else has ever thought of.
3 thoughts on “The Structure of Good Ideas”
I would say that there is an additional type of combination: combining an impossible old idea with a newly emerging enabler. When language first developed the natural instinct was to use gestures and pictures either literally or metaphorically. Picture languages such as Chinese calligraphy developed when only paper and ink were available. Because there were no computers at the time, the awkwardness of picture creation forced the idea of communication by picture into a very abstract and difficult form. The idea lost out to competition with phonetic languages. A person’s own internal brain communication as exemplified by his dream language is a picture and metaphor language. If ever there were to be a “thought” language it would seem it would be most analogous to a picture language. Icons are moving in that direction but are still too clumsy and non-standardized. The idea of communicating by picture or thought is awaiting an enabling technology, just as Leonardo’s helicopter design had to await the invention of a motor to power it. Perhaps “a good idea in waiting” would be the “impossible idea” combined with a list of future requirements almost like Archimedes’ “give me a lever and a place to stand and I’ll move the Earth”… a little grand at the time… So the ancient scribe might have said, “give me a painting machine, and I will tell you what I’m thinking.” I think the written word should consist of “words” that are animated icons but more like small dream vignettes with standardized meanings in a form much more accessible than is a Chinese character(ideograph)
Of course, again thanks for the post on this… The research is now published in Scientometrics. The paper ultimately changed a bit, but the basic idea still aligns with your blog post.
The paper can be found here: http://www.springerlink.com/content/y545661440001817/
Congratulations, that is great news! I will certainly read this final version of your paper.