Web Decision Making-Hunch.com and Caterina Fake (Flickr)

Posted on August 8, 2010 by

After exploring some recent efforts at data mining in medicine in this blog, I yeared for more info about the use of data mining in social media generally. This isn’t easy to come by if you are not in this arena. I did find some relevant writings on a fairly new site called Hunch and its co-founder Caterina Fake, who has focused on helping people find what they want on the net, essentially using crowd souring to make better decisions fast. This is a pregant field. Making swifter more precise and better decisions based on individualized wants and interests is a critical function, not just in consumerism but in all business. It is also essential in medicine. So, Hunch has captured my interest. Caterina Fake/Hunch first gets people talking about themselves, their opinions, tastes, beliefs, idiosynchrasies and once they have sufficient data they mine it for correlations to provide cutomized recommendations for the user. (the correlation search seems similar to that being done by Andre Brin and his wife at 123&me–see my recent blogs inculding the data mining of blood data  from prior clinical studies for therapeutics in Parkinson’s). Hunch learns interesting things about us. For example people who believe in alien abductions are more likely to drink Pepsi.(okay what else?) People who eat fresh fruit more likely to want to buy the Canon’s EOS 7D camera. On her site, Caterina describes the site as a decision making site customized for you. The Hunch site explains that researchers have establlished that decisions made by diverse and independent groups are superior to decisions made by individuals, including experts. The reason is that knowledge is spread among many people. In choosing what to ask us, Hunch’s question selection algorith (MIT developed), tries to ask questions which will help optimize and rank recommendation outcomes. It creates decisions trees.It is the people who answer questions that train the algorithm proper training to make nuanced recommendations. Now I am beginning to understand the concept of machine learning and searching among data for correlations you might not be able to find without the computer. I’m not sure how the Hunch algorithym knows what questions to ask to get at the important intuitive info but I am on the hunt for this. I am also about to try out Hunch to seek what areas of of strength it has. My goal is to keep an eye on companies like this for the lateral benefits they might have towards the development of medical therapeutics. If you are in this field, please comment. There is also an article in WIRED Mag. on Hunch but it is quite general. There are things I really like about the goals of this company–first–Hunch wants to establish via proof that it can do better in this process than other companies. For example Match.com uses the process to connect people. If the company evolves, it should be able to do so but better. Same with Amazon and books. From what I see, Hunch does not try to tell people what they want but rather to help to define within what they want how to understand what might be better for them based on those wants/needs. Perhaps it will expand to help people define wants and needs for a better more fruitful life.

Leave a Reply