Google Flu Trends - using signals to predict a real outcome
Today we were at the IBM Smart Work Summit in Melbourne
As we continue the conversation about HiveMind and the concept of mining the signals in an organisation to work out what people's expertise really is, one common question is "but can it work"? Peter Sheahan had the key note speech and in it raised a great example from Google which demonstrates clearly how the concept of mining hidden signals can deliver real world value.This comes from Google's Flu Trends site, what it shows is that by mining the common search terms related to cold and flus, they can (very) accurately track flu occurences based on searches for flu symptoms. Not only this, it's a leading indicator over the data from Doctors which takes time to aggregate. Similarly, we believe that by mining a broader range of signals within an organisation, for example what people blog about, e-mail about, contribute and comment on, check-in and perhaps yes, even search for in your Enterprise Search solution, we can mine a clear profile of their expertise.
I've been fascinated by these Hidden Signals in organisations for a while now. In some ways the original inspiration for HiveMind was this Tweet back in January 2009 where I tweeted the following:A good layoff indicator is LinkedIn, my network activity is way up - 1st action when laid off is to connect and seek recommendations. http://twitter.com/timbull/status/1159905073
The key take away was this message:
Total minutes spent on the site doubled in January to 96.8 million, from 47.6 million in December.Part of what is driving all the activity is people looking for job, and helping friends who are out of work. Recommendations are up 65 percent since December, says spokesperson Kay Luo.
I then created a more detailed response http://timbull.com/hidden-signals-in-social-software that posed the following question:
Most of us are already well aware that we leave a footprint on any social networking site, and in some instances, may come to regret that photo we posted on facebook. What I find more fascinating however is how these sites are generating new signals and measures that over time could be used to perhaps supplement or replace existing messages. One person updating there resume is not newsworthy - everyone updating it sends a strong message about job security. Good systems are those which can make visible these hidden messages and signals. What hidden signals are you observing in other systems which provide insight (of any kind)?
With HiveMind we are building a tool to help mine these signals from organisations and turn them into expertise. What's your thoughts? What signals are there inside your organisation that send messages on people interests and abilities, if only there was something listening for them?
Comments (0)
Posted 8 months ago
by Tim Bull
