Friday, 3 May 2013

Web 3.0 analytics, the possibilities

One of the great problems of using web based data to make predictions has been traditionally the web had a lot of readers and few writers.  With the rise of Web 2.0 that has ended.  We now have lots of writers as reading the web and creating the web with comments and tweets merge in to a single conversation, a conservation that it is pretty easy to overhear.

But the problem with Web 2.0 is that social engagement is not placed in to a context beyond just friends.  I chat with my friends about Chicago and US politics all the time, but most of us no longer live in Chicago and many of us don't even live in the US anymore.  What has been needed is an ability to confidently connected web buzz with real location, locations where economic, social and political processes are ground.

In come Web 3.0, which I define as a web linked to the real work by rich geo-spatial data.  This kind of data lets you know the locations of conversations.  Over the past year I have been exploring the potential of using open social data with geo-tags to make real world predictions.

2013 Bi-Election the rise of UK.

The UK Independence Party had never had a reputation of winning elections.  UK politics had been dominated by two major parties with the Liberal Democrats as the no vote, but in 2013 the Liberal Democrats were in the government and the UKIP part emerged as a right wing alternative to the Conservative senior party in power.

On May 2nd 2013 there were bi-elections in England.  As the elections took place I used the managing trendsmap tool to get a picture of the location and intensity of Twitter chat about the major parties.  The results were amazing.  Though the term labour was popular, and labour was the key opposition party, the UK was a buzz with chat about UKIP.  But the question was would buzz translate in to votes?  With the results coming in to the election it is clear that the twitter buzz was not just a few wonks chatting about something on twitter, the massive buzz on UKIP and the low buzz for Tory and LibDem memes reflected a real change in the populations voting behaviour.  The was a major right wing protest vote for UKIP taking place, and twitter reflected that.

Twitter memes mapped on the UK on May 2 2013 during an election show the intensity of support for UKIP all over England.

UK Near Triple Dip

In 2013 the UK economy emerged from a near triple dip recession.  This economic downturn grew out of a weak late 2012, especially weak consumer demand.  In 2012 during the Christmas shopping season I was tracking London high street Foursquare and Twitter traffic.  What I noticed then was the the quantity of tweets, and the content of tweets did not reflect a surge in consumer behaviour over normal periods, and that compared to Mall of America where a strong Christmas was taking place the London numbers and sentiment looked very poor.
5 days of Foursquare Checkins at Liberty in London before Christmas 2012, showing no sustained surge in checkins over the normal patterns.

Now it could have been that this was just cultural difference, with English people not interested in tweeting and checking in to stores as they shop.  Perhaps again the geeks glued to their mobile phones in the UK did not represent the consumer population.  And at this point I can't dismiss this, I don't know for sure if the people I track using social media are representative or not. But I do no that just as I predicted from twitter and foursquare data on the London High Street, 2012 had a weak consumer Christmas.

Work to be done

I think everyone can agree at this point that the activity on sites like twitter and foursquare are not just disconnected nose but does contain some data that is correlated to other social behaviours.  What is left for us social researchers to figure out is:
  • How representative is social media of the overall population, who is being over or under examined?
  • How do we effectively measure geo-social activity on the web?
  • How do we present this data so it can be used in a timely fashion?
  • How can we make more accurate and precise predictions based on social data?
  • What are the risks in using social data?

This is the work of the Web 3.0 Lab right now.  I feel strongly that I have developed some good qualitative methods for measuring the geo-social web.  My work now is the try and find how to capture enough solid data to make more precise predictions.  For example the UK avoided recession in 2013, my twitter data predicted a downturn but was utterly unable to estimate its size.  It turned out it was not a big enough slow down to be a recession.  As for the UKIP election results, I had no ability via twitter analysis to guess how large the UKIP vote would in fact be.

My study of much of this is still an art.  But I will keep working.

No comments:

Post a Comment