The woman behind the desk checked, and then with a straight face informed me that there had been 'technical difficulties' at Amsterdam Schipol, and that my bag would not now arrive until tomorrow evening. She asked for the name of my hotel and told me it would be delivered directly to my room. All fine and good, but there was me standing there in my jeans and sneakers, and my best suit and shirt were in my luggage. In Amsterdam. Worse, my keynote was scheduled for the following morning, which left me in something of a dilemma. To say I was furious with the airline would have been an understatement. The plane I disembarked at Amsterdam was exactly the same plane I got into again to fly onward to Cologne. I recognised the crew. I got off the plane, trotted a mile or more across Schipol Airport and then got back on to exactly the same plane, but in the meantime my bag had been removed and left who knew where.
And so I arrived at my hotel, checked in to my room and then proceeded to tweet my problem to anyone on Twitter who cared to read it. I named and shamed KLM, and then went off to find something to eat. An hour later, to my surprise, KLM responded to me on Twitter, apologising for the mix up and advising me that I should go and purchase whatever I needed, and they would foot the bill. Wonderful. Clutching my credit card, I went off and bought a new pair of shoes, two new shirts, underwear, socks, shaving kit and toiletries. I stopped short of purchasing an expensive new suit. I was wearing a serviceable jacket and anyway, KLM would probably only increase the airfares to compensate if I blew another 1000 Euros on a Ted Baker original.
The keynote went well and my luggage duly arrived the following evening. But how did KLM know to respond to my tweet? Answer - they were scanning for mentions of KLM on Twitter and other social media. This is known as sentiment tracking, a method that may well come in useful in education in the future. I'll give you some examples of how it's used now and how it works...
The Twitter example above is a very primitive form of Sentiment Tracking and Analysis (also known as opinion mining). It simply involves a KLM staff member regularly scanning the popular social media channels to intervene if there is any bad publicity or complaint, before it blows up into something unmanageable. Several tools are available for sentiment tracking on Twitter and other social media channels. Sentiment tracking is becoming much more sophisticated. Many large business do this now, because they want to know what is being said about their brand. They know that a complaint in a public forum can have a highly negative impact on their business if it's not dealt with quicky. But sentiment tracking can also be harnessed positively by businesses. Recently I wanted to buy some black, Italian hand made slip on shoes. I visited one or two online stores, and then without purchasing, I went off to do other things. An hour later, I searched on Google for some e-learning blogs, and landed on my first page. There at the bottom of the Blogger website this advert was staring back at me:
How did the system know how to target me? The online store (Amazon) had logged my IP address, and my interest in that specific product, and the fact that I had not purchased. It had probably sent a cookie. It assumed from this that I must still be interested. At the next available opportunity, Amazon targeted me with an advert through Google Ads via Blogger. The same applies when you mention something on Facebook, or simply let slip your date of birth, location or other personal information such as hobbies and interests. Before you know it, Facebook is pushing targeted advertising to your page, and it's highly effective. Facebook logs dozens of different items of personal data from your actions every time you visit, tag a photo, post a new status update or 'like' someone's comment.
I noticed the following three adverts on my Facebook page just now: You will notice that Facebook knows I am in the UK. It knows a lot more about me than that though. The last advert is because Facebook knows I am a Manchester United fan - that little detail is there in my profile somewhere. The middle advert is because it knows I am a guitarist, again from information in my personal profile. The first advert? I'm not sure why the first is there, because I have never let it be known that I wish to illuminate something 200 metres away from me. Perhaps someone else can shed some light on this. It's not in my profile that I like to bother pilots as they land their jet airliners, or that I have aspirations to be a covert operative for MI6. Sometimes sentiment tracking gets it wrong, and sometimes it just takes a wild punt and hopes for the best, a bit like playing Internet Battleships. But it could be a lot worse. Facebook might decide to send me links to a mature women dating site, or a wholesale Viagra dealer, just for a laugh. That would be hard to explain. Sentiment tracking is usually quite accurate though, picking up on your emotional statements, likes and dislikes, conversations, as well as links you have previously clicked. Sometimes it seems to take a random guess, as with the torch. But sometimes that guess can be disturbingly accurate.
How does sentiment tracking work? At the simplest level, the system uses Natural Language Processing techniques (NLP) to mine the words you type into your status updates or query boxes. At a deeper level, artificial intelligence applications capture the NLP data and process them into clusters that have collective meaning. A lot of modeling can be done with those kinds of data. Essentially, sentiment tracking makes sense of what you do on the web, and then transforms it into recommendations, actions or in this case, advertising. There are many problems with this kind of computation, including questions over how machines can differentiate between various emotional intensities, differentiate between polarities of opinion, or detect subjectivity in a statement. However, refinements in systems will continue to improve their accuracy.
When it comes down to group behaviour, sentiment tracking can be quite accurate. As we have demonstrated with our previous research into Technosocial Predictive Analytics (TPA), using a mashup of NLP, AI, GPS and geomapping, events such as flu epidemics and social movements can be tracked and even predicted quite accurately over geographical location and time. Have you ever shopped for a book on Amazon? You select your book and then Amazon displays a message saying something like '76 people who bought this book also bought...' and you suddenly realise that there's another book you didn't know about on a similar subject to your own purchase, and now you want that book too! It's a very effective marketing ploy, but there is also enormous educational potential. Amazon is using a form of crowd sourcing for its sentiment tracking, and is selling you a book you didn't know you wanted, based on the tacit approval of a cluster of people who are similar in their tastes, profiles or backgrounds to you. In effect, the individual acts of buying books, combined, create a desire line - a slime trail of social enzymes if you will - that can be mapped and recommended to future purchasers of similar products.
Clearly there are opportunities to harness the power of these methods in education. Imagine students being directed to new and highly useful content they were previously unaware of. Imagine new content being created automatically on the basis of the actions of like minded scholars in dispersed locations. Imagine content being changed and updated automatically, based on the activities of a global community of practice. Finally, imagine being able to track the actions, content creation and decision making of your groups of learners, and mapping these onto information graphics to track their collective and individual progress, knowing when to intervene and when to let them alone. This kind of learner analytics (or educational data mining) will emerge from the collective intelligence of crowd sourcing and the sentiment tracking of individual actions and behaviour. The technology already exists. We now have to determine whether we want this capability in education, and if we do, we next have to ask what will be the ethical, pedagogical and social implications?
In the next blog post: How Google is refining your web search
Photo by David Sky
Other images by Steve Wheeler
Tracking sentiments by Steve Wheeler is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License.