Friday, 10 January 2014

What Can We Learn From IBM Watson's Clinical Judgement?

On the day that IBM launches it's new business unit, IBM Watson Group, I became inspired, for the second time, by the possibilities and opportunities for applying this technology to tangible problems facing us today - making our planet, a "Smarter Planet", if you will.

The first time occurred as I followed Watson's progress with Jeopardy!, and I was stunned that it was not only fast and accurate; but also gave a degree of confidence in it's answers.  As someone working on a predictive analytics project for a learning organisation, this could be so useful when applying future anticipated insight to a business decision-maker wanting to estimate the results of a training course.  Since it's gameshow success, things went a little quiet as IBM built upon their success, developing Watson's commercial viability in a protective and secretive research environment in Austin, Texas.  Quiet, that is, until this morning - the second time I was inspired - following this 112-minute video: 2014 IBM Watson Special Event.


Ginni Rometty, President and CEO of IBM, said that when Watson won Jeopardy!, "I'm not sure many people really understood what was happening behind the scenes"; Mike Rhodin, leader of the new IBM Business Unit, spent 10 minutes outlining the approach Watson takes to learning: "It learns like our children learn...  It reads.  It asks questions.  If it encounters [conflict], it has to sort that out."

As various luminaries from the world of healthcare explained how they saw Watson's technology being applied to the industry, this sparked in me an interesting question: if Watson is going to deal with conflict in the world of medicine - a world fraught with ethical dilemmas, updating standards and emotional agendas - how will we know that Watson is making the correct judgements?  In other words, how can we be sure that Watson will learn correctly?


Mr Rhodin suggested a methodology in his earlier overview: "When our children come home from school, how do we know our children are learning?  We ask them questions and we check their answers.  If the answer's not right we help them to discover the right answer."  Sounds good - especially to a parent with 2 young children approaching school age!  Ok.  It builds algorithms on algorithms and, in this way, continues to learn.

However, in an industry with truly exploding data, an industry where Watson will know more than any human (i.e. be able to assimilate new research and evidence faster, and more efficiently, than any doctor or academic), who will be the "authority" to keep Watson in check?  Jay Katzen, President of Elsevier Clinical Solutions, told us "I'd have to read 174 articles, every single day, just to keep up!"

Despite what I tell my daughter, no parents are all-knowing (except, perhaps, my wife!) and most of us can probably think of a scenario where we have heard a parent giving advice to a child, with which we wouldn't agree.  Whether it's right or wrong makes no difference, of course: because often there is no right or wrong - just alternative approaches.  How will Watson make those kinds of judgements?  For example, will it be able to apply a weighting to it's billions of datapoints to promote or restrict sentimental, financial or political considerations?  (And would it be correct to do so?)

Machine Learning is not enough

Daniel Hillis put it best, half-way through the film, when he espoused a vision of machines and humans interacting to mutual benefit and advancement.  If cognitive science is about understanding how learning takes place, and cognitive computing about machines learning to learn, then what we should be aiming at is "cognitive collaboration".

Greg Satell (among others) beat me to coining this term, evidenced by his Digital Tonto blog entry 'The New Era of Cognitive Collaboration' where he discussed - what else?  - IBM Watson!  (It seems Watson really is breaking new boundaries, as all roads seem to lead back to it.)

What does this mean for learning, performance management and strategy?

If learning can be automated, is there still a place for Learning Professionals in the corporate world?  Absolutely!  Expertise, judgement and timing come together to create a critical mix of skills that make learning professionals the authority in this area.  In my organisation, we have seen a growth in learning datapoints of c.100%, over 3 years, as we assimilate more informal learning and reconcile that with greater appreciation of the value of learning in the 21st Century business.  Super-charged abilities to tackle big questions, like that presented by Big Data, mean that Watson will certainly not be under-utilised as data grows, whilst structure and formality subsides.

Technology is getting very smart indeed, and may be getting better able to assist us in our strategic direction.  Data Scientists and Business Analysts will ensure that the reports generated are as accurate and insightful as possible (one form of authority, perhaps); meaning that the learning professional may finally be able to rid themselves, once and for all, of the transactional and administrative tasks related to keeping the engine turning over and, instead, focus on the business of empowering the organisation with the correct skills, as required.  The crucial point, here, is that the learning organisation needs to be communicating strongly with the data community, to ensure the implementation of sound judgement and close the authority loop.

It seems, then, that the authority defining the direction of an organisation may remain the domain of the human for some considerable time to come

Tuesday, 29 May 2012

Lessons From Vienna: Is The Waltz Over For Classroom Training?

Vienna is warm in May: the parks bare witness to couples relaxing, some lost in their own thoughts as they contemplate the rivulets meandering by, whilst others rush past in a bid to keep pace with an ever-quickening world.  

Whilst attending a recent conference in the beautiful Austrian city, I challenged a colleague to meet me at 7.30am for a run in the park opposite the hotel.  When I arrived, he was wearing long cargo shorts, a heavy metal t-shirt and the type of trainers you may expect to better suit skateboarding: we shook hands and congratulated each other on not standing-up the other - no mean feat, considering the late nights and red wine! Contrasting with my running shoes, rugby shorts and lightweight t-shirt, I felt confident that this would be a good way to start the day; but the smug feeling of superficial superiority was soon to be smashed, as my colleague casually apologised for his attire - and that he wouldn't have dreamt of attempting his last half-marathon in such clothes.  Gah!
[...Pant!  Pant!  Cough!  Gah!]

It's the kind of jolting, affirming experience that isn't often available, virtually.  The running partner in question is, in fact, our lead facilitator for Live Virtual Classroom events and, in his spare time, leads much of our research into virtual learning media as a whole so, as we trundled along in the morning sun, he seems the perfect metaphore for a quandary which is facing many learning organisations today:  What role does F2F (face-to-face) learning hold in the 21st Century?

On the one hand, my colleague is a proponent of a futuristic view of learning: he has worked with, and produced points of view, on such tools as Saba Centra, Blackboard Collaborate, Sametime Unyte, Lotus Live, as well as virtual world experiences such as SecondLife and Unisfair Intercall.  On the other hand, here we are: running together during a 3-day face-to-face event in the beautiful, warm Summer sunshine of Central Europe - discussing some new ideas for the measurement of impact achieved by means of virtual learning.  The irony of the situation doesn't seem immediately evident to either of us.

Earlier in the event, I'd had the opportunity to rub shoulders with other colleagues, whom previously, I'd never met: including many of my own team!  Reflecting the multiculturalism common of many internal service providers, currently, my team consists of staff from India, Slovakia, Belgium and the UK; I'd met a couple of them previously, but for the majority, this was the first time of meeting - one of whom I'd managed, virtually, for over 4 years.  We had time to learn more about each other and to get out whiteboards and flipcharts and work through some of the tactical issues of the moment - both activities which seem to be more effective in person...  I was able to assess posture and body language when speaking and, perhaps as a result, able to comprehend a much greater percentage of their communicative effort - which, in turn, made all of us feel less stupid than at other points in our relationship.  So much value in the workplace comes from having the confidence to raise your hand and communicate your idea, that it's invaluable to build this capability (and I think this is contrary to the approach many take - hiding their linguistic insecurities behind Instant Messenger texts or emails).

'Fearlessness' (example) is a key theme running through the event: we enjoyed a wonderful day facilitated by Sue Liburd who, with sensitivity, warmth and passion, helped us to face up to the negative experiences in our minds, and replace them with positive stories of success.  The experiences in our break-out exercises were (at the same time) emotional, humbling, inspiring, personal and professional... Without exception, delegates emerged from that day feeling more positive, confident, understood and able to get their messages across.  But I can't help thinking that we could not have got anywhere near as engaged, via the User Interface on my laptop...

On the last day, as a group, we revisited some of the predictions we had made about our industry - and our own service offering - 5 years earlier.  On the face of it, some sounded wacky (heavy reliance upon virtual worlds for learning, see above), some sounded already dated (mobile learning); but we were pleased to have come so far in such a short space of time.  Common predictions for the future did not seem too unreasonable (greater saturation of learning via mobile devices, greater use of gaming to support learning) and some sounded mundane by comparison (better content management and linking of IP assets; better structuring of learning data to drive insight).  But, as I re-visit the day-to-day tasks and To Do lists that I'd managed to escape, the paradox about how we shall best deploy learning in the future is brought into sharp focus: I find that I have a report to produce on the evolution of learning media.  
Mobile learning, itself, has to evolve and keep pace with the speed of change!
5 years ago, we were eagerly anticipating greater dependance upon less formal, bite-sized learning on mobile phones.  With budgets shrinking, and especially with the effects of the contracted global economy, this kind of light-touch learning has brought our own organisation to deploy up to 70% of it's training content via eLearning platforms.  However, improved technology and analytical models have gone a long way to counteracting this trend: as, increasingly, learning organisations find more robust and innovative ways to demonstrate the value they provide to the skills of a workforce (and, therefore, to the bottom-line in the chart of accounts). CFOs and CHROs have been loosening the reigns on the impoverished learning organisations, and F2F is back on the agenda!  

Why?  Because certain types of learning depend much more heavily upon the ability of the individual learner to make meaningful connections.  Yes, I can learn many things via eLearning (from how to create a pivot table, to the cultural norms and traditions of my team-members offshore); but I cannot yet receive the kind of emotional bonds, real-life experiences, or simple conversations that make a learning environment complete.