Sign Out
Logged In:
 
 
 
 
 

Archive: January 2016 (Features)


Tuesday, 26 Jan 2016
Rob Chidley

On the face of it, this seems a bit of a strange title for a talk to The OR Society but Kenneth Cukier, Data Editor of The Economist, who presented this year’s Blackett Memorial Lecture soon made it clear that it was actually a very appropriate title.

Traditionally, O.R. has been about identifying and understanding the problem, deciding what data is needed, collecting it and making sure it is as “clean” as possible then attempting to isolate the causes. Big data, on the other hand, is not particularly concerned with a specific problem and it does not matter if the data is messy provided there is lots of it. If any correlations can be discerned from the data, again, it may not be critical as to whether A causes B, A is caused by B or that A and B are actually independent. If the data suggests that people tend to buy more “Pop Tarts” just before the weather turns bad then it makes sense to stack the shelves when bad weather is forecasted: if they buy more, great; if not, it is no great loss. Sometimes it does not matter if there is no causal relationship.

In the lead up to the 2008 Presidential Election, the Obama campaign had to choose from three or four website homepages that they thought would be the most effective. As it turned out, the most informal one “won” – no one knows why but it need not matter. Importantly, the effectiveness of the different homepages was measurable: the best-performing image and link drew in $60 million more in contributions than the least performing one.

Where big data has proved immensely successful has been in areas involving machine learning. If you write an algorithm which tries to answer all of the questions there are very few cases when you will succeed. Even in a simple game such as checkers (draughts), if you code all the rules and all the moves you can think of, the chances are you will still be able to beat the machine but if you write an algorithm which allows the machine to work out the best play for itself and let it play itself hundreds, thousands, millions of times, you will have produced a truly formidable player.

Spell-checkers, grammar checkers and translators are other areas that greatly benefit from a big data approach. Microsoft used four different algorithms to carry out grammar checking. The one which performed the best when taught using a small sample of cases performed the worst with large data, in fact it only improved a little. But the one which performed the worst with small data (trained with half a million words), performed far better than the others when given large data (one billion words).

In another example, Kenneth explained how machine learning had identified twelve critical markers when looking for cancerous cells. The experts had previously only recognised nine of these. In the case of cars, researchers in Tokyo fitted a seat with hundreds of sensors to identify people via their posture. It may be used as an anti-theft device: whether the person sitting is the legal owner. If every car was so fitted then it might be possible to determine when a driver was falling asleep or not paying attention and perhaps be used to avoid potential accidents. He also recognised the vast potential the Internet of Things could provide, most of which we have not yet thought about.

He finished on a more cautionary note. With machine learning we do not know how the rules have been derived or, indeed what those rules are. Not only that, but they are constantly changing as more experience is gained. So, for example, if the algorithm used by VW had been derived via machine learning would VW still be to blame? If a pedestrian is accidentally killed by a driverless car who would be responsible? If a train hits someone, unless the driver has broken the rules, there is usually no case to answer and we accept that, but trains do not have the ability to swerve, brake or take any other avoiding action. With the IoT, maybe the car will be able to alert the pedestrian via their mobile phone that they were so engrossed in when they stepped off the pavement!

Legislation covering data, data ownership, data protection, machine learning and this whole area is seriously in need of a major review.

This article originally appeared in the January 2016 edition of Inside O.R.

January 2016



Monday, 4 Jan 2016
Rob Chidley

2016 could not have got off to a better start for The OR Society as we learned that our incoming president Ruth Kaufman will be awarded an OBE for services to Operational Research.

Ruth Kaufman OBE
When did you find out you were in line for this award?
Just a few weeks ago, a letter arrived out of the blue saying, in rather more flowery language, words to the effect of: ‘The Prime Minister proposes to recommend that The Queen approves your appointment as an Officer of the Order of the British Empire.’

How did you react? Given that I am a republican opposed to most of what this Prime Minister is doing, my first reaction was probably unprintable. My second reaction was ‘why me?’ Other people have done a lot more than I have. It took a while for me to realise that it is, literally, an honour that somebody out there thought it worth the effort of nominating me and pulling together all the supporting evidence. Once the news had sunk in, I realised that it’s not just gratifying for me personally, but also it will be really helpful in my push, as the new President of The OR Society, to increase the visibility and profile of OR.

What does ‘services to Operational Research’ mean? Good question! I don’t know exactly what they had in mind because, as the recipient, I don’t get to see the 10-page form that the nominator had to submit. But as head of an OR group in a government department, one-time chair of the Government OR Service, member of The OR Society Board for 7 years, and now OR Society President, I have persistently tried to do two things: to look internally, to improve the quality and effectiveness of OR practice and governance; and to look outwardly, to extend OR influence across professions or disciplines, across application areas, between OR commissioners and OR providers.

I’ve been involved with setting up Pro Bono OR, with building the ‘Making an Impact’ practitioner sessions at conferences, with raising the profile of OR in government, with representing OR at Executive Board level in a government department – every one of these has been a result of teamwork, but I’m very proud of my role within the team.

How did you get into OR? And what has kept you engaged with it as a discipline? I fell into OR accidentally, but it has kept me hooked. I studied maths at the University of Sussex’s cross-disciplinary school of social sciences and, not knowing what OR was, I looked for a job that would make use of both aspects of the degree. I found one that ticked all the boxes – as an OR analyst with London Transport. My work in OR has developed from doing the analysis to facilitating the analysis, to integrating it into strategy and change management. What keeps me in OR is the kick I get out of using analytical rigour and insight to make improvements happen.

"What keeps me in OR is the kick I get out of using analytical rigour and insight to make improvements happen."

This honour comes at the beginning of your time as President of The OR Society – what are your aims for the Society over the next two years? The OR Society is aiming to grow its reach, so that more and more people take an interest in its activities; to grow the visibility of OR and The OR Society; to promote the engagement of its members with each other and with the Society itself; to grow the “people pipeline”; and to nurture OR research. I want to focus on making this all happen: building on our recent successes – the growing analytics network, the launch of Impact magazine, the influx of student members, the Pro Bono initiative – in the context of the boom in demand for OR skills. There are a whole lot of opportunities out there, and I want to seize them.

 

Ruth tweets at @ruth_kaufman

 



 Archive 

  2017 (12)
  September (1)
  2016 (14)
  September (1)
  2015 (8)
  2014 (5)