Stephen Heppell – “This is not scary, this is exciting!”
This is a short paper full of big ideas about Big Data. It is a frustrating read in many cases because (surprisingly for an academic) there is no evidence for most of the assertions.
The paper can be found here
There are some propositions I am happy to go along with, for example “it is disappointing that our measures of effectiveness, and our management data are both so poor in 2015. As a result, a lot of what we do in schools is simply convenient rather than optimal.” But the observations that follow from this seem poorly informed. “An athlete in any sport would have a precise understanding of their nutritional requirement and the impact of various meal options on their performance”. I’m no sports nutritionist but “an athlete in any sport”? There isn’t a precise understanding of what foods make for strong performances in all sports. Cycling is the one sport I know well and here many individual elite performers often have strong historical data to help them determine what might be better foods to consume. But that often means they stick to foods that have worked well in the past. If they have never eaten kale and borlotti bean stew ahead of a day of racing they wouldn’t know how it might impact on their performance. A gifted amateur athlete outside a development programme won’t even have anything like that data. They would just have generic guidance to rely on.
This is also pretty obviously facile “as part of a research project, we asked students for indicators that their learning on a particular day was exceptional; one said “that would be how fast I eat my dinner” because he knew that on a really good learning day he would eat fast to get back to work!” What is the significance of this? The student’s own view of what is significant data doesn’t offer any help for the analyst of a data set from a wide range of different individuals. There may be many possible reasons why a student eats quickly. Where lessons are timetabled and they start at precise times there would be no reason to consume a meal speedily. Also “If a school declares that Wednesday will be a Discovery Day, a day of immersion and of project based, mixed age work, and on that day the children come to school faster and stay longer, we would have learned something important about engagement.” Would we really? Correlation isn’t the same as causation, what if that was a very wet and cold day?
In other places it’s unnecessarily opaque. For example what does this sentence mean? “Knowing that in office environments a minimum lux level for conversations would be around 250 lux, whilst for close work like typing and writing it would be above 450 lux we started exploring.” Presumably a ‘lux’ is a measure of ambient light? I don’t think the use of the term or the numbers add anything to the argument. Much simpler to say that research from Offices shows that most school exam rooms do not have an optimal level of ambient light. Stephen Heppell has been a proponent of new builds as a tool for educational improvement so it’s not a surprise that he finds deficits in buildings. These are important and shouldn’t be discounted. Levels of CO2, ambient light and noise must make a difference. The problem is that he doesn’t quantify that difference. Without quantification it’s possible that a school could spend a great deal of money to achieve only marginal gains. Knowing that a factor has a positive or negative impact on learning is just the start of a process of decision making. Suggesting that schools act quickly whenever they find one possible source of poor outcomes isn’t helpful. A phone company may find that consumers say they would prefer a titanium model, but if the extra 20% cost of the material only increases sales by 3% and reduces margins, then it’s probably not a good investment.
All of this is frustrating because the overarching suggestions make excellent sense. It is a shame that they have been padded out with examples and ideas that seem half-baked. The big problems for Big Data in schools aren’t examined at all and that is a serious weakness. There is nothing about ownership of data and how schools have to manage the increasingly rich datasets they hold. A more level headed analysis might allow that individual schools won’t be tackling Big Data in the commercial sense, at least not in the medium term. Businesses like Amazon and Google (where Big Data is a reality) have datasets with trillions of data points. Even if a school were to retain ten pieces of data on every school meal eaten in a year that would only add up to a couple of million data points for an average secondary. Even more problematic is linking that to attainment, progress and other data. It is worth pointing out that no UK state school is at all likely to be in any kind of position to achieve this in the next five years. Schools are at the start of their data analysis journey and their data is relatively small. Stephen Heppell makes the exploration of the potential of data analysis less rather than more likely by portraying highly ambitious aspirations as relatively close at hand.