Start The Week (Radio 4): Big Data
Here’s a link to this morning’s Start The Week, discussing “big data” and mathematical modelling of data. Well worth a listen. Contributions are from James Owen Weatherall on physicists in finance, Marcus du Sautoy, Kenneth Cukier and sociologist Tiffany Jenkins.
While there is an unstoppable logic to gathering and making sense of all the data out there to better inform ourselves, I welcomed Tiffany Jenkins’ reminders that ultimately human (non-mechanised) judgments need to be all over this and not sidelined by it.
As she points out, physics, if done well enough, may well be useful in explaining and predicting how a market functions – fine as far as that goes – but has little to say on what the market’s role is or should be. The question of whether the point of financial markets is to grow profits for investors, benefit wider society, grow GDP, increase happiness or whatever: these are questions of political and moral philosophy, which are rightly the stuff of politics and democratic decision-making. Modelled data can help inform the decision-making but ultimately both the decision and the decision-making process (which involves the prioritisation of some goods over others) is qualitative in nature. Further, the decisions about what data is to be gathered – and even to some extent what it means – are also made using holistic forms of thinking and problem-solving that draw as much on humanities-derived methods of critical analysis and sense-making as on pure ‘scientific’ methods.
In research, we are often divided into ‘quantitative’ and ‘qualitative’ research – and I am a specialist in the latter – but as such I have frequently worked on projects where I collaborate with ‘quant’ colleagues. The process of getting ‘the story’ from qual and quant data is not one of quant being the cake and qual being the icing on top. The two sources of data and modes of thinking work best when they interact throughout the project, from the planning, through the fieldwork design and into analysis. The numbers only help if you’re asking the right questions. And getting to the right questions is actually a qualitative process.
Even in the interpretation of the quant, qualitative judgments are made (though this is often not formally recognised). For example (and this is purely hypothetical), if data were to show that 60 per cent of 18-21 year olds surveyed were regularly browsing toys aimed at 10 year olds, quant researchers would rightly give this special attention. It would either be an anomaly that needed to be explained away, perhaps due to a glitch in the survey wording; or it could be an insight into something significant, like a fashion for indulging in a second childhood during early adulthood. But what would mark it out as significant is not inherent in the data itself. The red flag in the researcher’s mind when she sees the figure derives from the researcher’s personal wider knowledge of life and the norms of the area they are researching. Systematic deductive reasoning can only produce such a red flag after a very circuitous and laborious process, if it manages it all. Sometimes it is just quicker and better to use human brain power and experience. The concern is that to a scientist, such judgement may seem unsatisfying and the logic seems to drive us to leave more and more of our decision-making to machines. But in reality they can only make good choices for us with very close human supervision. Perhaps we should be paying more intention to that human supervision and the goals we are trying to achieve. I can recommend Michael Sandel‘s What Money Can’t Buy on the limits of how far maximising efficiency can get you before you have to engage with moral philosophy and political and social choices. It’s not very far at all.
The other issue with the mathematical modelling of data about human behaviour, as we know from Kahneman and the behavioural economists, is that sometimes the models have been based on some very outdated (and, frankly, wrong) assumptions about human psychology and how decisions are really made. The idea of humans as ‘rational actors’ is of course nonsense, though many market ‘models’ have been based on it. It’s only by grasping the nature of the myriad innate biases and barriers people work around that you can make sense of it all. And as with any system, rubbish in, rubbish out.
But big data is an opportunity for bring really useful insights to decision-making. As long as we’re not tempted to under-think what we’re using it for.
- Retailing and Big Data (c24.co.uk)
- IBM TryTracker Brings Big Data To Rugby (techweekeurope.co.uk)
- Millions of lives blighted by toxic Big Data (computerweekly.com)
2 thoughts on “Start The Week: on “big data””
A really useful perspective. Sometime I think “big data” is a “woods for the trees” situation: the right decision is what matters, not how we get there. Sometimes big data will get us to a different, better, more nuanced conclusion. Often I suspect it may just provide a useful point of triangulation.
David Brooks seems to be focused on big data at the moment; his article “What data cant do” (http://nyti.ms/WJcp9s) says it all better than I could.
Hi Simon, Thanks for the link to the David Brooks article – brilliant. Enjoyed your last blog post too btw, will be back on soon.
Shore Qual Research & Consultancy
Landline: 01865 556533 Mobile: 07850 328880 email@example.com
Website: http://www.shorequal.com Blog: http://www.shorequalblog.com Twitter: @ShoreQual