How did I get into data science and analytics?
I’d love to say it was part of a grand plan for career development, and I’m lucky that it almost looks that way, but I arrived in data science as an intern during my year out between an undergraduate and postgraduate degree in economics.
As it happens, I’m completing this gap year as I type.
What has kept me in the field?
Despite the ungraceful entry, I’ve stayed in the field and become increasingly immersed. One of the aspects that continues to interest me is how quickly the field develops. Technology and methodology that is current today may be outdated in as little as a month.
This stands in contrast to the world of undergraduate economics, the courses of which often feel as if they could have been written in the early 20th Century.
So it’s a refreshing experience; however it would be wrong of me not to incorporate economics into my interest in data science.
Fortunately the two marry rather well. My main field of interest is behavioural economics, a sub-field of economics which seeks to understand economic decision-making through the lens of behavioural science. It relies almost entirely on the outcomes of controlled and natural experiments. Natural experiments draw on data generated by people living their everyday lives, rather than circumstances manipulated by a researcher. This data could involve online spending patterns, current account applications, or the number of visits to the doctors. The list is infinite, and harvesting the data is a data-science problem; to the rescue come tools such as Hadoop and Apache Spark, cornerstones of big data analysis.
Looking at this from the other direction, in data science we work with predictive models which can generate behavioural questions. These models allow for sophisticated identification of relationships between variables in the real world. A common example is the use of big data in the Obama 2012 campaign. The campaign aimed to identify individuals most likely to be persuaded to vote for Obama. Campaign volunteers were duly sent in the direction of these malleable voters.
The question then migrated from a data science to behavioural science problem: how do we actually persuade these people to vote for Obama? In practice it was answered with a three-pronged technique.
- Hand these voters a small commitment card.
- Ask them to articulate their plan for voting
- Let them know their neighbours’ voting intentions.
I won’t explain the reasoning here, but there is a very good article which goes through the thinking behind these tactics and introduces ‘nudge’ theory.
These three tactics, drawn from research in behavioural science, each aimed to increase the chances of a voter favouring Obama.
So there are links in both directions here, from behavioural questions to data-science solutions, and from data science questions to answers from behavioural research.
What advantages have come with data science and analytics?
I can say this without hyperbole: working in data science has given me the most useful and transferable skill set to which I have been exposed. The tools used, the techniques practiced, and the understanding gained as a result of asking questions in data science marries with virtually any technical discipline.
Most of all however, there is a certain way of thinking encouraged in data science; it’s a reasoning process in which open-mindedness is its highest virtue, and useful (and above all unexpected) insights its treasured output. It is the field behind countless innovations and is becoming increasingly prominent as business rely more on data-driven decisions above intuition. As a discipline, economics is moving towards a data-driven future, so are the people with whom I work, and so am I.