Never before have we been able to analyse so much data, so quickly. But that isn’t always a good thing. Director of the Faculty of Health’s Centre for Health Services Management Joanne Travaglia and researcher Hamish Robertson have teamed up to examine how unchecked assumptions about data are reinforcing privilege, prejudice and inequity, and why a more critical approach is needed.
We have reached a point in history where understanding the world means understanding big data. The rapid growth of interest in this field has been so marked and so universal that some theorists are predicting it will result in a major paradigm shift in industry and research.
Presented as an uber (pun intended) example of disruptive innovation, the buzz around big data and data analytics is palpable. But studies of how the big data field developed, its risks and benefits and how individuals, communities and organisations can manage the implications of ever-increasing volumes of digital data remain scarce.
Instead of the conventional cross-sectional or longitudinal studies that grew out of the ‘small data’ age, there is a perception that big data is neutral, immediate and comprehensive. But things aren’t, of course, quite that simple.
While there is some evidence that big data medical research is producing positive findings, not all of the big data industry is aimed at socially beneficial outcomes. In some places, governments are trying to retreat from conventional data methods on the basis that big data will suffice.
However, it often serves quite different purposes to conventional methods like a census or regular social surveys. And it is often volunteered data, so it’s both technically free and monetisable – which is doubly appealing to some governments and the companies that lobby them. The risks, therefore, are real.
Breaches of centralised data have always been a concern and remain a major problem in healthcare and elsewhere. The tracking of behaviour – including predictive advertising – often raises discomfort in even the most avid social media user. Indeed, we would argue that big data is becoming problematic and the cracks in its uncritical application to social issues are already beginning to show.
Of course, this isn’t a new problem. Since Victorian times, at least, data (of every type) has been used for social engineering purposes. Social inquiries into vulnerable groups were, and can continue to be used to ‘identify’, quantify and isolate the source of such problems. It’s here that we saw the origins of enduring ideas such as ‘scientific’ racism, eugenics and the workhouse system – counting and classification were central to these ideological and practical systems. Indeed, Victorian-age social policy ideological inheritance is alive and well in the 21st century. That’s one reason why we face many of the same problems and why the promised solutions prove so intangible still.
Throughout history, systematically collected ‘scientific’ data has also been used to justify the isolation, torture and abuse of people with intellectual or physical disabilities, people who are gay, lesbian, bisexual or transgender, First Nations peoples, and virtually every other vulnerable group.
At the very same time these groups continue to be regularly excluded from other forms of data collection. This absence has significant implications for health services delivery and health services management.
“Big data is becoming problematic and the cracks in its uncritical application to social issues are already beginning to show.”
Take, for example, cardiovascular disease. It’s the leading cause of premature death amongst Australian women, with Aboriginal and Torres Strait Islander women and those from lower socio-economic backgrounds experiencing even higher rates of morbidity and mortality than the general population.
But we know from numerous other studies that women are significantly under-represented in the majority of clinical trials. Women’s symptoms are also different to men’s, and comparatively under-studied, and so they continue to go unrecognised by the general public and clinicians. As such, women to miss out on effective treatments at virtually every step of their interaction with healthcare, from prevention, to diagnosis, referral and treatment.
How can we design, implement and evaluate the efficacies of interventions and services if data from half the population is missing? We can’t.
There are also socio-political risks to consider. It’s already apparent in some of the critique of proprietary algorithm use in social policy domains such as welfare and justice work. Cathy O’Neill, for example, describes much of the work in this field as “weapons of math destruction”, by which she means many of the unchecked assumptions in such models and methods reinforce existing systems of privilege and prejudice.
Individuals on the margins of society can become even more vulnerable through the application of advanced big data technologies. Once big data is applied to social problems, which is to say ‘people problems’, then we need to pay much closer attention to its application. This is especially true in the social sciences where we have seen this kind of led-by-the-nose enthusiasm for the promised ‘big fix’ before.
So how do we as students and researchers of technology, business and humanities address this within and across our disciplines?
Our argument is that we need not only a sociology ‘of’ big data, that is, how to understand and use big data in sociological contexts, but also a big data sociology. The latter would need to examine the sociological implications of big data use and possibilities from a critical perspective. Data is not, and never will be entirely neutral. It is a tool, and like every tool is dependent on the intentions of those who use it.
To begin, a big data sociology needs to confront the hype front on. No more techno-mythology or endless streams of unaccountable ‘innovation’ stories. At least not without close inquiry, analysis and follow-up. Secondly, the impact on society’s vulnerable groups have to be transparent. No more hiding behind the algorithms (‘the maths made me do it’) or proprietary systems, copyright law and weak, individualistic ‘privacy’ protections.
These responses may be unavoidable in some situations but they also tell us something about the risk attached to some big data applications. If the provider wants to hide how their system actually works, then they have probably got something to hide that is of importance to the people it targets.
And that’s why we need a big data sociology to confront these issues directly and separate the big data hype from its potential.