Big data for categorizing people should be used with caution, expert says

Big data that could be used to identify people as likely child abusers or alcoholics should be treated with the utmost caution, according to University of Queensland researcher Dr Philip Gillingham.

The School of Nursing, Midwifery and Social Work data expert said a fascination with exploring the potential of technology was fraught with human and financial danger.

"It is possible that bringing together and mining multiple databases will provide terrific insights into social problems," Dr Gillingham said.

"For example, datasets in the USA have demonstrated the relationship between homelessness and mortality, and in Australia big data has been used to understand patterns of criminal offending.

"However we also see potential for subjective judgments, errors and inappropriate responses to be magnified on a much bigger scale than ever before.

"You could match the data of homeless people and say a large number are alcoholics, so they should be targeted with alcohol rehabilitation, but what caused their situation is never uncovered.

"We need caution to ensure that we aren't going to waste resources and insult and stigmatise groups of people."

Dr Gillingham collaborated with computational sociologist Timothy Graham of UQ's School of Social Science to deliver a critical perspective on the potential impact of big data in human services.

He said there had already been contemplation in New Zealand of using databases to predict the likelihood of someone being a child abuser.

Yet preliminary exploration found holes in the data, potential for misjudgements and not a great deal of insight compared to what was already known.

"Existing tools already tell us the most likely perpetrators, without spending millions of dollars," Dr Gillingham said.

"The phenomenal cost – and whether that money could be better spent on services – is something that is quite often overlooked."

Dr Gillingham used a personal example to illustrate how big data and predictive modelling could lead to wasted resources and wrongful targeting.

"I have characteristics that align with people who enjoy golf," he said.

"I'm constantly bombarded with advertising at home and online with golfing products.

"But the actual truth is that I hate golf, absolutely hate it."

Dr Gillingham warned that big data could result in patterns that distracted from core issues and could be open to politically-influenced interpretation.

Big Data in Social Welfare is published in journal Australian Social Work.