In a 2019 report, the UK government’s Social Mobility Commission revealed that social mobility in the country has been ‘virtually stagnant’ since 2014. “Even when those from working-class backgrounds are successful in entering professional occupations, they earn on average 17 per cent less than their more privileged colleagues,” states the report. It’s a similar story stateside: an American born to a family in the country’s bottom 20 per cent of earnings only has a 7.8 per cent chance of reaching the top 20 per cent when they grow up.
“Low social mobility means that people’s backgrounds are determining what they do in life more than their abilities and work ethic,” says Lee Elliot Major OBE, the UK’s first – and still only – Professor of Social Mobility and author of two books on the topic: “Social Mobility and its Enemies” (published in 2018 by Penguin) and “What Works? Research and evidence for successful teaching” (published earlier this year by Bloomsbury). “The result is we’re missing out on lots of talented people in our populations. It’s both a social and economic issue.”
It was this topic – social mobility and how to best address it – that the worlds of business, academia and non-profit tackled during Data for Good Exchange (D4GX) 2019 London. Co-hosted by Bloomberg and J.P. Morgan on Thursday, November 21, 2019 at Bloomberg’s EMEA Headquarters in London, the evening panel discussion marked the first time an event in the long-running conference series focused on data science for social good was held in Europe. As the evening’s moderator, D4GX Conference Director Vicki Cerullo, noted, “There was just so much alignment around this issue and how much we care about this from both organizations.”
D4GX London’s centrepiece was a panel discussion featuring the University of Exeter’s Professor Major, MyBnk CEO Guy Rigden, J.P. Morgan’s Head of AI Research Manuela Veloso, PhD, as well as Gideon Mann, Head of Data Science in Bloomberg’s Office of the CTO – who helped found the Data for Good Exchange back in 2014.
While Major and Rigden have developed long careers tackling social inequality, Veloso and Mann are industry experts from the fields of AI, machine learning and data science. More recently, Veloso and Mann have been exploring how such technologies and methods can be used to directly benefit society.
“AI is a global technology. It has to have a global reach. Not just in business, but for non-profits and the public sector too,” explains Mann.
Today, data science is increasingly being seen as a vehicle to help improve social mobility – to gain insight into society’s most vulnerable populations, uncover previously unseen opportunities, and measure progress on how those underprivileged groups are being helped to overcome the challenges they face.
Data science tools are able to facilitate a personalized approach that can far better target different people, from different backgrounds, at different moments in their lives. Guy Rigden and MyBnk, a financial education non-profit for young people across the UK, aspire to provide the same level of highly targeted support. Unfortunately, getting young people to engage on the topic of financial literacy solely using people-power is challenging, he notes. “Imagine trying to play a game without knowing the rules – or that there’s even a game. Most young people we work with don’t know what they don’t know. Regrettably, this may result in them making uninformed decisions and taking on unsustainable debt.”
Unsurprisingly, Veloso is confident in her alternative: “the scale of the world, the scale of the social good that we’re trying to address, must inevitably be addressed by machines.”
One technique that data scientists are using to explore social outcomes is simulation. With a test-and-learn approach, different ‘what if’ scenarios can be created and the consequences of different decisions explored. The idea is that these digital outcomes can help people in the real world make better, data-driven decisions to improve their prospects in life.
Of course, questions of bias arise when analysing the datasets on which such simulations are based. Veloso concedes that biases will exist in the data, but only because it is human data.
“Data is captured from human behaviours and it inevitably has biases,” she says. “This is because topics like human behaviour are subjective, not objective. Using a broad variety of data is one way of avoiding bias. The second technique is to inject principles into the AI and machine learning systems, in addition to the data – representing the data in a way that the machine can process. The third way is to accept that the machine will make biased decisions, provide it with feedback on the bias, and hope it won’t make that kind of decision again in the future.”
A challenge that can only be conquered in partnership
For the disadvantaged young people that MyBnk works with, there’s no financial safety net to help them recover from mistakes. Data-driven simulations can avoid such issues by helping them determine which financial choices will pay off – a personalised risk assessment, if you like. What’s needed now is greater collaboration across the sectors – the combination of local, day-to-day knowledge of issues with high-level technical know-how – to make it a reality.
“The problems we face in the public sector are too complicated for any one partner to solve,” explains Mann. “Data scientists have the tools, but they don’t often fully understand the problem domain. Non-profits and the public sector understand the problem, but don’t often have the tools. It’s only through multilateral partnerships between these two groups that we can begin to address systematic problems that impact our society.”
In his closing remarks at the program’s end, Samik Chandarana, Head of CIB Data Analytics and Applied AI & ML at J.P. Morgan stressed that it was this spirit of partnership that brought the two companies together to co-host this important discussion.
“We have two internationally-recognized brands in financial services who truly believe that data and the use of data and technology really can significantly enhance the work we do in social change,” said. “Public policy and social programs should always be based on good and fulsome data. Data science and its associated practices is a key tool to analyse data. By bringing technical practitioners together with organizations that are operating to affect the social landscape, that’s how we can accelerate this change.”
“My hope is that tonight is one night among many where we get to exchange ideas and build relationships to solve these problems together,” added Mann.
Watch the entire Data for Good Exchange 2019 London panel discussion below: