Ai needs to save us from Ai

25th January 2018

Article first published in Digital Agenda, 3rd January 2018.

The recent acceleration of innovation in deep learning and machine learning made Ai the hot topic in boardrooms, media and government throughout 2017. If there’s one thing that most economists agree upon, it’s that change is coming and its scale will be unprecedented and the consequences profound.

According to a new study from the think tank Future Advocacy, at least one fifth of jobs are at risk of being automated. The study says the highest levels of automation will be in Britain’s former industrial heartlands – the Midlands and the north of England, as well as industrial centres of Scotland.

These are areas of the country that have already suffered from deindustrialisation, and remain unemployment hotspots. It’s going to be a further kick when so many people are down.

On top of larger unemployment rates – and not entirely unrelated – these parts of the country are also focal points for other, more personal issues. Uncomfortable, complicated things that will inevitably impact all of us at some point in our lives. Things we choose to ignore because it’s often easier to think that these issues only affect other people.

You see, people in these areas of the country – especially men – are ending their own lives in record numbers. It’s a sobering reality. We assume that the type of people who die by suicide are unwell, disturbed or unlucky; people who stumble at life’s biggest hurdles and are too weak to get back up. Most of us think we’re made of sterner stuff.

What we don’t realise is that 75% of people who take their own lives have never been diagnosed with a mental health condition, or that only 5% of people who suffer from depression go on to take their own lives.

A 2015 study by researchers from the University of Zurich in Switzerland found that between 2000 and 2011, one in five of an estimated 233,000 annual suicides were linked to unemployment.

The reasons behind every individual suicide are unique and complex, but we know common factors that contribute: relationship breakdowns, bereavement, socio-economic factors and mental health problems.

Jobs and the economy are key factors, but we should not fall into the trap of assuming correlation equals causation. It is not that austerity and job loss themselves kill, but that people who are already vulnerable face another pressure.

Job loss and waking up ‘unskilled’ one day can be brutally crushing and make people feel like they’re in the way, useless, worthless, unworthy and a burden on families. This is the ying to the Ai yang – when we use data to automate a job and make the economy tick harder, we also erode some people’s sense of purpose.

So, while the intelligent machines herald the beginning of a more efficient world, do we have a corporate obligation to help create a more compassionate world too?

The same efficiencies we can create on the production line using the new wave of intelligently artificial algorithms might be the answer to tame the genie we’re about to let out of the bottle.

Of course, there is no one-size-fits-all algorithm we can design for the complex issue of suicidal ideation and there are many geographic and cultural factors to consider. However, using the predictive capability of data and machine learning, we can at least give some vulnerable people more insight into their feelings, which might offer a level of support that currently doesn’t exist.

People generate a plethora of personal data on smartphones through online movements and that data can help us to better understand ourselves in ways we were never able to do before.

Unlike traditional approaches, deep learning systems reveal novel ways of looking at social data, finding patterns or indicators that scientists, psychiatrists and community mental health teams may often overlook. True insight often lives at the boundaries and untapped value may be nestled in the gullies among disparate data sets. Combining different sources of information will make those insights sharper still.

Applied machine learning could also serve as the missing insight tool for HR departments, which are about to deal with high redundancy levels. Or maybe our own data will help employees find an untapped talent they can use for finding a new role. We have many opportunities to do the right thing and find the people out there who are within our reach to save.

High unemployment creates uncertainty among the unemployed about their possibilities on the labour market which will create huge stress and lead to some serious consequences. But even though I believe we’re about to create a huge crisis, I also believe that the falling cost of human-related data capture has now opened the door for the Ai community to begin to explore the possibility of a truly synoptic overview of the human condition and support us during the coming age of automation.

Machine learning is wonderfully positioned to improve our knowledge about areas like suicidal ideation and help us deliver interventions that are sensitive, safe, and address one of the leading causes of death.

But, be warned, if it is oversimplified, non-transparent, and not subjected to the “do no harm” principle, it will simply add to the current challenges and potentially worsen the situation, it’s a fine balancing act. But, make no mistake, for some areas of society, automation is taking us towards consequences we cannot afford to ignore.