The Civil Society Strategy, published by the government last week, sets out the potential of digital and emerging technology in the voluntary sector. "Machine learning" has been one of the tech buzzwords of the past year, so I was interested to hear how the Cystic Fibrosis Trust was using it to help doctors make more informed decisions.
If you’re confused about the difference between machine learning and artificial intelligence, you’re not alone. Artificial intelligence is the science of creating machines that are capable of intelligent behaviour. According to Stanford University, machine learning means getting computers to act without being explicitly programmed. The Children’s Society is already using this technology to develop an in-person translation tool for young refugees. But why should it matter to your charity? Think about the amount of data many charities hold. By using technology to analyse and learn from it we will be able to identify sophisticated patterns in donor behaviour, develop new services and improve the quality of advice we offer, to name but a few examples.
Machine learning and artificial intelligence are dependent on having plenty of high-quality data, but your charity might already have these in place, especially if you’ve prepared for the General Data Protection Regulation. The Cystic Fibrosis Trust, for example, holds the UK CF Registry, which contains annual healthcare information on the majority of the 10,000 people with cystic fibrosis in the UK. This includes hundreds of data items for each individual, with some information going back 20 years.
The charity has partnered with the Alan Turing Institute for a new research project that indicates that machine learning can predict with a 35 per cent improvement in accuracy on existing statistical methods whether a cystic fibrosis patient should be referred for a lung transplant. It achieved this by using machine learning to develop a prognostic model, which helps doctors assess the patient and calculate the risks of a certain course of treatment. It’s effectively using the data to help clinicians and patients make better, more informed decisions about treatment.
Dr Janet Allen, director of strategic innovation at the trust, told me that the project fitted well with the charity’s mission to support people with cystic fibrosis and use the "most cutting-edge tools to improve the quality of life of people with this condition". Some charities might be concerned about the ethical implications of machine learning after the Information Commissioner’s Office found that Google’s DeepMind partnership with the Royal Free Foundation Trust broke data protection law. But Allen says that the trust "has in place strict governance on how the data can be used for research". Data provision is overseen by an independent committee and is only ever provided in line with the informed consent of people with cystic fibrosis and their families, the Health Research Authority ethical approval held by the registry and data protection legislation. Allen says the data released does not contain identifiable information.
Allen thinks machine learning offers exciting potential for the future of the sector. For those with cystic fibrosis, machine learning can help to identify beneficiaries at risk of severe progression and better insights into clinical trials data. The machine learning methods in this study could be used to treat other diseases in the future, such as heart attacks or cancer.
The four questions each charity should be asking itself about machine learning are: how could we use technology to learn more from our data? How could this help beneficiaries? Who do we need to partner with to do it? And how can we look after people’s data when doing this?
Zoe Amar is the founder of the digital and marketing consultancy Zoe Amar Communications. @zoeamar