Conflict in the code: An introduction to ethical machine learning for healthcare
In the first of our three-part series on ethical machine learning in healthcare, we explore some basic concepts.
What is ethics?
In its simplest form, ethics is a system of moral principles that affect how people lead their lives. Sometimes used interchangeably with the term ‘moral philosophy’, ethics covers concepts such as the criteria for what is ‘right’ and what is ‘wrong’, human rights and responsibilities, and what is good for individuals and society.
Our understanding of ethics originated from different religions, philosophies and cultures, but at the core of ethics is concern about something or someone outside of ourselves, such as other people, the interests of society, or what we commonly refer to as the “greater good”.
What is machine learning?
A subset of artificial intelligence (AI), machine learning is the study of algorithms and statistical models used by computer systems for pattern recognition and predictive modelling. Machine learning is the process of a computer looking for patterns in data and trying to draw conclusions. In practise, data scientists provide a computer with historical or representative data, and a computer evaluates which model best fits that data. The learning process can be continuously refined, training the algorithm with ever-increasing accuracy.
Ethics in research and medicine
Scientists traditionally operate within a framework of research ethics, which means that they need to consider objectivity, confidentiality, non discrimination, social responsibility, and the protection of human subjects through informed consent and the right to withdraw. Data scientists are no different.
The concept of ethics in medicine is not new. One of the mostly commonly understood and practised systems of ethics is that of the Hippocratic Oath. Written nearly 2,500 years ago, the Hippocratic Oath is taken by physicians who swear that they will uphold specific medical standards. Doctors, often upon graduation from medical school, pledge to ‘do no harm’, maintain privacy and respect their patients.
Ethical machine learning for healthcare
Machine learning involves computer-aided research, sometimes in real time. We therefore need to train our computers to follow good research ethics, so that the results are applied using responsible medical ethics. Both our researchers and our medical professionals need to apply their ethical frameworks to ensure that our machine learning is ethical.
We can then fulfil the promise of machine learning not just to ‘do no harm’, but to do a lot of good.
The next article in the series will explore the role of machine learning in healthcare today, and examine some common issues.