Guiding principles and tangible actions must work to ensure the safe and effective use of emerging technologies like AI
By: Keith Bigelow, GM of Analytics & AI, GE Healthcare
Everyone can be a little bit biased, or so social scientists say.[i] In the field of Artificial Intelligence (AI), where machines are learning to guide decisions as inconsequential as picking a restaurant and as life-changing as helping clinicians diagnose disease, guarding against this reality has never been more important.
Fortunately, ethical standards are not new to influential professions, especially healthcare. The Hippocratic oath – the oldest and most famous code of ethics – has guided doctors for centuries. Today, nearly all medical students in the U.S. take some form of the pledge.[ii] As advanced analytics and AI continue to proliferate, some of the principles guiding the industry should evolve.
AI presents a unique challenge in ensuring safe, fair and effective usage. Outputs generated by algorithms are only as good as the data put in. It can be difficult to demonstrate how an algorithm arrived at its decision. But these challenges are met with incredible opportunity. AI has the ability to make sense of the massive volumes of healthcare data. It could take away mundane and repetitive tasks and enable precision health.
To enable the benefits and mitigate the potential harms requires not just updated guidelines but also new approaches to technology development and evaluation adapted to AI. The good news is that the industry knows we need to get this right – reflected in the increasing body of research on the topic and the formal statements of commitment by companies taking major steps into the space.
Today, GE Healthcare officially becomes one of those companies. We are publishing the following AI principles, which we will apply to improve healthcare quality, cost, access and the patient experience.
AI Systems exist to augment human intelligence and must:
Be designed for the benefit, safety and privacy of the patient
Be a trusted steward of the data and insights
Be transparent and produce explainable outputs
Guard against creating or reinforcing bias
Optimize the safe development, production and compliance of therapeutics and healthcare solutions to deliver Precision Health
These principles are both timeless and timely. With more than 100 years of industry expertise, GE Healthcare has long been guided by the need to do right by patients and improve lives in moments that matter. But, as we push the boundaries and explore the possibilities of AI, we must guard against the adverse effects and biases we know are possible if development and deployment are not done correctly.
Here are a few steps we are taking beyond putting pen to paper.
Commitment to diversity: In addition to employing a diverse data science team, 26 percent of whom are women, 80 percent of whom are a minority, and 44 percent of whom sit outside the U.S., GE Healthcare is committed to using diverse data sets that mirror the global population who our technologies serve. We are also validating our algorithms with a wide range of healthcare systems. Diverse teams, diverse data and diverse validation are critical to proactively identifying and neutralizing potential biases.
Commitment to transparency: No data enters our platform without a source associated with it, clearly outlining the intended use and affirming data usage rights with each entry. This is step one to transparency. The second step is tracking and managing all data once it is in our platform, which makes the algorithm less of a “black box” and more easily traceable, reproducible and fixable, should problems arise. The final step is treating advertising claims responsibly. As we have done for decades with our medical devices and monitoring solutions, we will assess every statement for accuracy and context before it is used externally. As the consumerization of healthcare accelerates, now more than ever it is crucial to ensure that the misuse of customer data in the consumer market[iii] does not permeate into the Healthcare domain.
Commitment to partnership: We can’t do this alone. GE Healthcare is working with academic institutions, regulatory entities, governments and other industry partners to ensure we do AI right – practically, methodically and for the benefit, safety and privacy of the patient. For example, I recently had the opportunity to attend and present at an AI in Medical Imaging workshop with thought leaders from government and regulatory bodies, academia, leading hospitals and solutions providers to discuss the opportunities in front of us as well as the common challenges we face. These proceedings will result in a set of papers with recommendations regarding data policies and regulatory approach.
Ethics shouldn’t arise after a breach of trust or crisis. They should be engrained into who you are and what you do. GE Healthcare may be publishing AI principles today, but these commitments and actions are cemented in our past, present and future. The work to ensure safe, ethical and effective use of AI will never stop. I am proud to lead this effort for a company that has successfully navigated healthcare innovation for more than a century, and I am proud to work with a diverse and empathetic team to continue to pioneer positive transformation.
[iii] http://www.unfitbits.com/assets/European Journal of Cultural Studies-2015-Crawford-479-96.pdf