Ethics and integrity of data use in AI and machine learning

At the recent Demystifying AI, machine learning and robotics in healthcare conference, Associate Professor Tam Nguyen (Deputy Director of Research, St Vincent’s Hospital Melbourne) led the panel discussion regarding the ethics of data use in artificial intelligence (AI) and machine learning applications.

Key takeouts

• Australians fear for the use of their personal data outside the direct provision of care
• Change of perspective from ethical review panels regarding the use of social media for research – most ethics panels will no longer allow the creation of publicly accessible data sets from social media as the primary intention of the publicly available consumer posts are for the use of research
• Robust AI solutions require large and diverse data sets. How can this be managed while preserving the privacy and the access rights to these individual data sets?
• There is a need to develop access barriers by shifting the data from the AI models and building an ecosystem that allows the two worlds to connect seamlessly
• Prior to engaging in application related AI work, organisations need to provide clear transparency, accountability and purpose for the use of the data
• There is the potential to encrypt data so even if the raw content of the data is accessed there is nothing people can do with that – consider the trade off between preserving privacy and efficiency
• Instead of sharing data, suppose we share AI models – it turns out that one can sometimes reverse engineer the data even if only given the AI model
• Liability of a privacy breach is currently being managed in contracts. One anticipates the development of new laws to manage privacy issues and data breaches with AI and machine learning applications
• The regulation framework in Australia was established based on clinical trials of drugs where factors such as dose, adverse events and contraindications are determined. The TGA and National Health and Medical Research Council require a completely new framework for approving these new types of technologies including AI and machine learning

The panel consisted of Dr B O’Dea from the Black Dog Institute, Mr R Dearn from Ashurst, Dr D Bauer from CSIRO, Dr S Harrer from IBM and Professor J Bailey from The University of Melbourne.