Investigating the benefits and impacts of automating science
Project Duration: October 2022 to September 2025
Source: CSIRO Tardis, CSIRO Biofoundry
The Challenge
Artificial intelligence (AI) machine learning (ML) and robotics technologies are being used to develop new methods of automating science and engineering. Laboratory processes are increasingly being performed by machines. Simulations are used to model physical phenomena, test different approaches to environmental management, and test product designs before they are manufactured. AI/ML is also being used to design new experiments that complete gaps in models of physical phenomena.
The automation of science may have considerable impacts and benefits. The roles and responsibilities of scientists and engineers may change as more of their traditional work is performed by AI/ML systems. There may also be trade-offs that researchers make in deciding which parts of the research and development process should be automated to be considered. It is also anticipated that automating science and engineering will have flow-on effects to society.
The increasing use of AI/ML in science and engineering has the potential to be as disruptive to how science and engineering are performed as it has been in many other fields. It may make performing advanced scientific and engineering tasks more accessible to those outside of research institutions and major companies by reducing the specialist expertise needed to perform them. It may also offer the possibility of ‘building in’ ethical research practice into automated research tools by automatically preventing potentially harmful or dangerous experiments or designs from being performed or created.
The increasing automation of scientific practice also allows us to explore the question of whether ethically performing scientific processes will necessarily result in ethical research outcomes. Would it still be possible to perform irresponsible research with automated tools that implement ethical research practices? Do the ethical responsibilities of researchers change if some or all the ethical decisions they would usually make in performing research are already made for them by the automated tools they use?
To make informed decisions about how to further automate science and engineering responsibly, we need to understand how we are already using automation and how we might use it in the future to mitigate potential risks and ensure these techniques are used to create positive benefits.
Our Response
CSIRO’s Responsible Innovation Future Science Platform and D61 are collaborating to explore how scientists and engineers understand and navigate the impacts and benefits of increasing automation in their work. Three research and development areas where new applications of AI/ML are being employed (industrial component design, synthetic biology, and environmental management) will be case studies for exploring how increasing automation is changing the impact of our science.
Through interviews with scientists, engineers and end users developing and using AI/ML in these fields, this project will identify how different forms of AI/ML are being integrated, and what they see as the benefits and impacts of automation in performing research and development.
The findings of this research will provide a greater understanding of how automation is changing science and engineering, and how further AI/ML-powered automation in these fields may be done responsibly. The findings will also inform the development of a framework to guide responsible automation of science and engineering.
Impact
Understanding the potential impacts and benefits of automating science and engineering will better prepare researchers in deciding how to integrate AI/ML technologies into their work.
The findings of this research will inform how responsible automation can be developed and deployed across our diverse portfolios of science for positive benefit. Identifying common impacts and benefits, and how researchers in different fields understand and use automation, will be useful in directing potential future work in how to further apply AI/ML technologies in science and engineering.
Team
David Douglas Justine Lacey David Howard
References