With the UK looking to become a leader in the field of AI, the new RSS AI Task Force has been consolidating our work in the area and unifying the efforts of our members. We sat down with the Chair of the task force, Donna Phillips, to find out more about her and this important work.
Can you tell us a little about your background and what motivated you to take on the role of Chair of the new RSS AI Task Force?
I have spent most of my career to date as a data professional for government departments and arm’s-length bodies such as the Office for National Statistics, Department for Education, Ofcom and now the Independent Anti-Slavery Commissioner, and I’ve also worked for the social research agency Verian. I’m a statistician, but I’ve been a data science enthusiast for the last ten years, and I was first able to incorporate data science into my work in my role as Head of Statistics at Ofcom.
At Verian, I led on the development of data analytics and AI within our social research and analysis processes, working closely with Faculty AI. That really fed my interest in the ethical use of AI and its benefits in social research and statistics, for example by automating resource intensive manual processes. I’m also very enthusiastic about data ethics, inclusive data and encouraging people from all backgrounds to use and understand data.
I’ve been an RSS fellow for a number of years, and when I saw the opportunity of joining and, potentially leading, the AI Task Force, I immediately knew that I wanted to apply.
What do you see as the role of the RSS in the rapidly changing landscape of AI and machine learning, and what are the key areas in AI and statistics where the RSS can have the most influence?
The RSS has a pivotal role to play in the rapidly evolving AI and machine learning landscape. By sharing our expertise in statistics, data science and ethical standards, the RSS can provide expert guidance on the development and evaluation of AI models, having input into government AI policy development and emphasising the importance of statistical rigour and ethical considerations.
Thinking about the immediate future and getting the task force rolling, what do you hope to achieve in the first year?
The task force has four sub-groups to drive our work for the rest of the year – policy and ethics, evaluation, practitioners and communications. By the end of the first year, I hope that RSS members will see a lot more content from the task force, that we will have contributed to development of government policy and encouraged more AI professionals to join the RSS.
Given the cross-disciplinary nature of AI, how will the taskforce coordinate activities and foster collaboration across different branches of the RSS?
Some of the task force are already members of other sections and groups, such as the Data Science and AI group, the Campaign Advisory Group and the Data Ethics and Governance group, so we will be seeking more opportunities to collaborate with them. We also want to engage directly with members, asking for input, and we’re also planning to hold events.
AI brings a number of ethical challenges, like issues of fairness, bias and transparency. How will the taskforce approach these topics, and what role can statisticians play in addressing them?
Statisticians, data scientists and other data professionals play a key role in this area. We can develop, suggest and evaluate frameworks for fairness and bias assessment. We can create and advise on the development of explainable and transparent models and also how to design ethical data collection methods. We can ensure that AI systems have human accountability.
These are the kind of things the task force will continually raise as important issues to policy makers and practitioners.
Looking beyond the task force’s immediate work, what trends in AI do you think will most significantly shape the field of statistics in the next five to ten years?
In the next five to ten years, the field of statistics will need to respond to an AI landscape that is rapidly evolving in terms of complexity, scale and application. Rather than AI replacing the statistician’s role, I think that we will see a convergence of AI and statistics. AI will significantly influence statistics, for example, through automated modelling, increased focus on prediction and causal inference, the growing use of synthetic data and privacy-preserving analytics and the growing importance of evaluating large-scale AI models.
With that in mind, is there anything you’re particularly hoping to see in the government’s AI Action Plan?
The task force has written a
position paper in response to the government’s AI opportunities action plan that is published on the RSS website. I support the government’s very welcome goals of making the UK a leader in AI and setting ambitious targets. As work develops, there are a few areas where it would be good to hear more, such as helping to minimise risk by avoiding inappropriate models and a focus on effective ones; the regulation of AI and the use of AI and ethics; and stronger recognition of the value of professional accreditation and investment in the development of skills, now and in the future.
Finally, if you could change one thing about AI and statistics in public debate, what would it be and why?
If I were to change one thing about the public debate it would be to shift it towards a more nuanced understanding of AI and statistics, where technical aspects are considered in a balanced way alongside the ethical, societal and legal issues. I would hope that the importance of statistics, and statistics education, is recognised as being key to understanding the potential of AI and mitigating its risks.