Should statisticians be more involved in shaping the future of data ethics? Tom King, secretary of the Royal Statistical Society’s Data Ethics Special Interest Group, tracks the current progress in the data ethics landscape and urges RSS fellows to get involved at this vital stage of development
When the RSS responded to the government consultation on the Centre for Data Ethics and Innovation in September, we could see the potential to lead internationally as there is no established precedent elsewhere in the world.
Indeed, in reviewing the international landscape, we are finding many countries starting to talk about ethics but mostly around regulation and privacy. In the UK there is a lot of talk about ethics, especially for AI, but there is a lot of fragmentation in sectors and applications which we hope the new Centre for Data Ethics and Innovation (CDEI) and Ada Lovelace Institute will convene across.
In its announcement about the CDEI, the government stated that its Office for AI – a joint unit between the Departments for Culture, Media and Sport (DCMS) and Business, Energy & Industrial Strategy (BEIS) – will work with the Open Data Institute to explore the potential of data trusts - which it's hoped will allow organisations to share data in a safe, fair and ethical way. We think these are a good idea if the details can be worked out – particularly regarding governance.
We also welcome the prospectus of The Ada Lovelace Institute, whose formation was in part prompted by the work of the RSS going back some years through our Data Manifesto. We hope these new bodies will start to comment on what data applications are doing and that these strong central voices will help to drive the need for coherent ethical training in data science which has yet to be clearly articulated.
It seems that suddenly everyone is publishing reports on data, and the quality of some is very high. The Royal Academy of Engineering last week published its report on Data Sharing, which takes a very practical approach, but data governance is an issue where they can only see some of the outstanding questions.
The Academy of Medical Sciences report last week, ‘Our data-driven future in healthcare’ carefully challenges issues such as the role of commercial organisations in health technologies – saying that they have an important role in development but that the public is opposed to patient data use in activities such as direct marketing.
Other sectors have much to learn from the established patient engagement practices in health which are now applied seriously to data. The Nuffield Council on Bioethics has led on public dialogues in this area and we expect the citizen juries, now popular in health, will be refined for other purposes.
Data governance in health has been particularly cut off from other sectors, so we are pleased that Office for Statistics Regulation (OSR) is following up its ‘Joining up Data’ (PDF) report by looking at health data. The OSR has also published guidance on Data Governance, under one of their pillars ‘trustworthiness’. This is emerging as one of the priorities for the Data Ethics Special Interest Group as we find it means different things in different contexts.
One thing is clear: the need for statisticians to engage with these oversight mechanisms is growing and statistical expertise is essential to inform governance and focus on emerging risks. We know from some research that interdisciplinarity is key to effective governance, including the participation of statisticians who understand where the data is coming from and how it is being used. At the moment, we do not feel that RSS members are well represented in these debates.
While we welcome the inclusion of the RSS executive director on the Ada Lovelace Institute, some RSS fellows are concerned that there is a lack of practising statisticians involved in these organisations.
As Andy Garrett, who chairs the Royal Statistical Society’s Data Science section, points out: 'Both the Centre for data Ethics and Innovation and the Ada Lovelace Institute boards currently include eminent experts from various fields, but expert statisticians are notable by their absence on both. Without including those who are closer to both the data and the underpinning statistical methodology, they may not be sufficient. We should look to get a seat at the table.'
To this end, the Data Ethics Special Interest Group wants to know more about member concerns and interests in this area. We encourage any fellows who would like to get involved or voice their concerns to contact the committee via its webpage.