Data Enabled Decision Making

This priority focuses on new methods to help support people making decisions in a world that is becoming ever more data rich. This could be assisting a person in making a decision, providing a system that can make decisions autonomously (but with the full confidence of those affected by them), or a combination of the two. The data in question will often be complex, incomplete and/or mixed mode which introduces new challenges to decision making.

The priority requires researchers to take an integrated approach in which every element reflects the ultimate need for the outputs of that process to in some way benefit a person making a decision. This will include, but not be limited to, data wrangling, data analytics, interaction with data and data visualisation. There could also be opportunities for work on hardware and computer architectures for enabling faster, more efficient or even real-time decision making.

It is recognised that there are 'users' in many parts of the data science ecosystem, both machine and human, expert or non-expert. Researchers are encouraged to consider these users from the outset, rather than in the final stages once a technique, approach or application has been developed. This should be in line with the People at the Heart of ICT Cross-ICT priority.

There will be many challenging research questions around the decision making aspect of this priority. How do we account for and incorporate uncertainty in these decisions? How can we minimise unintended bias being introduced into the decisions made, through the choice of data, methods and questions used? And more broadly, what does a good decision look like and how do we measure it? Alongside this, we expect researchers to consider responsible innovation in aspects of their work. For example, is there a need to devise means to make algorithms accountable for their behaviour? Does this "accountability of algorithms" (or lack thereof) affect how and where such systems can be implemented?

We recognize that work in this space often crosses in to other research or user domains, but work funded under this priority should be primarily novel ICT and should therefore be generalizable to a degree.

Effective work in this area requires access to appropriate data sets and facilities, which is often challenging as this relies on open policies of domain or user partners. Appropriate access to data sets, users and decision makers will be essential to meet the aspirations of this priority. There are opportunities already available to researchers to generate or procure data as part of research grants, and examples of best practise of effective partnerships with data owners in current investments which they could emulate.

Implementation of the Priority

The research landscape in this area is complex with many research communities and funders involved. For this reason, we will first set out the EPSRC portfolio in this area and work alongside other Research Councils and the Alan Turing Institute to articulate where the boundaries and opportunities are.

The 'New Approaches to Data Science' Call is a joint call led by the ICT, Digital Economy and Mathematical Sciences Themes. It is expected to fund a number of proposals which will have strong relevance to this priority, although the decision making aspect may not be a feature of all.

It is recognised that there is high academic and industrial demand for skilled people in this area in the UK. This priority will therefore be a key influence on EPSRC strategy on any future student training priorities. In addition to this, applicants applying for large programmes in this area should articulate how they are contributing to the skills base. This could be through, for example, staff development, dissemination of techniques to a wider audience or public engagement.

We will hold an event in 2018 to showcase examples of best practise in this area and lessons learnt from working in this way. We will expect all researchers with projects directly relevant to this priority to present their progress so far and how they are meeting the aspirations set out above.