his is the second in a short series of posts looking at how Dimensions Research Strategy came together. In our first post, Ross Andrews talked us through the early discovery work with development partners – the real-world conversations that are shaping what Dimensions Research Strategy can do.
In this second instalment, we speak to Kate Byrne, SVP Product, Academic at Digital Science, who is working withthe product team for Dimensions Research Strategy. Kate came to Digital Science after ten years working inside research libraries and university research offices, before joining the team designing the tools she once relied on.
We talked to her about what connected data actually makes possible, the questions institutions don’t know to ask, and why now is the moment to solve this differently.
Why this product, why now?
Institutions have the data, but not always a clear way to turn it into action. The question is how we make that easier. This product is designed to sit at that intersection point, across research, intelligence and strategy.
At Digital Science we already have some fantastic products that cover a number of areas in this problem space. Dimensions gives you this incredible dataset to be able to answer questions in this space, and to do querying and analytics to explore the area. With products like Symplectic Elements and Figshare, we have these toolsets that allow you to really own your organisation’s activities, curate them, make works openly available and reuse that data for a wide variety of purposes. But we wanted to go a step further.
The newer technical possibilities of some of the emergent AI capabilities have created an opportunity for us to answer that question in a new way. We were able to bring together some incredibly capable technology specialists and subject matters experts with some of our amazing community members who were willing to step up and be development partners. This is the moment that we can work together on core problem that’s never quite been solved.
The Digital Science difference rests on connected data – grants, publications, patents, policy, attention, and the links between them. Why is that the right foundation for strategic questions, and what does it let an institution see that a single source can’t?
Connected data is really essential because important strategic questions are almost always inherently multi-dimensional. The really juicy stuff – the non-obvious stuff, the special insights – comes from that intersection point of multiple data points.
Look at the work of a group of people: the funding that goes in, the publications and datasets that come out, the attention and impact that follows. In real life, that’s all one organic flow of real people doing real things. If we artificially chop up that flow into separate datasets, then we’re not going to really truly understand the complex landscape of what’s actually going on in an organisation.
We need multi-dimensional data to have a hope of doing analysis that represents the multi-dimensional reality of an organisation, and get answers that can help make the right decisions for those multi-dimensional problems.
You spent years working in libraries and around academic institutions before joining Digital Science. How much of Dimensions Research Strategy is designed to answer problems you saw play out yourself, on the institution side?
Looking back to when I started out in this space 20 years ago, a lot of the challenges were really at the level of collecting and curating the data necessary to do this kind of analysis. This was early enough that we were doing citation curation and H-index building by hand, working across multiple tools to locate and harmonise the data. Getting to actually do the analysis was the bonus!
Today we’ve got so much more data at our fingertips, but it has become a challenge of scale and speed. Finding the time,focus and resource to do meaningful analysis at the scale and speed that you need to work in today’s world has become essential to allow organisations to meet the challenges they face. .
Those institutions can be so diverse that two different departments may be completely unaware of what each other are doing. What do vendors most often get wrong about how institutions actually work?
I think it’s really easy to get focused on the outputs – the publications, the content – and see that as representative of the organisation. But when you talk about a university, these are large, complex organisations. Their organisational structures are complex. The relationships between those departments are complex. They can be the size of a small town!
It’s really easy to assume that an organisation like a university is going to be homogeneous in its needs, where actually the needs of a law faculty are going to be completely different from the needs of a biomedical science faculty, and different again to a computer science department. Dealing with the reality of those domain and cultural differences is super important if you’re going to actually help provide contextually relevant analysis and decision support. These are diverse organisations full of diverse individuals, and vendors need to understand that, adapt to that, speak their language, and take the time to understand those differing user groups and their needs.
Do you see this platform also democratising access to that type of knowledge and data internally – broadening the types of people who can engage with it beyond just the analysis team or the research office?
Yes, that’s absolutely my hope. One of the things that I have really loved in my ten years at Digital Science is getting to learn more about all of the different pockets of specialist research support or administrative teams in different corners of universities around the word; helping researchers connect to either other parts of the university, or external groups like funders or industry partners. When I talk to people across those roles, they’re not necessarily connected to each other, but I see so many points of commonality in what they’re trying to do and I believe that Dimensions Research Strategy will have a lot to offer many of these teams.
With such a fragmented landscape, how do you help surface the questions people don’t know they want to ask – the unknown unknowns?
As humans, it’s very easy to gravitate towards the things we know. There are always those researchers who’ve been the standouts – the leading PIs, the big names. But when you’re making strategic investment decisions, it’s just as important to know who the next big name is going to be, which research areas are trending upwards, where you’re investing time and money that isn’t currently matching the outcomes you’d hoped for. Pushing beyond the obvious answers and surfacing up-and-coming strengths is where things get more interesting – and to do that, you need interconnected, multi-dimensional analysis: looking at the comparative experience levels of researchers versus output versus funding inputs, so you get to explore where you’re going, not just where you’ve been.
Doing this in practice is partly a data problem, but it’s also a design one. It’s an interesting challenge to create an interface that will both accurately interpret and respect the intent of a query and that will sometimes challenge assumptions rather than just serve back whatever’s asked. If somebody’s asking the system to do something that doesn’t quite make sense, we want it to guide them towards something that will. That’s part of what our guided AI approach is doing – opening up more flexible ways to ask a question in the first place, rather than assuming the first framing is the right one. But it’s a balance: if the user gives over all control, you lose the benefit of their judgment and insight. It’s about having the right experience and the right interplay to benefit from both aspects of that, inside a very generative environment.
What made Benchmarking the first priority use case for initial release?
It’s very hard to make decisions about where you want to go without being able to understand the context you’re existing in right now. At the core, that means being able to look at a cohort and compare it with other cohorts. Benchmarking is the gateway to doing that.
A lot of benchmarking has been quite rigid and structured in the past, so the part of benchmarking that really excites us is being able to deepen our capability here and move into more nuanced spaces – where you could ask more fine-grained questions and really use it to understand how work in specific fields is growing and changing.
And the next use case will be Expertise – what’s the thinking there?
Universities are nothing if not their people. Being able to centre in on those people and understand their work, and help connect them to the right opportunities, is again one of those fundamental capabilities when we talk about helping make decisions. It’s really important that if we’re thinking about people, we’re taking the time to actually check in on the current data – what they’ve been doing, what they’re doing now – not just our perceptions of what people are doing.
Given the platform does surface data around people, how have you thought about the balance between what the AI does and where human decision-makers stay in control?
I think that’s a really important point, and it’s something that we can only address through constant attention and vigilance as we design the product. Our jumping-off point has to be human in the loop, always – not just in the loop, but in the driver’s seat.
It’s not appropriate for us to build systems that make those decisions for our users. But if we can streamline their work, get them the information they need quickly, let them interrogate the underlying data and ensure they trust its provenance – then they can spend the time they’d have used compiling all that on thinking it through and making informed decisions.
Why is the development partner and early adopter community so central to how we’re building?
Real, on-the-ground feedback from development partners and early adopters is key to making sure we’re building the right things – our roadmap needs to stay flexible so we’re learning and adapting in real time, not just following our own read of the landscape.
I’m coming up to ten years at Digital Science, and before I joined I was a member of the Digital Science community. I was the system owner for both Symplectic Elements and Altmetric at the University of New South Wales. One of the things I always really valued about the people I worked with at Digital Science was the voice that I had as a community member, and the opportunity to contribute my feedback from life on the ground, of living and using the system. And that’s something that I’ve carried forward with me through my time at Digital Science, as a really important part of fundamentally how we make sure that we are building things that really do solve problems for our customers and deliver value in their day-to-day.
When you’re talking about solving a problem, the best thing you can do is spend some time with people who live with that problem every day, understand their experience, and use that broader context to think about how technology can help make people’s days a little easier. There is no substitute for lived experience.
To find out more about Dimensions Research Strategy, read our launch blog here.