Ways to ‘see’ a system: tracking innovation trends
www.nesta.org.uk/blog/ways-to-see-a-system-tracking-innovation-trends/
www.nesta.org.uk/blog/ways-to-see-a-system-tracking-innovation-trends/
If you want to change a complex system, whether that’s our food environment or how our homes are heated, it’s critical to surface the latest developments in research, technology and public discourse.
Understanding the dynamics in a system is a first step in directing future innovation towards whichever social challenge you’re working on.
This is the case whether you’re a funder of research setting out to design a new programme of grants, an investment fund backing climate tech start ups or you’re a Whitehall civil servant working out how to crowd in private and public investment around a policy mission.
Whilst everyone's end goals might vary, they are all going to be asking similar questions:
Without reliable answers to questions like those, it is hard to target limited resources at the right things for maximum possible impact. You might miss an emergent, poorly funded innovation - or accidentally duplicate resources.
To date, ‘mapping systems’ has generally meant piecing together information through a painstaking qualitative process. While this can be one vital way of pinpointing the dynamics at play and the points of potential leverage, it also creates a static - and fairly subjective - snapshot.
Over the past few years in Discovery, we’ve been exploring how to create more actionable insights about the direction of innovation in a given system. This agenda has now come into sharper focus thanks to a surge of interest in ‘mission-driven government’ in the UK and across the world.
Missions in domains such as healthcare or industrial strategy often involve policymakers attempting to steer the path of innovation towards particular social problems, whilst trying to crowd in private sector funding too. More timely, accurate insight into the system we’re seeking to shape can be an extremely useful compass for people doing this kind of work.
However, too often, we have a tendency to look at how a system is changing from a single, narrow perspective depending on our vantage point as a funder, policymaker, entrepreneur etc. We also tend to rely on anecdotal evidence, or insights from historical research which might support our own views (well known to behavioural scientists as ‘confirmation bias’).
But it is an essential task - if a far trickier one - to build up a quantifiable multi-dimensional picture of the system we’re seeking to influence. It’s vital because we know that socio-technical transitions (such as changes in our food system) are influenced by a multitude of factors.
These range from breakthroughs in R&D, to policy developments, business strategies, media coverage and public acceptance. Because data about these domains comes in so many different forms, looking across them in parallel can leave us in the position of trying to compare apples with oranges.
To overcome this challenge, our team has been honing an experimental approach called Innovation Sweet Spots, which harnesses the latest advances in data science and machine learning to detect emerging innovations across Nesta’s missions in health, early years and decarbonisation.
Our approach involves analysing trends in research publications and public funding, patent applications, VC investment, news media coverage and parliamentary debates. Because these datasets are heterogeneous and the kinds of innovations tracked are very different, we use a taxonomy of trends which is based on comparable measures, such as rate of growth and magnitude.
What constitutes an innovation ‘sweet spot’?
We’re looking for the convergence of strong research activity and significant growth in private investment alongside interest from policymakers and the media. Taken together, encouraging signals of this nature would point to a maturing, highly promising innovation or solution.
Innovation Sweet Spots: data-driven horizon scanning approach
Anticipating a ‘tipping point’ for innovations such as heat pumps can help us to anticipate the timings of societal shifts - such as the post-carbon transition. This is similar to the idea of critical mass, where a process (such as the adoption of an innovation like mobile phones) takes off, hitting an inflection point and becoming self-sustaining.
Tipping points, critical mass, inflection points - foundational concepts in the field of systems thinking - are related to the idea that even relatively small actors can have an outsize impact if they can identify the right points of leverage.
Ten years ago, it simply would not have been possible to map activity with this degree of breadth and depth. Specifically, artificial intelligence (AI) and machine learning have now made it possible for public and social sector organisations on a budget to incorporate and analyse large-scale international datasets which are being updated continuously.
This presents a real opportunity to tease out coherent stories of change from noisy, complex systems.
We analyse various large datasets to capture different trends in innovation in each of Nesta’s mission areas. To date we have investigated trends in low-carbon heating innovation and food technologies, and we’re now researching trends in technologies to support early child development.
The choice of datasets used across these projects is inspired by Everett Rogers’ framework on the diffusion of innovation. To capture research trends in the UK, we use the open-access Gateway to Research database. It contains information about projects supported by UK Research and Innovation, which is responsible for around half of total UK public spend on research and development.
We also examine global research trends by using OpenAlex, a large, open-source dataset of publication data, including research paper summaries and metadata such as the authors’ affiliations and number of citations.
Looking at patent data allows us to cover innovations that are closer to commercialisation. For this purpose we have explored platforms like PatSnap as well as open-source data from Google Patents database.
To track investments into businesses, we use the Crunchbase database, which contains information about the amount of money raised by companies across the world from investment deals.
We complement the insights about research and investment with ‘public discourse analysis’ to better understand how technologies are being talked about in the public arena.
We have used openly available records of parliamentary debates (Hansard) and the open access platform of The Guardian news website - and we will explore other sources for analysing media coverage in the future.
One of the main tasks is to identify instances of technologies or innovations in the research, investment and public discourse datasets.
In our first iterations of this approach, we experimented with unsupervised machine learning methods, such as clustering and topic modelling, as well as simpler keyword searches.
Now, we are also using supervised machine learning for training classifiers (machine learning algorithms used to organise data) of different technologies as this gives us the greatest control to tailor the results specifically to our mission areas. Supervised learning, however, requires labelling a lot of data and so we are also trying out generative AI (large language models) for this purpose.
To interpret and communicate the trends around research, investment and public discourse, we have prototyped a data-driven typology that categorises these trends as dormant, emerging, hot or stabilising.
Our typology takes into account the magnitude (average level) and growth (eg, increase in the past five years) of time series related to a particular technology, such as the number of new research projects or the amount of investment across different years.
Growth is one of the main attributes when evaluating emerging technologies, whereas magnitude can help us see how established the innovations and technologies are and where they might be in their life cycle.
Whilst data science can illuminate the patterns in a system, on its own it cannot answer the ‘why’ or the ‘so what’ behind the trends.
Once the trends are identified, we also draw on the wider toolbox of strategic foresight methods to explore how these trends might play out for society. This helps us to critically appraise the hype cycle. Simply because a technology is ‘hot’ does not mean that it will be impactful, or indeed that the net effect will be positive for society.
For example, when we explored innovation in food systems, we combined quantitative analysis with qualitative 'sensemaking'. This included surveying a panel of investors, scientists, industry experts, and policymakers on the downstream social impact of the innovations surfaced through the analysis (for example, food reformulation), and unpacking the potential impact on obesity specifically.
The approaches we’ve been testing for Innovation Sweet Spots have an ever-expanding range of applications. In the longer-term we see potential to help social impact investors, policymakers and foundations to:
We want to explore how best to provide decision-makers with rolling insights (instead of a one off snapshot which becomes dated) so we are testing ways to make horizon scanning less manual and more cost-effective to do in real time. You can read more about our ongoing internal experiment to automate horizon scanning in this Discovery article.
Please stay tuned for the next edition of Innovation Sweet Spots on innovations aimed at supporting child development needs. If you’d like to stay in touch with this work, please contact Karlis Kanders.
