As part of our new research programme we have used machine learning and literature search to map key trends in collective intelligence research. This helps us build on the existing body of knowledge on collective intelligence, as well as identify some of the gaps in research that can be addressed to advance the field.
At Nesta’s Centre for Collective Intelligence Design we are interested in the potential for collective intelligence (CI) to be amplified by new methods and emerging technologies. And in the spirit of collective intelligence, we want to ensure that our work complements and builds on the existing academic canon, addresses current gaps and challenges and takes full advantage of new research methods.
We think it’s important to open up the conversation beyond some of the narrower academic definitions of collective intelligence to include all forms of group intelligences (human and machine) acting at scale to achieve an objective. This is why we also consider crowdsourcing, citizen science and deliberative democracy in our working definition of collective intelligence.
By codifying this body of knowledge to build an evidence base for what works, we aim to bring rigour to the field and contribute resources for designing collective intelligence that can be used by ourselves and others to address some of the most complex issues in health, international development, environment and many more!
Much of the existing knowledge on collective intelligence is discoverable through online research and innovation databases which is where we’ve starting our mapping endeavour. In collaboration with the Innovation Mapping Team at Nesta, we’re using a machine learning enabled approach to understand how the collective intelligence research community describe their work and the kind of subjects that feature most often in the literature and project descriptions. By applying topic modelling to our dataset of 7000 items (open access articles, funding abstracts, company descriptions) we’ve identified prominent thematic clusters and their relatedness to each other.
The cluster diagram above shows the dominance of each topic within the corpus relative to the others (size of text) and how closely related the topics are by meaning (position and distance between text headings). We can see that the largest topic heading is crowdsourcing, indicating that this is most commonly cited type of collective intelligence. Other well represented topics include citizen science, sensor data, data quality and multi-agent systems.
By delving deeper into the research using traditional academic literature search we have identified four key trends that are currently shaping the field:
These early insights suggest that academics and practitioners working within and adjacent to collective intelligence have a lot to learn from each other across disciplinary boundaries. Whether you call it citizen science, crowdsourcing, deliberative democracy or collaborative learning… these are all great examples of collective intelligence in action.
This work complements our ongoing efforts to build a repository of practical projects and tools demonstrating collective intelligence. Together, these mapping studies are helping to shape the research agenda for the Centre. Some of the questions we hope to address in the next 12 months include:
In the coming months, we’ll be refining our research questions and supporting our Small Grants winners to answer others about designing collective intelligence. And of course we’d welcome your feedback, get in touch at [email protected]