Quentin Loisel is a Marie Curie doctoral fellow within the Health CASCADE project. His research focuses on bridging the gap between humans and technology in co-creation.
Quentin is part of a team that has been looking at how we can more efficiently review co-creation literature. A possible solution to collating this vast and varied body of work is through technology. In this blog, Quentin presents the use of an AI tool to help collect the relevant work that Health CASCADE can build on.
You wake up one morning and your curiosity overwhelms you: what is [insert object of your curiosity]? What do you do? Do you open a book, consult Wikipedia or watch a documentary? You think that someone must have worked on the subject! Researchers do the same.
To approach a new subject, they start by learning about all the scientific hypotheses, experiments and theories that have been formulated. They seek to acquire the knowledge already produced by other scientists before them. To do this, they will study all the books and papers available. We call this “doing a literature review”. The idea is that at the end of this process, the researcher should know the subject as well as possible so they can propose the next innovative research question.
We can therefore say that we are “Standing on the shoulders of giants”. This metaphor suggests that in order to progress intellectually, we must build on the work of our predecessors. In fact, knowing what has already been done allows us to avoid doing the same research and re-invent the wheel.
Source: https://www.slideserve.com/cid/on-the-shoulders-of-giants
We can therefore say that a literature review is about climbing a giant. Each reference read by the scientist is like climbing a step on the staircase that leads to the shoulders of this giant. Sounds cool, right?
So why am I telling you about this? Because, as you might have guessed, in order to do scientific research on « co-creation », we are going to have to climb the giant of the same name and for that we need a staircase. Because yes, to climb a literary giant, you must build a staircase of references!
Step 1: Finding the steps of our staircase
So, the steps of our staircase are the relevant references of our subject. How to get them?
The traditional method would be to ask a specific research question. Then, we define the relative keywords and use them to search for all scientific articles in databases. Many papers containing our keyword would then be suggested to us. In this large package, there are many materials that do not exactly address our topic. We have to sort them out! To do this, the method consists of doing a ‘screening’ and reading the titles and summaries of each reference. We then judge, one by one, whether they are relevant to our subject and we keep the best ones.
But we have two problems with this approach. The first one is the fact that there is not one but many definitions of co-creation and therefore a large number of relative keywords. This means that when you do a search, you end up with thousands. The second problem is that our consortium has not one but many research questions. Each of these questions explores the theme of co-creation from a particular angle. Nevertheless, the keywords used are very similar. We can expect that our researchers will find themselves sorting through many articles and probably the same ones for the most part.
The traditional approach is therefore not the most suitable for our subject but also for our common project. It is necessary to find a way to mutualize our efforts and create a database of all the articles relevant to the community working on co-creation. Everyone will then be able to search inside for material related to the subject that interests them concerning co-creation. To do this, we need to expand the keywords to include all topics related to co-creation.
So, instead of focusing on one issue, we try to approach an entire theme. Instead of narrowing our field of research, we open it up. It may sound simple when you put it that way, but it’s an innovative approach that involves thinking differently and building our own method. More importantly, it gives us a greater amount of material to sort through: no less than 150 000 references.
“We had a big bunch of steps of all kinds, but there were only a few in the pile that interested us. No choice, we have to sort it out!”
Step 2: Finding the steps of our staircase – among thousands of others
Do you see the problem? Take a couple of minutes to get the measure of the task. Imagine that you take one minute to read the title, the abstract and make your judgement. To finish your screening, you will have to work for 2,500 hours, or 62 working weeks, or 1 year and 2.5 months of work (40 hours/week, without holidays)!
Incredible, isn’t it? Yes, but it is a very common problem. In fact, every year the scientific community produces more scientific work than the previous year and we have an exponential growth in the number of articles. This makes it increasingly difficult for researchers to synthesise and update all this knowledge. This concern was especially felt during the covid-19 pandemic when the scientific community quickly mobilised to study the situation.
So how do we proceed? Get ten people to work on the project full time for three months? Hardly feasible or efficient. Besides the process of hand screening is not entirely fault proof.
Step 3: what if we ask technology for help?
The solution could come from artificial intelligence (AI) technologies. AI that would allow us to efficiently identify interesting material. Which would analyse new publications in real time and keep us up to date. A tempting prospect!
Source: https://asreview.nl
But even if the need to find technological solutions was felt all the more during the Covid-19 crisis, part of the scientific community remains cautious about using these technologies. This caution is understandable, as this is a young scientific area. Nevertheless, the results are very encouraging and many services have appeared.
One of them is ASReview. A free open-source AI developed by a team at Utrecht University in The Netherlands. The advantage is that this technology has already been successfully tested, which gives confidence in its effectiveness. When used optimally, we can expect to find 95% of relevant articles by screening only 10% of the article pool. We could go 10 times faster and do the task in 6 weeks. With a team, we could hope to accomplish our task in one big week!
Step 4: let’s get started?
So here we are.
Our giant “co-creation” is very big. We need a big staircase to climb it! But before that, we have to sort out the references to have only the relevant ones. Going through everything will take too long, and being too selective will mean we miss interesting material. But we have a solution to help us! An artificial intelligence tool. It will help us with the sorting! It will be our assistant.
We have a plan! Now we need to put together a dream team and off we go!