CRI Research focus - Open Science : collaboration networks
Network approaches for collaborative learning and solving
04 04 2020
CRI Research focus - Open Science : collaboration networks

Behind each complex system there is an intricate network that encodes the interactions between the system’s components”, Albert-László Barabási wrote in his book “Network science”. Society, for example, requires the cooperation between billions of individuals, while communication networks integrate billions of cell phones with computers and satellites… In line with this idea, Marc Santolini explores how collective phenomena emerge from complex webs of interactions between elementary parts. After a PhD in physics at ENS Paris, he did his postdoc at the Barabási Lab in Boston. In this lab that pioneered network science, he started to wonder:How could network science be used to better connect people and improve how we learn and solve problems together?One of his inspiration came from Michael Nielsen’s book Reinventing discovery. The author covers the advent of citizen science and how massive collaborations would require to “design serendipity”, i.e foster the hazardous encounter of people whose collaboration will accelerate knowledge. “If we allow people to share their projects, their results and their profile on a digital platform, we could then use this information to help them connect to the resources and people they need to advance their projects”, Marc Santolini continues.

These questions led him to create JOGL, or Just One Giant Lab, a nonprofit initiative hosted by the CRI. Co-founded with Thomas Laindrain (La paillasse) and Léo Blondel (Harvard) in 2016, JOGL is a platform that allows anyone to share results, find collaborators and contribute to cracking challenges. The aim is to build a decentralized research institute allowing everyone to learn and collaborate together in the digital age, overcoming the barriers of traditional institutions. Numerous projects have already been submitted to the platform: a mobile health platform to improve refugee’s health, for example, or an algorithm measuring vaccination hesitancy using social media data. Behind the scenes, there is a large network database where users are connected to the projects they are part of, the skills they have declared, or the users they follow. This information is then used to provide recommendations of what people they could interact with or what projects they could be helpful for.

In parallel to working on the design of this open science platform, Marc Santolini started several research projects to better understand how collaborations work. As a long term fellow at CRI, he is leading the project iGEM TIES (Team interactions study) with a team of postdocs, developers and interns. During the iGEM science and engineering competition, held annually, around 300 multidisciplinary teams of students work together to design projects using synthetic biology. How do these students collaborate and how do these interactions lead to better team performance? To answer these questions, Marc’s team extracted data from open Wiki pages from more than 2,000 teams that have participated in the competition over the past 10 years. From these wikis, they could reconstruct the team interactions and study their structure and dynamics. Moreover, they are now recording real-world interactions of team members using a bluetooth-enabled smartphone app they have developed. The measured network features have been associated with team success (medal, prizes, finalists) to explore the types of collaboration underlying team performance. “What distinguishes successful teams is their propensity to collectively collaborate on a large number tasks, with a bursty dynamics of interactions” Marc Santolini explains.“In a way, we are doing quantitative scientific anthropology. It's a modern version of what Bruno Latour was doing in the 80’s, when he was sitting in laboratories as an anthropologist to study how social structures affect scientific output.” While the iGEM study focuses on small teams (10-20 members), Marc then wondered what happens with large teams of 100, 500 or 1000? His team tackled this question of massive collaborations by studying open-source communities on GitHub, a platform where developers build and design software. “When a community exceeds 100-150 people, all projects adopt a “fractal” work-sharing structure, where leadership –defined by the largest number of contributions- over tasks is nested over various scales. For example, at the global level, someone overlooks the whole project, but if you zoom in at the local level of a subtask, for example the user interface part of the project, you observe the same leadership structure with someone overlooking all elements from that specific part. This fractal behavior is typical of self-organization phenomena observed in physics since the 80s”.

Beyond team collaborations, his teams is also tackling learning and innovation. For example, in partnership with Orange labs, they are studying how knowledge propagates across learners using a fine-grain phone call dataset collected during a 6 months training in Madagascar. Using models borrowed from epidemiology, they describe a contagion process where engagement and performance in the training is “transmitted” by social interactions. Another line of research focuses on the dynamics of knowledge production. By analyzing the scientific publications from arXiv, an open archive of electronic preprints of scientific articles, they explore the universal patterns behind scientific innovation. “We observed that all research fields follow a similar rise-and-fall life curve. By studying the early stages of a given field, we find shared facets of the scientific innovators. For example, they tend to be at the beginning of their careers, work in small teams or have a highly multidisciplinary profile”. Overall, these various projects tackle the fundamental questions of how people work together, innovate and learn, with the final goal to guide the design of the JOGL platform of open science innovation.

As a junior scientist and team leader, Marc compares a young research team to a start-up: “You are suddenly drawn into a world where you need to wear an ever-increasing number of hats : executive, administrative, hiring processes, operational, managerial, strategic, communication…” While he has found at CRI the freedom to work on topics he is passionate about, he recognizes that “building a team has been an incredible challenge”.

More news