The Science of Scialog: How Conference Design Sparks Collaborations
Well-designed interactions can help scientists form the collaborations that have become increasingly important in today’s research environment, according to a new study based on data from four series of Scialog conferences.
The paper, “Dynamics of social interaction: Modeling the genesis of scientific collaboration,” is published in Physical Review Research, a fully open-access APS journal. Northwestern University Ph.D. candidate Emma Zajdela is first author, and she and her adviser Daniel Abrams are corresponding authors. Co-authors are Research Corporation for Science Advancement’s Senior Program Directors Richard Wiener and Andrew Feig, RCSA Data Analytics Specialist Kimberly Huynh, and Northwestern undergraduate Andy Wen.
“The goal of Scialog workshops is to rapidly catalyze new collaborative teams among people who have never worked together before, and in most cases don’t even know one another, and get them to come up with really innovative projects that push the envelope in one scientific area,” said Wiener, who helped created Scialog in 2010 along with RCSA Senior Program Director Silvia Ronco.
“What we try to understand with this paper is whether that process is working in the generation of new collaborative teams,” he said. “It shows that we are effectively doing that.”
Scialog is short for “science + dialog.” Scialog meetings are designed to stimulate intensive interdisciplinary conversation and community building around a scientific theme of global importance. Teams of two to three researchers who have not previously collaborated compete for seed funding for high-risk, high-reward projects based on ideas that emerge at the three-day conferences.
“Scialogs offer an amazing testbed for theories of how scientific collaborations form,” said Abrams, a professor of applied mathematics. “Research Corporation has retained great data for years showing not just who co-attended sessions together, but also other key factors, like how well participants knew each other beforehand, and what alternative conference scenarios would have looked like (for comparison to the real-world scenario).”
Zajdela, who has been developing mathematical models to understand and predict complex social phenomena, said Scialog has been a fertile ground for her research.
“I have been fascinated for the past few years with the idea that mathematics can help us understand and even predict human behavior,” she said. “The incredibly rich and detailed data collected over the past decade by the organizers of Scialog have allowed us to develop a new mathematical theory for how people who interact during gatherings form teams. We imagine this theory could be extended far beyond Scialog to other contexts like education, business and even romantic settings.”
The paper analyzes data from four Scialog series – 12 conferences over the pre-COVID period from 2015 to 2019 going back to the second Scialog initiative. The first Scialog initiative was excluded because RCSA made significant changes to the structure of subsequent meetings.
“For me the big takeaway is that certain aspects of the workshops have more power toward generating collaborations than others,” said Feig. “The Northwestern team’s insight is to think about effective interaction time – the number of minutes in a room with somebody divided by the number of people who were part of that conversation.”
“The time spent in small groups, when structured around a task, was far more effective at generating the interactions necessary to launch a collaboration and build the trust the collaboration is founded on,” Feig said.
And what is the task? Discuss what it would look like to work together.
“That prompt is important,” Feig said. "It's not just ‘go sit and have coffee’ but really think about what science would make sense for a small group of people who don't know each other, given their respective backgrounds. That might not work all the time, but it works in the constructed atmosphere of Scialog because we've pre-screened and brought people together from a variety of disciplines and structured them into groups in a way that makes them likely to have something that they want to talk about.”
Designing group interactions at Scialog meetings begins with an initial survey of the approximately 50 Fellows chosen to participate in each conference. The survey asks how well Fellows are acquainted with other participants, as well as their interest in various topics. Survey responses are part of the mix of information used to find groupings to maximize varied conversations among researchers of different disciplines, genders, research approaches, and interests.
“Our approach is to get people to mix more, which we believe results in more creative science,” Wiener said. “We design Scialog so there are opportunities to go through a series of steps in just a couple of days, from being unaware of each other, to getting to know each other, to having significant scientific conversations and exchange of ideas, to wanting to work together.”
The model, which some participants have called “speed-dating for scientists,” moves people toward collaboration through a series of group discussions.
Medium-sized groups of about 10 Fellows and one or two discussion Facilitators hold 75-minute conversations, offering participants an opportunity to gain a general awareness of what type of science they do, their approaches, and their ideas. Mini-groups of three or four Fellows follow the larger sessions, allowing for more intensive scientific dialogue.
“Our research shows that Fellows who came together in these small discussion groups were 10 times more likely to write a proposal together,” Wiener said.
Data specialist Huynh uses a set of algorithms to tackle the complex computational problem of optimizing groupings and pairings at Scialog conferences to maximize the opportunity for everyone to meet as many new potential collaborators as possible.
"For small groups, we make sure they don’t previously know one another,” said Huynh. “For larger groups we make sure it isn’t the same people meeting over and over.”
Though this paper only looks at which Fellows choose to initiate a collaboration, Huynh is also mining longitudinal data on Fellows to help the team analyze the longer-term outputs of Scialog collaborations, which include the science they produce (represented by publications) and other grant support to advance promising Scialog projects.
RCSA surveys Scialog Fellows again each year to track how their knowledge of the rest of the cohort changes with time.
“What we’re seeing is a broader group of collaborations forming across the network,” said Feig. "Some people choose to work and publish together, whether or not they received funding at Scialog.”
Though the importance of the small-group interactions is striking, both Wiener and Feig said future analysis could shed more light on the impact of conference design.
“Because there are other opportunities for people to interact during a Scialog conference, some of those intensive discussions are happening more informally,” Wiener said. “How much is conference design driving those collaborations, and do the larger groups accomplish more things that aren’t necessarily reflected in the data?”
RCSA held eight virtual Scialog conferences from 2020 to 2022, collecting new data that could inspire future lines of analysis – especially around virtual conferences. Do interactions on Zoom spark collaborations the way in-person conferences do?
“That work is nascent, but we’re very excited about the recordings of discussions from our video conferences,” said Wiener.
“Because we were on Zoom, we had the ability to record all these conversations,” he said. “Now that we’ve moved back to in-person meetings, we are still recording so we can look at the differences between video and in-person interaction. We’re looking at some of the social dynamics within a conversation and how different people with different social capital or different personality types are engaged in conversations, as well as what impact that had on ultimately choosing to find partners for collaboration.”
These new data may enable the team to look at other questions to inform conference design: Who spoke more? Do gender, field or status of institution affect who spoke and for how long? Is speaking more or having an outgoing personality type favorable or unfavorable in finding collaborations? How do different types of facilitators involve participants who are quiet? Does the time facilitators spend on different themes affect the collaborations that are formed?
Zajdela and Abrams said they are now planning to study the tendency of people with similar backgrounds to socialize, or in this case, to form collaborations. This research could provide valuable insight into improving the diversity in science overall.
Both Wiener and Feig say the Scialog data could help answer critical questions about the advantages and disadvantages of various ways to host scientific conferences in the post-COVID world.
“People are very interested in doing virtual conferences because they have some clear advantages in cost, carbon footprint, and the ability to be more inclusive of participants,” Wiener said. “But the big question is: What do you lose?”
Whether choosing to go virtual, in-person, or hybrid, Feig said the Scialog analysis reminds those designing conferences to keep sight of their goals.
“We should always be asking ourselves if we are accomplishing what we think we are and how we should modify the structures we’re using to make them even more effective," he said.
“So much time and money are spent on scientific conferences,” said Wiener. “We can’t say what’s effective for all conferences, but we can say that this is an example of how to analyze whether what you’re doing is meeting your goals. And at RCSA we advocate, by example, doing that kind of analysis.”
Abrams said the study shows how important thoughtful conference design can be to the future of science.
“The way people are assigned to attend sessions at conferences has a significant impact on who collaborates with whom,” he said. “A small change to the itinerary at a conference may have impacts that go on for years.”
