New Model Uses Team Interactions to Predict Next Steps in Collaborative Work
For Immediate Release
Finding the right information at the right time is critical for solving complex problems. Researchers have developed an algorithm that helps individuals locate needed information more efficiently by drawing on both a user’s own history and behavioral cues, as well as the history and behavioral cues of teammates collaborating to solve the problem.
“This collaborative process of sharing information to solve problems is called social foraging – it’s a well-established concept in animal behavior research,” says Sandeep Kuttal, co-author of the work and an associate professor of computer science at North Carolina State University. “But until now, this behavior hasn’t been incorporated into software systems designed to support problem-solving. We think it holds promise for improving software used for collaborative work – such as software engineering, scientific research, or crisis response.”
“Previous algorithms look solely at the history of a single user,” says Kuttal. “Our goal is to consider both a user’s interactions and how their teammates interact with the system, in order to improve recommendations for the user.”
To achieve this, the researchers created a new predictive model called Programmer Flow by Information Scent for teams (PFIS-T) which incorporates both teammates’ activity in the system and explicit cues from communication between teammates to predict a user’s next action. PFIS-T was developed specifically for software engineering teams doing maintenance tasks – such as debugging, code reuse and feature development – but it could be adapted for other collaborative applications.
The researchers evaluated PFIS-T using data from a laboratory study involving 30 software engineers working in 10 three-person teams to perform code-maintenance tasks. The researchers compared the actual behavior of the three-person teams to both the behaviors predicted by PFIS-T and the predictions generated by a previous model that relied solely on an individual user’s history.
“We found that team cues were even more important than we anticipated,” Kuttal says. “The PFIS-T model performs best for teams whose members communicated frequently.”
Overall, PFIS-T predicted 81.5% of team navigations, improving accuracy by up to 16.7% over the model that looked only at individual user history.
“One takeaway here is that PFIS-T could complement AI-based next-step predictors by providing a socially grounded signal about how humans actually navigate complex information spaces,” Kuttal says. “We’re optimistic PFIS-T can improve existing tools and serve as the foundation for developing new ones.”
The paper, “Where Will They Click Next? A Social Foraging Model for Collaborating Teams,” will be presented April 17 at the ACM CHI conference on Human Factors in Computing Systems, being held in Barcelona, Spain. Corresponding author of the paper is Shahnewaz Leon , a Ph.D. student from NC State.
This work was done with support from the Air Force Office of Scientific Research under grant FA9550-21-1-0108, and from the National Science Foundation under grant 2313890.
-shipman-
Note to Editors: The study abstract follows.
“Where Will They Click Next? A Social Foraging Model for Collaborating Teams”
Authors: Shahnewaz Leon and Sandeep Kaur Kuttal, North Carolina State University
Presented: April 13-17, ACM CHI conference on Human Factors in Computing Systems, Barcelona, Spain
Abstract: Modern knowledge work is increasingly collaborative, especially in information-intensive domains such as crisis response, scientific discovery, and software engineering. Software engineering epitomizes these trends through practices like pair programming and collaborative debugging. Yet existing computational models of information foraging remain individual-centric, leaving teams without support for social foraging—leveraging partners’ actions and communication to navigate complex projects. We introduce PFIS-T, a predictive computational model of social information foraging. Building on the PFIS model family, it integrates implicit cues from teammates’ recent navigation and explicit cues from synchronous communication to predict a programmer’s next action. We evaluated PFIS-T with ten three-person debugging teams, finding that it substantially outperforms the strongest individual baseline, PFIS3, predicting 81.5% of navigations and improving accuracy by 16.7%. These results show how predictive models can operationalize social foraging and point to opportunities for collaborative IDEs and interactive systems that adaptively surface social trails to improve coordination and awareness.
