Spatial analysis of collaboration

Recommended paper

Research from MIT’s Senseable City Lab investigates patterns of collaboration among MIT faculty by evaluating “output” (patents and papers) and “structure” (department and spatial organization) over a period of 10 years.

How might spatial proximity lead to greater collaboration? How can we use findings of spatial analysis and social networks to design greater outcomes for collaboration, such as in laboratories, workplaces, or schools? (Dare I say government?)

And — why is paper-based research more siloed by department, whereas patent-based research more cross-disciplinary?

Mural in Walker Memorial at MIT

An exploration of collaborative scientific production at MIT through spatial organization and institutional affiliation

Lead Researcher: Matthew Claudel; Advisors: Fiona Murray, Carlo Ratti



The analysis reported herein reveals clearly diverging trends in the composition of collaborative teams between papers and patents, with relatively stronger institutional collaboration for papers and more pronounced cross-disciplinary collaboration for patents. We then construct a multi-layered network of authors, and find two significant features of collaboration on campus. Firstly, a network topology and community structure that reveals spatial versus institutional collaboration trends among co-inventors. Secondly and most importantly, we empirically demonstrate a persistent relationship between physical proximity and intensity of collaboration, well fit with an exponential decay model as already observed in collaboration networks at a larger geographic scale. This relationship is consistently observable for both papers and patents, and is also consistent for exclusively extra-departmental collaborations. This is a particularly interesting result, in that it demonstrates the significant role that spatial proximity still plays in collaborative knowledge creation process–even at the micro (architectural) scale considered in this study, and despite the abundance of tools for digital communication and virtual collaboration.


A topology-based evaluation of collaborative networks reveals the composition and interaction of collaborative clusters. In the publication network, communities are more strongly related to departments, whereas in the patent network, communities are more heterogeneous overall, and have a tendency to align with buildings. This suggests that co-inventors organize around ideas or projects, while co-authors organize around disciplinary areas– a finding that reconfirms the statistical results related to team composition. Overall, these results reflect faculty members’ practical motivations for working together: co-inventors collaborate around projects, benefitted by shared equipment and a breadth of expertise, while co-authors collaborate within domains of scholarship to advance the knowledge of a particular subject.
Finally, we tested the effect of spatial proximity on collaboration. Regardless of co-location in a specific building, we found a persistent relationship between spatial proximity and the frequency of collaboration, well fit with an exponential decay model. Building on that observation, we suggest the possibility of a significant ‘distance limit’ for co-authors and an optimal ‘goldilocks distance’ between inventor pairs.
One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.