The Bartlett School of Architecture



The goal of the project "SocioBuildings" was to investigate the effect of a building's configuration on modulating people's interactions.

The project combined methods of direct observation of interaction patterns in space, an online administered self-reported survey and RFID sensing technology to capture face-to-face interactions in a single case study of a University Building in Cambridge.

Using three different methods for capturing interaction patterns allowed comparing the benefits of each method. Specifically, comparing traditional data gathering methods derived from the social sciences research community (self-reported surveys and direct observation by human observers) and an automated method commonly used in the computer science field (RFID sensors) was expected to shed light on the quality, rigour and validity of research methods.

Therefore the project plan included testing whether automated data captured by RFID sensors delivered comparable findings, complementary findings or contradictory findings to direct observations and self-reported surveys.

Additionally, the project aimed at investigating spatial and social rationales for interaction patterns and thus creating new insights into the spatiality, temporality and dynamics of interactions. The data collection period covered 10 full working days, thus providing a richer data set than typically explored in similar previous studies of direct observations.

In order to bring the spatial dimension into the analysis of interaction patterns in the workplace, the spatial configuration of the building was analysed using Space Syntax technologies, in particular visibility graph models. The distances between co-workers were modelled as axial lines of potential movement flow, broken into smaller segments to provide a more detailed account of proximity relationships between people in space.


This collaborative research project brought architectural researchers from UCL together with computer scientists at the University of Cambridge, and researchers of the SocioPatterns project at the ISI Foundation in Turin, Italy.

The UCL contribution was led by Dr Kerstin Sailer and supported by Rosica Pachilova as research assistant.

The team in Cambridge consisted of Dr Daniele Quercia, Dr Cecilia Mascolo, Dr Christos Efstratiou, Dr Ilias Leontiadis and Chloë Brown.

The project was supported by Dr Ciro Cattuto at the ISI Foundation.

The project was partially funded by an EPSRC Post-Break Award to Dr Kerstin Sailer to deepen insights on the interplay between social and spatial networks. 


The research found that variable degrees of overlap could be established between the two approaches (participant data collection versus automated data collection) with rather few comparable findings.

For certain space usage behaviours high levels of variance between the automated and manually collected datasets were found, pointing towards predominantly complementary and contradictory findings. It was shown that the goodness of the fit between automated and manual data depended on the way data was aggregated.

In summary, evidence suggested that both human and machine based data gathering revealed crucial insights into behaviours of building users. Substituting manual methods with automated ones could not be supported by the data of this study.

For a more detailed account of research results, please refer to the following publication:

Additional results were reported here:


To our knowledge this is the first systematic comparison of the performance of different methods of capturing interaction patterns in a built environment setting.

Discovering the ground truth of who interacts with whom how often in the workplace and how those contacts evolve over time is an intrinsically difficult task, since each method works with assumptions and has to acknowledge its own biases and limitations: for instance self-reported surveys are known to suffer from recall bias; direct observations can be considered problematic since observers intervene in the field, but also due to sampling and external validity issues; whereas RFID sensors may not capture each interaction due to signal noise and can suffer from accuracy and reliability problems.

This issue is far from being solved, yet this project provides an important first step to determine the degree of overlap of findings generated through different methods with an ultimate aim of improving the reliability and accuracy of research methods.

This has considerable impact for future research and research methods, but also for a wide variety of applications, specifically in indoor-location tracking and the associated development and use of sensors, which is a growing field of interest.

spatial analysis big data