Benjamin N. Waber

Mining Face-to-Face Interaction Networks using Sociometric Badges: Predicting Productivity in an IT Configuration Task

Social network theories (e.g. Granovetter 1973, Burt 1992) and information richness theory (Daft & Lengel 1987) have both been used independently to understand knowledge transfer in information intensive work settings. Social network theories explain how network structures covary with the diffusion and distribution of information, but largely ignore characteristics of the communication channels (or media) through which information and knowledge are transferred. Information richness theory on the other hand focuses explicitly on the communication channel requirements for different types of knowledge transfer but ignores the population level topology through which information is transferred in a network. This paper aims to bridge these two sets of theories to understand what types of social structures are most conducive to transferring knowledge and improving work performance in face-to-face communication networks. Using a novel set of data collection tools, techniques and methodologies, we were able to record precise data on the face-to-face interaction networks, tonal conversational variation and physical proximity of a group of IT configuration specialists over a one month period while they conducted their work. Linking these data to detailed performance and productivity metrics, we find four main results. First, the face-to-face communication networks of productive workers display very different topological structures compared to those discovered for email networks in previous research. In face-to-face networks, network cohesion is positively correlated with higher worker productivity, while the opposite is true in email communication. Second, network cohesion in face-to-face networks is associated with even higher work performance when executing complex tasks. This result suggests that network cohesion may complement information-rich communication media for transferring the complex or tacit knowledge needed to complete complex tasks. Third, the most effective network structures for latent social networks (those that characterize the network of available communication partners) differ from in-task social networks (those that characterize the network of communication partners that are actualized during the execution of a particular task). Finally, the effect of cohesion is much stronger in face-to-face networks than in physical proximity networks, demonstrating that information flows in actual conversations (rather than mere physical proximity) are driving our results. Our work bridges two influential bodies of research in order to contrast face-to-face network structure with network structure in electronic communication. We also contribute a novel set of tools and techniques for discovering and recording precise face-to-face interaction data in real world work settings.

A Human–Computer Interface Using Symmetry Between Eyes to Detect Gaze Direction

In the cases of paralysis so severe that a persons ability to control movement is limited to the muscles around the eyes, eye movements or blinks are the only way for the person to communicate. Interfaces that assist in such communication are often intrusive, require special hardware, or rely on active infrared illumination. A nonintrusive communication interface system called EyeKeys was therefore developed, which runs on a consumer-grade computer with video input from an inexpensive Universal Serial Bus camera and works without special lighting. The system detects and tracks the persons face using multiscale template correlation. The symmetry between left and right eyes is exploited to detect if the person is looking at the camera or to the left or right side. The detected eye direction can then be used to control applications such as spelling programs or games. The game ldquoBlockEscaperdquo was developed to evaluate the performance of EyeKeys and compare it to a mouse substitution interface. Experiments with EyeKeys have shown that it is an easily used computer input and control device for able-bodied people and has the potential to become a practical tool for people with severe paralysis.

Productivity Through Coffee Breaks: Changing Social Networks by Changing Break Structure

In this paper we present a two-phase study undertaken to experimentally study in a real world setting the effects of social group strength and how to increase the strength of groups in the workplace. In the first phase of our study we measured interactions between workers at the call center of a large bank based in the United States using Sociometric Badges. We confirmed our hypothesis that the strength of an individual’s social group was positively related to productivity (average call handle time) for the employees that we studied. In the second phase of our study we show that by giving employees breaks at the same time we increased the strength of an individual’s social groups, demonstrating that low-cost management decisions can be used to act on these results.

Organizational Engineering Using Sociometric Badges

We show how a wearable computing research platform for measuring and analyzing human behavior can be used to understand social systems. Using a wearable sociometric badge capable of automatically measuring the amount of face-to-face interaction, physical proximity to other people, and relative location, we are able to construct a dynamic view of an organizations social network by viewing interactions as links between actors. Combining this with email data, where e-mail exchanges indicate a social tie, we are able to form a robust view of the social network, using proximity information to remove spurious e-mail exchanges. We attempt to use on-body sensors in large groups of people for extended periods of time in naturalistic settings for the purpose of identifying, measuring, and quantifying social interactions, information flow, and organizational dynamics. We discuss how this system can lead to an automatic intervention system that could optimize the social network in real time by facilitating the addition and removal of links based on objective metrics in a socially natural way. We deployed this research platform in a group of 22 employees working in a real organization over a period of one month, and we found that betweenness in the combined social network had a high negative correlation of r = −0.49 (p

Sensible Organizations: Changing Our Businesses and Work Styles through Sensor Data

We introduce the concept of sensor-based applications for the daily business settings of organizations and their individual workers. Wearable sensor devices were developed and deployed in a real organization, a bank, for a month in order to study the effectiveness and potential of using sensors at the organizational level. It was found that patterns of physical interaction changed dynamically while e-mail is more stable from day to day. Different patterns of behavior between people in different rooms and teams (p < 0.01), as well as correlations between communication and a workers subjective productivity, were also identified. By analyzing a fluctuation of network parameters, i.e., “betweenness centrality, ” it was also found that communication patterns of people are different: some people tend to communicate with the same people in regular frequency (which is hypothesized as a typical pattern of throughput-oriented jobs) while some others drastically changed their communication day by day (which is hypothesized as a pattern of creative jobs). Based on these hypotheses, a reorganization, such that people having similar characteristics work together, was proposed and implemented.

Studying Microscopic Peer-to-Peer Communication Patterns

This paper describes first results of an ongoing research effort using real time data collected by social badges to correlate temporal changes in social interaction patterns with performance of individual actors and groups. Towards that goal we analyzed social interaction among a team of employees at a bank in Germany, and developed a set of interventions for more efficient collaboration. In particular, we were able to identify typical meeting patterns, and to distinguish between creative and high-executing knowledge work based on the interaction pattern.

Sensing Informal Networks in Organizations

We present an examination of informal network structure within the sales division of a global manufacturing organization. Sociometric Badges were used to collect data on face-to-face interactions over a total of 8 weeks, the latter half of which was spent in a redesigned workspace. These data were supplemented by employees’ e-mail and instant messaging log activity. The allocation of an individual’s communication among colleagues reflected the company’s structure as a post-bureaucratic organization. The observed interteam communication patterns differed from those expected to arise based on the various functions performed by each team throughout the sales cycle, suggesting that the communication needs of each team were not wholly provided for by the available media. A subset of workers who were encouraged to utilize flexible seating arrangements in a remodeled space had a higher proportion of face-to-face interactions with colleagues outside of their team, while employees seated far away from each other were less likely to exchange e-mail. This research has implications for companies hoping to understand the structure of informal networks within their organization as well as those considering workplace redesign as a method of stimulating communication within these networks.

Sensor-Based Feedback Systems in Organizational Computing

Radical change is needed in todays organizations. While e-mail, instant messaging, wikis, prediction markets, and the like have proliferated across myriad sectors, the fundamental practice of management has failed to keep pace. Sensors can automatically measure social behavior occurring in physical space as well as the virtual world. Moreover sensor-based feedback is poised to help create the change necessary to improve performance and satisfaction of workers. In this paper we summarize previous work on sensor-based feedback systems and propose new systems at the individual, group, and organizational level. Our goal is to help direct future research towards these promising avenues.

Understanding Organizational Behavior with Wearable Sensing Technology

We describe how recent advances in wearable sensing technology allow for unprecedented accuracy in studies of human behavior, particularly organizational behavior. We use one such platform, the Sociometric badge, to understand organizational behavior in two studies. In the first, we describe the collection of data over a period of one month in a German banks marketing division. We found that physical proximity had a high negative correlation with e-mail activity, and by combining behavioral data and electronic communication data we were able to very accurately predict self-reports of personal and group interaction satisfaction and performance. Next we describe an experiment at a data server configuration firm, and we discovered behavioral variables that had extremely high correlations with objective productivity measures. In both studies the fine-grained behavioral variables measured by the Sociometric badge played a critical role in predicting outcomes.

Web mediators for accessible browsing

We present a highly accurate method for classifying web pages based on link percentage, which is the percentage of text characters that are parts of links normalized by the number of all text characters on a web page. We also present a novel link grouping algorithm using agglomerative hierarchical clustering that groups links in the same spatial neighborhood together while preserving link structure. Grouping allows users with severe disabilities to use a scan-based mechanism to tab through a web page and select items. In experiments, we saw up to a 40-fold reduction in the number of commands needed to click on a link with a scan-based interface. Our classification method consistently outperformed a baseline classifier even when using minimal data to generate article and index clusters, and achieved classification accuracy of 94.0% on web sites with well-formed or slightly malformed HTML, compared with 80.1% accuracy for the baseline classifier.