Social Serendipity

Mobile phones have been adopted faster than any technology in human history and now are available to the majority of people on Earth who earn more than $5 a day. More than one billion mobile phones were sold during 2003, six times as many as the number of personal computers sold that year [BBC (2004)]. Such an infrastructure of handheld communication devices is ripe for novel applications, especially considering their continual increase in processing power. And while digital communications has enabled everything from telecommuting to long-distance relationships across different continents, it has done little to encourage interactions of collocated people. In this paper we describe an architecture that leverages technology designed for communication at a distance to connect people across the room, rather than across the country.

The recent universality of mobile communication devices, combined with the growth of online introduction systems, facilitates an opportunity to generate entirely new types of applications. Although never intended as such, many devices that incorporate low-power wireless connectivity protocols such as Bluetooth that can be used as beacons to identify a user to others nearby. Our application leverages this phenomenon to facilitate dyadic interactions of two physically proximate people through a centralized server. A survey of fifty mobile phone users showed that if it becomes possible to instigate introductions to nearby strangers with similar interests using their phone, 90% of the respondents would use the service regularly. We present such a system, and have named it Serendipity.

The Opportunity

Serendipity consists of a central server containing information about individuals in a user's proximity and several methods of matchmaking. These profiles are similar to those stored in other social software programs such as Friendster and Match.com. However, Serendipity users also provide weights that determine each piece of information's importance when calculating a similarity score. The similarity score is calculated by extracting the commonalities between two users' profiles and summed using user-defined weights. If the score is above the threshold set by both users, the server alerts the users that there is someone in their proximity whom might be of interest. The thresholds and the weighting scheme that defines the similarity metric can be set on the phones and correspond to the existing profile types such as meeting, outdoors, silent mode, etc. When it has been determined that the two individuals should have an interaction, an alert is sent to the phones with each user's picture and a list of talking points.

Once we have quantified the social network amongst the subjects, we then would like to have a method for connecting people who aren't already friends. The Serendipity application was developed to facilitate dyadic interactions of two physically proximate people though an introduction situated in the immediate social context. A central server contains profiles about individuals in a user's proximity and several methods of matchmaking. These profiles are similar to those stored in other social software programs such as Friendster and Match.com - however, they can also contain (at the user's discretion) implicit information about an individuals: sleeping schedule, common hang-outs, inferred friendships, even the usage of games like "Snake". If a user-defined similarity score is above the threshold set by both users, an alert is sent to the phones with each user's picture and a list of talking points. In response to the feedback from an initial trial with forty participants and over one hundred introductions, our next version will incorporate several other introduction techniques, such as the approach described by Terry et al., which relies on a mutual friend to make the introduction [Terry et al. (2002)]. Alternatively, to preserve a user's privacy and to minimize disruption we also have enabled a feature of sending only an anonymous text message alert that there is a person nearby who shares similar interests; both users must respond "yes" to actuate the dissemination of any personal information.

Enterprise. Although static employee surveys can be easily analyzed, the output reflects a severely limited view of an organization's social network. We propose that the dynamics of the social network can be inferred from proximity data. Examples of the possible significance of BlueAware include the ability to automatically build a network model of the individuals within an organization, in order to quantify the effects of, for instance, a management intervention. Additionally, incorporating Serendipity into the workplace could instigate synergistic collaborations by connecting people who may be working on similar material, or someone who may have related expertise to another employee's current problem. Finally, forming groups based on their inherent communication behavior rather than a rigid hierarchy may yield significant insights to the field of organizational behavior.

Dating. The growth of online dating has soared over recent years as the stigma associated with personal ads diminishes. Serendipity provides users an alternative to encounters with people that they have only seen on a computer screen. Although we need many more users than our current number in order to test the efficacy of Serendipity as a dating tool, we are dialoguing with several online dating companies about the possibility of integrating a similar system in their own product line involving millions of active participants.

Conferences. It has been well established that there is a need for introduction systems at events such as large conferences and tradeshows [Borovoy et al. (1998)]. Salesmen can generate their own proximity webpages similar to the one described above to publicize their products and expertise (rather than interests and photos). Conference participants can customize their profiles to only be connected with individuals who can address their specific area of interest. As we have shown during the initial deployment in May, Serendipity can be an effective tool for networking at conferences.

Implementation

While we have a client running on all 100 phones that does repeated Bluetooth scans, Serendipity depends on have a critical mass of users. To expand the number of potential people beyond simply those with a specific model of phone, we have deployed Bluedar throughout many social setting settings on campus (the Media Lab coffee machine, the Sloan student lounge, the Infinite Corridor, the Muddy Charles pub…). These devices repeatedly scan for any visible Bluetooth device and send back the device type and unique hardware address to our server. This enables anyone with a Bluetooth device to participate in Serendipity by simply registering the BTID of the device and linking to an online profile. When Serendipity detects two people are nearby each other at the coffee machine, it sends an introduction message to each person's phone. These introductions are not designed to create strong links between individuals, but rather serve as ice-breakers. These informal interactions have been recognized as a valuable source for professional "weak ties" - critical for individuals as well as organizations. It has been shown that an individual's opportunities for upward mobility in society are frequently the result of these types of relationships. In organizations, a social structure that incorporates extensive weak ties is thought to maximize cohesiveness and adaptability to change while encouraging cross-organizational internal collaboration [Granovetter (1983)].

User Studies

Serendipity was initially deployed in early May 2004 at an elite conference consisting of senior corporate executives and professors. Personal profiles were created for forty of the conference participants who picked up their assigned phone upon arrival in the morning. Over one hundred introductions were made over the course of the day, primarily during the intersession coffee breaks. As it was the first time the system was deployed, a significant amount was learned about these types of situated introductions that helped refine the system in subsequent versions.

The conference setting necessitated several modifications from our original design of Serendipity. Because all the subjects were proximate to each other during the talks, it was necessary to develop a method for preventing introductions to be made while the talks were progressing. Simply hard-coding the conference break schedule into phones was not advisable due to the uncertainty in the talk lengths as well as the fact that it would then also prevent introductions between people who both happened to be outside during a particular talk. Instead, we were able to use several personal Bluetooth devices of our research group to prevent these unwanted introductions. We had volunteers disperse themselves throughout the auditorium each carrying a visible Bluetooth device whose name was changed to "BLOCK". Any of the forty phones inside the auditorium during the talks were able to detect at least one of these "BLOCK" devices. When this name was detected, the Serendipity application was paused and no information was recorded about devices in proximity or sent to our server.

While we succeeded in preventing introductions during the talks when we knew they were not appropriate, we had not taken into account the density of people mingling during the breaks. Several users complained of receiving multiple introductions to people within only a few minutes of each other. This led to a social disruption as one conversation was just getting underway, another conversation was initiated. One user solved the problem by simply turning his phone off while in conversation and then turning it back on when he was ready to meet someone else. In our subsequent version of the software we formalized this feature as "Hidden Mode" as well as imposing a maximum of receiving one introduction every ten minutes.

Some other surprising results included many users who were working for large corporations appreciating being introduced to other coworkers in the same company. For a couple of the participants, the introduction component of the application was not clear; they did not know what the picture messages about people nearby were meant to accomplish. However, besides the comments about the disruption of multiple introductions, the initial user feedback was primarily positive. Most of the initial subjects did not voice any privacy concerns, however this turned out to be not the case for a longer longitudinal study that is scheduled to last for the duration of the 2004-2005 academic year.

Currently Serendipity is running on the phones of one hundred users on an academic campus. Seventy of the users are either students or faculty in the same technical lab, while the remaining thirty are incoming students at the business school adjacent to the laboratory. We are currently receiving information from the devices regarding the other subjects typically observed over the course of the day. The profiles of users from the technical lab are currently bootstrapped from information available within their public project directory. Users also have the opportunity to input personal information and change any aspect of their profile.

Enterprise Introductions

Today, knowledge management is a $5 billion industry, but despite the benefits of such systems, most people interact with the social software in the isolation of their offices. That, however, might soon change with the growing popularity of mobile applications that support the desire of individuals to affiliate with others. Such technology could enable companies to untether knowledge management systems from the desktop so that they can be used in social situations where they might be most beneficial: nearby the water cooler, in the hallway, around the coffee machine. For example, when a critical mass of people in the same group gather together in the lounge, we have a setting that automatically messages the remaining members of the group that an ad hoc meeting may be taking place. These alerts can be for additional aspects of lab life as well. For example, it is now possible to be sent an email to a list every time the number of people in the kitchen exceeds a threshold. This has turned out to be a good indicator that there is free food.

Privacy Concerns

According to a forecast by the International Data Corp., nearly 80% of all mobile phones will have Bluetooth capability by 2006. If that prediction holds true, applications like Serendipity would have the potential to dramatically transform the ways in which people meet and connect with each other. As technologies converge, new cellular phones can identify each other with Bluetooth and can recreate the functionality of the Lovegety by leveraging the information already stored in existing online profiles. For that to happen, though, researchers need to address a number of privacy concerns. Specifically, many people might prefer eating their meals or riding an elevator in the silence of their own thoughts, and they could easily take offense at having their movements tracked by a Big Brother-like system.

Obviously, an application like Serendipity introduces a significant number of privacy concerns that must be addressed. In the research project at the Media Lab, all subjects will have given their explicit consent to participate and will know that, when their device is consciously turned to "visible" mode, others will be able to detect their presence. If users want to prevent their phones from logging data, they could simply choose the "invisible" mode. (But, of course, if everyone were to do so for extended periods of time, then that would defeat the whole purpose of the study.) In addition, centralized (instead of peer-to-peer) control helps ensure that people share only the information that they want to share. With Serendipity, a server helps mediate which people have access to certain data. A user might, for example, specify that certain pieces of information be shared only with those who have the same interests. Or the user could specify a hierarchical level of sharing (with friends, for instance, but not with friends of friends).

In general, companies have found that people are usually willing to relinquish a portion of their privacy in exchange for something. Consumers, for example, have been willing to divulge personal information (such as the names of their friends and relatives) to receive complimentary benefits (such as free gifts or reduced rates for a service). Within an organization, that type of quid pro quo arrangement could take various forms. Employees might, for example, be compensated (in financial or other terms) for playing active roles as intermediaries. Already many companies offer small bonuses (or "finder's fees") to employees who refer their friends and acquaintances to fill certain jobs at their organizations. Such approaches could help applications like Serendipity gain widespread acceptance within a corporate setting. If that were to happen, the technology would finally enable social software to be used where it could potentially have the greatest benefits - in social settings.

Beyond Serendipity

Bridging social software introduction systems with current mobile phone technology enables a diverse suite of applications. Conference participants will be able to find the right people during the event; large companies interested in facilitating internal collaboration could use Serendipity to introduce people who are working on similar projects, but not within one another's social circles; single individuals could go to a bar and immediately find people of potential interest.

Proximity Webpages. The application provides the user the option to view any information a proximate person has deemed public, regardless of their similarity score. While most interactions instigated by Serendipity require information to be sent to both users, proximity webpages allow users to simply see public profiles of nearby people without disclosing information about themselves.

Alternate Introduction Mediation Techniques. Although the current matching algorithm simply looks at similarity thresholds and scores described above, there are many other methods of matchmaking. One such approach described by [Terry et al, (2002)] relies on a mutual friend to make the introduction. Such a method can be incorporated into Serendipity by alerting the mutual friend rather than the two individuals. Alternatively, to preserve a user's privacy and to minimize disruption we also have enabled a feature of sending only an anonymous text message alert that there is a person nearby who shares similar interests; both users must respond "yes" to actuate the dissemination of any personal information.

Human-Machine Interactions. By equipping physical infrastructure with embedded computing and a Bluetooth transceiver, a variation on this system can be used to notify human users of nearby resources or facilities. For instance, the system can notify the user of an approaching free taxi, or a nearby public restroom. If instead of human users we consider mobile platforms with embedded computers (e.g., trucks, buses) we can envision other applications. For instance, busses could wait until passengers from other busses had gotten on-board, or delivery vehicles could more efficiently service pickup/drop-off requests.

Role-Based Access Control (RBAC) is a technique used to assign user permissions that correspond to functional roles in an organization [Sandhu et al. (2000)]. By capturing extensive user behavior patterns over time, our system has the potential to infer not only relationships between users, but also their permissions. For example, if two students working in different labs begin Tuesday collaborations at a coffee shop, they should be granted constrained entrance access to each other's lab.

Public Release of Serendipity. While Symbian Series 60 phones have become a standard for Nokia's high-end handsets, they represent a small fraction of today's Bluetooth devices. We are in the final stages of developing a MIDP (Java) version of the BlueAware application that will run on a wider range of mobile phones. The final test of Serendipity will be its public launch on www.mobule.net. We hope that not only will the application prove to be robust, but also quite popular within the realms described above, as well as those unanticipated.

Our society is more connected than ever before due to two parallel paradigm shifts in computing: movement from desktop to mobile computing, and from individual to social software. Mobile phones have become standard attire across the globe. In millions of pockets and purses are wireless transceivers, microphones, and the computational horsepower of a desktop computer of just a few years ago. Today the majority of this processing power goes unused. However, once the emphasis of mobile applications shift towards supporting the desire of individuals to affiliate with others to achieve their personal goals, this will soon change. We are catching glimpses of introduction services with the advent of online dating and knowledge management, yet the real potential of these new applications will be realized by an infrastructure of socially curious mobile devices, allowing us to untether social software from the desktop and imbue it into everyday life. It is our belief that the mobile phone market is at a critical tipping point where the functionality will shift from the traditional telephone paradigm to a much broader social-centric perspective. We hope that this work represents a step further in that direction.

N. Eagle and A. Pentland (2005), "Social Serendipity: Mobilizing Social Software", IEEE Pervasive Computing, Special Issue: The Smart Phone. April-June 2005. pp 28-34. PDF

Patent. M.I.T. Case No. 10705T, "Interaction Cueing Using Wireless Networking and a User Profile Database," by Nathan Norfleet Eagle and Alex Paul Pentland, United States of America. Filed May 6, 2004, "Combined Short Radio Network And Cellular Telephone Network For Interpersonal Communications"