Network Connections Home
Thesis Presentation Message Board Wiki

Introduction  |  More is Different: Why "Social Software," Why Now?  |  Social Network Analysis: The Big Deal About Small Worlds  |  The Big Picture: Friendster, Ryze and Meetup  |  Case Study Analysis: Friendster, Ryze and Meetup  |  Directed Searching: Working the Network  |  Conclusions  |  Sources  |  Printer-friendly Version
Directed Searching: Working the Network

Before distilling my observations about site functionality into a series of conclusions, there is one topic, applying to Friendster and Ryze but not Meetup, so important that it must be broached separately: directed searching of a social network. While the fact that we are connected to a given person through a short chain of intermediaries may sound exciting, we are in fact connected to everyone on the planet by only a short chain of intermediaries. Without the ability to do a directed search over the network for the person you're looking for, the fact that you are only a few short hops away is not necessarily actionable knowledge.[62] The more we learn about small world topology, the more the result of Milgram's social experiment is interesting not because it demonstrates that we are connected by short paths of intermediaries, but because it reveals that in the real world somehow we are able to find them.

Milgram's missive said: "If you do not know the target person on a personal basis, do not try to contact him directly. Instead, mail this a personal acquaintance who is more likely than you to know the target person [my italics]."[63] So, what exactly is a person taking into account when determining which of their acquaintances would be "more likely to know" a specific person than they are? Of all the possible directions in which each person in the chain could have sent Milgram's letter, how could path selection have been so accurate that, on average, successfully completed chains reached their target in fewer than six steps? It follows that we must employ methods of interpreting the social information at our disposal to make our search more likely to hit the desired mark.

The question is what kind of social information provides the context people need to be able to accurately gauge social direction and distance? It's a question of the utmost importance for the designers of "who you know" social software, like Friendster and Ryze. The value of such software lies in being part of a robust virtual social network, but also in having the tools to traverse the network in an efficient way. The efficacy and efficiency of the search process is key to the success of the software. Without a practical means of executing a directed search you are not traversing the network, but trawling it -- casting a net as far and wide as you can and dragging it across the network in the hopes of finding what you're looking for. Cast. Drag. Repeat. Trawling is a far cry from traversing; it might turn out to be effective, but it's certainly not efficient. So, to determine whether the transaction cost of searching the network is sufficiently low, we need to ask if the social information provided by Friendster and Ryze equips a user to trawl, to traverse or both.

As it turns out, when it comes to homing in on a person's location in social space, people have a very strong sense of distance and direction, and it is measured not by square miles or hop count but by spheres of affiliation. Watts says:
" real social networks, individuals possess social identities. By belonging to certain groups and playing certain roles, individuals acquire characteristics that make them more or less likely to interact with one another. Social identity, in other words, drives the creation of social networks."[64]

So, the social contexts in which we participate--groups, institutions, activities--determine to a large extent how we identify ourselves and others, and therefore it is through the lens of these affiliations that we make sense of social space.[65] Milgram himself alluded to the primacy of social structure over network structure in human cognitive processes when he said, "Social communication is sometimes restricted less by physical distance than by social distance...We should think of the two points as being not five persons apart, but five 'circles of acquaintances' apart -- five 'structures' apart."[66]

These structures are the social dimensions created by our affiliations: geographical, racial, professional, educational, organizational, etc. Depending on who you're looking for and what you're trying to achieve, different social dimensions will be the best way to assess social nearness. Usually, however, it is the consideration of more than one social dimension that enables people to ascertain the best path toward someone to a high degree of accuracy. Knowing someone lives in Arizona would not help me too much in finding a path to him or her. However, knowing that they live in Arizona, went to Oberlin and work in new media would help me. I have no personal affiliation with Arizona or anyone from there, so I would disregard that as a social dimension. However, I know several people from high school who went to Oberlin (almost went there myself) and I currently work in new media. The latter two social dimensions would be the contexts I would activate to help me find my target. Indeed, after studying affiliation networks and creating a social network search model, Watts and Strogatz concluded:
"As long as individuals are more likely to know other people like them (homophily), and--crucially--as long as they measure similarity along more than one social dimension, then not only will short paths exist between almost anyone almost anywhere, but also individuals with only local information about the network will be able to find them."[67]

So, in order for Friendster and Ryze to provide effective network traversal tools, they need to enable users to measure social dimensions by making real life affiliations explicit within the system. The more affiliations made explicit, the better, as different social dimensions will be meaningful to different individuals executing a directed search.

As we've seen, Friendster user profiles provide more information about affinities than about affiliations. Affinities may be good conversation starters, but they don't reveal whether or not the user in question is someone you would want to start a conversation with in the first place. "My favorite movie is Gremlins and I like peaches" is simply not as useful for assessing social distance as say "I'm from Cleveland and I used to work in the publicity department at MTV." In general, affinities are far less effective when it comes to making a directed search because you don't have to belong to anything to have a proclivity for something. Wildly disparate people from all over the world could be peach loving Gremlins aficionados. Moreover, unless you're the peach farming president of the Gremlins fan club (affiliations), knowing you love Gremlins and peaches is not a social metric that helps me find a path to you. Occupation, location and hometown are the only affiliation-related fields in a Friendster user profile, and of the three only location is searchable. One is better than none, but nowhere near enough to execute directed searches regularly and reliably. Friendster simply does not provide enough social dimensions with which to place someone in social space.

Ryze user profiles, on the other hand, no doubt because they were designed in the spirit of real world business networking, provide an enormous amount of affiliative information. Universities, former companies, current company, positions held, location, hometown -- and those are just the preformatted fields in a user's profile. Ryze users tend to fill their pages with even more affiliative information of their own accord. A user's list of interests serve as the only affinity search pivots, all other Ryze pivots are affiliative in nature. Interestingly, provided you have more than a basic membership, Ryze search functionality enables you to search for users by more than one pivot simultaneously. This would seem to imply that not only is Ryze better equipped to help users find paths to target users, but it also enables users to skip the path entirely and go directly to the user they want. In the real world, geographical and time constraints might force me to be satisfied with finding a target person through a chain of intermediaries, but social software like Ryze lets you go straight to the source.

Ben Hammersley commented upon this phenomenon in his article about "who you know" social software, Click to the Clique:
"Given the right knowledge of the people between us, I could probably plot a chain of people between myself -- here in the depths of the Swedish countryside -- and you, wherever you are. If such information was available, and it turned out it might be advantageous for us to chat, then we could ask each of the middlemen in turn for an introduction and get on with it. Or we could skip the middle guys altogether. That's the idea of the growing number of social network sites on the net today." [68]

Skipping the middlemen altogether is certainly the most efficient means of traversing the network.[69] Since that's the idea of the growing number of social network sites on the net today, whether or not the site supports directed searching is a make or break question. Ryze's functionality, by supporting searches across multiple social dimensions (affiliations), achieves that end. Friendster's, by highlighting affinities, does not.[70]

Ryze supports both traversing and trawling (you can search by interests or random members if you really want to); Friendster only supports trawling. There is one advantage to trawling a network: serendipity. Sometimes people don't actually know what they're looking for or where they're going, but they like what they find when they get there. John Seely Brown describes personalized news feeds in a way that resonates here: "Such a model neglects how difficult it is for people to know and describe what they want...It also neglects the importance of serendipitous news--news that people didn't set out to find -- to the way people understand the world."[71] But trawling the network should be the user's choice, not the only option. Social software that attempts to turn online interaction into offline relationships by building robust online social networks for users to tap into should make user affinities and user affiliations explicit in order to facilitate both undirected and directed searches.

Conclusions >>>

[62] Watts, Six Degrees: The Science of a Connected Age, 136.

[63] Milgram, "The Small World Problem," 65.

[64] Watts, Six Degrees: The Science of a Connected Age, 116.

[65] Ibid., 148.

[66] Milgram, "The Small World Problem," 66-67.

[67] Watts, Six Degrees: The Science of a Connected Age, 153.

[68] Ben Hammersley, "Click to the Clique," The Guardian, 9 January 2003, (,3858,4579200,00.html) (31 March, 2003).

[69] If you're a hub or a high profile personality in the system, users regularly skipping the middleman could be problematic; after all, middlemen act as filters by essentially vouching for the person they're introducing. But the typical network member is not a hub and not inundated with communication from unknown users.

[70] Facilitating directed searches for users is not an issue for Meetup, which has a fundamentally different design principle than does Friendster and Ryze. It is interesting to note, however, that more than half of the most popular meetups are consistently affiliation based and not affinity based.

[71] Brown and Duguid, The Social Life of Information, 218-219.

Introduction  |  More is Different: Why "Social Software," Why Now?  |  Social Network Analysis: The Big Deal About Small Worlds  |  The Big Picture: Friendster, Ryze and Meetup  |  Case Study Analysis: Friendster, Ryze and Meetup  |  Directed Searching: Working the Network  |  Conclusions  |  Sources  |  Printer-friendly Version

Thesis  |  Presentation  |  Message Board  |  Wiki  | Home

Contact: Alicia L. Cervini
Interactive Telecommunications Program, 2003