Blog Post

Interview with Terrell Russell and his work on Contextual Authority Tagging

One sunny afternoon last week, I had the opportunity to sit with Terrell Russell, a fellow doctoral student in the School of Information & Library Science at UNC Chapel Hill, and discuss his research. Russell is working on Contextual Authority Tagging, a method for locating expertise within a group.  He graduated from NC State with bachelor degrees in Computer Engineering and Information Technology and Service Organizations and a masters in Computer Networking.  He is the co-founder of ClaimID.com, an online identity manager and OpenID provider.  He also runs Cloudalicio.us, a visualization tool useful for looking at tagging data over time.

To learn more about Terrell Russell, visit these websites:

http://weblog.terrellrussell.com/
http://claimid.com/terrell
http://cloudalicio.us/
http://www.terrellrussell.com/projects/contextualauthoritytagging/

In talking about his work, Russell explains that currently we conceive of expertise location and assignment as being driven by what self-reporters say they are good at. So for instance, in a work setting, it is possible that the organization thinks of Joe Smith as an expert on statistics because he has a degree in statistics from a good university and his resume shows that he has worked as a statistician for many years. Joe is able to demonstrate a set of credentials that has meaning for the organization degree, work experience, etc. But what about all the people who actually know a lot about a given topic or have particularly good skills for which they are not credentialed?

Think of the self-taught gourmet cook who only cooks for the people she invites to a dinner party. She doesnt have anything on her resume about cooking in restaurants or having gone to cooking school, yet she is an expert in cooking because it is her hobby and she has spent a lot of time practicing and researching her skills. How do we determine who is an expert in a given topic area if we dont have a set of official, recorded credentials to back up the notion of expertise? Russell says, We know a lot of stuff and it isnt being recorded. He is interested in gossip-level knowledge rather than credential-level knowledge. So hes interested in the fact that all the friends of the gourmet cook actually talk amongst themselves about how wonderful her cooking is. Its not written down anywhere; but folks know this to be true.

Now lets add to this the notion of identity in the virtual world. In the physical world, I show up at school for a meeting and everyone knows who I am. They can see me. They can see how I act and what I say. They have a sense of my identity and it is pretty much consistent when I go the next meeting with different people. All those people get a chance to know me in different situations. I am always who I am in those situations. My name is the same and my face is the same. Resolving my identity in two or more physical contexts is not a problem. But in the virtual world, I can be many different people. In World of Warcraft (WoW), my name is not Laura. People in the game know me as something different than what I am in the physical world. If I were to go into an AOL (American Online) chat room, my name would be different from both my real name and my name in WoW. Would the people in AOL, if they were also in WoW, be able to resolve both these identities and know that I am the same person? No, probably not. So if I were good at playing warlocks in WoW, would anybody in AOL know that and seek out my expertise while in AOL chat? Probably not. Additionally, no one in WoW or AOL would know that I have some expertise in systems analysis in the physical world. I cant prove who I am and what I know to them. This complicates the notion of expertise.

In the virtual world, reputation is unclear because who you really are may be obscured by the many faces you don in the virtual world. In the physical world, people determine reputation by hanging opinions and knowledge scores off  your credentialed identity. Russell uses the example of credit cards. You can walk into a store, flash a piece of plastic, and they let you walk out with loads of stuff without laying one thin dime on the counter. The way this works is that you have built up some sort of credit reputation (a credit score); and your identity, as proven by the credit card, is associated with that score and thus a credit line. You are rewarded for your good reputation and history of paying back your debts.

So how do we reconcile all these identities, ascertain reputation, and determine what someone knows in the virtual world? Russell asks, Wouldnt it be cool if I could right-click on you and find out what you know?

Since self-reporting is problematic due to ambition or bias and most especially in the virtual world where you cant truly prove self anyway, Russell thinks we should consider collective opinion as well. Russells research is predicated on the idea that a groups opinion about what somebody knows has some bearing on what the person actually knows. This is an assumption obviously; and his dissertation is meant to justify this assumption. Russell is interested in how the individuals opinion of himself and the groups opinion of the individual's knowledge converge to give us some idea of expertise. Where do they agree?

Russell plans to co-opt social tagging and repurpose it to help answer this research question. Instead of tagging webpages, how about tagging people? This is different from the recommendation system you might find in LinkedIn. That type of recommendation system doesnt ask Jill, the recommender, to talk about what she thinks Mark, the recommendee, actually knows. It also doesnt aggregate recommendations in a way that is searchable and meaningful. In other words, there is no way to search for expert in information retrieval and get results based on recommendations.

Russell will run a series of modified DELPHI experiments where he will ask individuals to rate themselves on what they know and then rate others on what they think the others know. After many repetitions, the hope is that some sort of convergence between what the individual says about himself and what the group says about the individual will shake out. In other words, what they all agree on about the individual will rise to the surface; while what they disagree on can be left to sink to the bottom.

So, perhaps I am a subject in Russells study and I say that Im good at systems analysis, macram, and playing a warlock in WoW. Then the rest of the subjects say that Im good at systems analysis and some say that I probably am good at warlocks in WoW; but no one mentions that Im a kick-butt knotter. So we could assume then, that I actually am good at systems analysis and maybe also at playing warlocks, but that it is unlikely Im good at macram (or at least that my expertise in this area has not yet  been demonstrated to my fellow research subjects).

Then the hope is that this type of social-people-tagging can be scaled up to the Web so that we can actually click on Terrell and find out what hes good at. But first things first Russell intends to get the conceptual, logical framework tested and built first. Then, a system of some kind, available on the Web, can be built that allows us to tell who is good at what.

82

No comments