How will we collect network data in the future? Implications for quantitative and qualitative measurement

As David Lazer argues in a February post for the Harvard-based Complexity and Social Networks blog, the primary way of collecting social network relational data is through self-report. There's a whole can of worms associated with this approach, not least of which is the extensive literature on the social scientific analysis of capturing everyday activity/interaction. Self-report challenges validity because people remember things differently at different times based upon a whole slew of confounding factors. Retrospective accounts have large holes in them. So how to capture interaction data for valid social networks?

Lazer proposed two articles which have used email and other e-data to follow interaction. Yes, digitalia could be the harbinger of something great to come (particularly in reference to research questions which deal specifically with computer-mediated interaction), but a couple of points: e-mail is not an indication of all interaction, and until we have the right to capture data from all meat/digital interaction (don't envision that happening, and don't want to - which is why I refuse to watch Reality TV), we won't be able to "accurately" report true human interaction. There are too many communication strands for us to connect with others to.

However, I hope to ensure the validity of the networks I'm generating by relying upon the explicit all-or-nothing presence of a connection tie in my research, using participants' Second Life Calling Card lists. Sure, some debate whether the presence of a person on a CC list actually reflects a friendship (it could, for example, be a result of a free-for-all popularity contest, it could be the result of a business transaction), but as I maintained on Terra Nova in response to these critiques,

I want to make clear that the survey bases it collection of friend and acquaintance names on calling cards, but the pertinent social network is pulled from this bounty of both strong and (occasionally very) weak/acquaintance ties using questions which delve more deeply into the respondent’s relationship with each person listed. In fact, that some people have so many calling cards is a great permanent record of their interaction patterns in SL.

In the pilot research and pre-survey interviews with SL Residents, the overwhelming finding was that social ties are directly related to the avatars listed on CC lists, and that collecting these would be an indication of some form of association. Placing someone on a CC list represents an explicit formation of a tie.

Using this method also helps in the name generation, without relying upon priming questions. More names are added in a secondary question which asks for avatars the respondent anticipates will be added to his or her calling card list in the near future. This metric is less-"random", and indicates a different strength of tie.

There's an interesting thread happening right now on the socnet listserv, about the possibilities of using qualitative methods to measure networks rather than relying upon a more quantitative approach with traditional sociometrics. Certainly some researchers like Fisher (1982) and Kirke (1996) have argued that at the interpersonal level of definition, qualitative work (like interviews) is essential to properly classify the experiences of the participants. Perhaps a combination of both approaches - an idea championed by McCarty (2003) and others - holds the key to capturing accurate networks, even moreso than following the digital traces we leave behind us.