Made out of people

Abandoning Users

(I wrote this during HCIC, watching a speaker talk about a collaborative system.)

There's a certain ethical constraint on building collaborative and social systems that I've been thinking about for a while. This happens in MMPORGS, YASNS, MUDs and MOOs, email lists, and similar.

Researchers become very interested in building systems, and we study adoption a great deal: how many users can we get, how fast can we get them, and how many hours can we get them to use the system? Those are all good things: the CSCW paper that does wonderfully is the one where the speaker stands up and says, "It was more addictive then crack; we have the entire english-speaking world using it; and they never log off."

Unfortuantely, the next line is--all too often--"then we completed the six week study and turned it off. What a wonderful lesson we learned!"

It happened several times that I've seen recently: CHIPLACE, for example, was an online site for allowing the CHI community to meet up and chat. It was a success--and it was shut down after that year's experiment ended. Fortunately, it was adopted by someone else, who was able to put it up.

I know that danah has talked about this before, but I'm back to thinking about the responsibility of designers to their users. The server resource may be expensive, or require hand-management, or cost per user: issues that are inexpensive during an NSF- or organizationally-funded project, but a big deal after the project ends. (My officemate ran a MUD for part of his dissertation research; he seems to have a critical mass now. He's left it up--but there's a lingering question: he graduates this summer. What happens to his MUD then?)

I guess I feel that there's a new sort of reseach ethics here to consider. Just as a medical experiment can't abandon its users midway ("enjoy the withdrawal, folks!"), neither ought a social experiment. But the reverse is a problem, both in the medical and the social fields: "Yes, we had you for a year, we showed a significant improvement; we now end the experiment and wait a few years for FDA approval."

I sincerely don't know what to advocate here. Am I asking designers not to build social services? Not really. We learn a lot from them, and I build them myself--although I do tend to try to make them decentralized, if possible, so that users can hang onto them at the end if they want to.

And for studies that don't have that nice feature? It would be nice to be thinking from the beginning about a way that a user can adopt the system. Don't build it over your hyper-confidential technology; make it something that the community can--for a few hundred dollars--re-use and repurpose. Allow a volunteer user to come in as an administrator (as Jack does). Put up the server code online, or make it available to your users.

This is a wonderful advantage to being software people, rather than medical people. A doctor can't send people home with a lifetime supply of an experiemntal medicine--but porting a server to a desktop isn't too bad.