msurveys.com
instant messaging
SMS/mobile
avatars
Second Life
blog
about msurveys.com
subscribe
Monday, January 15, 2007
Turning the Second Life client into an interviewer terminal
With the opensourcing of the Second Life viewer/client last week, I've spent most of the weekend dusting off my C++ knowledge, and managed to compile the viewer in Xcode on my MacBook Pro.
An idea that was previously mentioned - having "real" interviewers walking around Second Life and conducting surveys - now suddenly seemed feasible, at least technically. By the end of the weekend, I had something basic working...
It goes like this: the interviewer logs into Second Life with the modified viewer, and goes around looking for respondents. At this point everything is as usual, they can walk, fly, chat etc. just like anyone else. Once they find someone, they may ask them if they were willing to take a survey. If the respondent agrees - and here's the clever bit, such as it is - the interviewer can switch their avatar to a "data collection" mode, so the avatar automatically starts saying the questions as they have been defined in the survey back-end, and the answers from the respondent are automatically recorded in the survey back-end too. Once the survey is finished, they're back to a "normal" Second Life user.
Here's a quick video I recorded earlier (sorry it runs a little faster than it should, something happened when transferring it between Win and Mac).. "Belle Isbell" is the interviewer coming up to me, and once I say ok to take part in the survey, she becomes a data collection program - until the end of the survey, when she thanks me and walks away. All my answers to her are now stored in the survey back-end.
In reality, especially in a busier setting, we'll probably want to run the survey in the IM channel rather than the public chat channel, not least for privacy reasons, but I left it in the chat channel as it's easier to see.
An idea that was previously mentioned - having "real" interviewers walking around Second Life and conducting surveys - now suddenly seemed feasible, at least technically. By the end of the weekend, I had something basic working...
It goes like this: the interviewer logs into Second Life with the modified viewer, and goes around looking for respondents. At this point everything is as usual, they can walk, fly, chat etc. just like anyone else. Once they find someone, they may ask them if they were willing to take a survey. If the respondent agrees - and here's the clever bit, such as it is - the interviewer can switch their avatar to a "data collection" mode, so the avatar automatically starts saying the questions as they have been defined in the survey back-end, and the answers from the respondent are automatically recorded in the survey back-end too. Once the survey is finished, they're back to a "normal" Second Life user.
Here's a quick video I recorded earlier (sorry it runs a little faster than it should, something happened when transferring it between Win and Mac).. "Belle Isbell" is the interviewer coming up to me, and once I say ok to take part in the survey, she becomes a data collection program - until the end of the survey, when she thanks me and walks away. All my answers to her are now stored in the survey back-end.
In reality, especially in a busier setting, we'll probably want to run the survey in the IM channel rather than the public chat channel, not least for privacy reasons, but I left it in the chat channel as it's easier to see.
Techorati Tags:
Comments:
<< Home
Excellent stuff Mario. It seems to me there are opportunities galore in a research conext. I've just joined 2nd Life to get more familiar with the experience. The environment seems to have a potential to enable a new world of fusion of virtual techniques (such as shopping, ad/product placement; market simulation (faux product launches); perhaps even avatar ethnography...then again, maybe not!
Interesting concept, altough I find typing numbers not so convenient: in a real use, the fact that an avatar starts the interview would led most people to try to articulate their answers in words and arguments.
If that happens, I'd try to use llDialog and see if it works better.
On a more technical note, I think there is some need for a generic, very simple way to tie the viewer to external programs: it can be done with pipes, a text interface that would allow external programs to control an SL avatar and chat (on any channel), listen, move or IM.
That is easy to do and would make the SL client an order of magnitude easier to play with for non-programmers.
If I have time (non-likely), I'll work on such a custom viewer.
If you are interested in marketing and SL, you may be interested on the little project I am working at the moment, a tool that extracts the "buzz" from an SL location (or a network of locations): it listens for chat, mines it for keywords and aggregates it in various ways.
If that happens, I'd try to use llDialog and see if it works better.
On a more technical note, I think there is some need for a generic, very simple way to tie the viewer to external programs: it can be done with pipes, a text interface that would allow external programs to control an SL avatar and chat (on any channel), listen, move or IM.
That is easy to do and would make the SL client an order of magnitude easier to play with for non-programmers.
If I have time (non-likely), I'll work on such a custom viewer.
If you are interested in marketing and SL, you may be interested on the little project I am working at the moment, a tool that extracts the "buzz" from an SL location (or a network of locations): it listens for chat, mines it for keywords and aggregates it in various ways.
Thanks for all the comments.
Nic/Richard, I agree (maybe obviously, given my interest in this) that there is potential and application for this kind of thing. Especially if you view SL as the shape of things to come, it will be useful for marketers and market researchers to start engaging with this medium, and try and find and make use of any new opportunities that come with it.
Kitten Lulu, I agree that adding some sort of natural language recognition facility would be useful. In the simplest sense, and this would be trivial to implement, I could accept "Yes" for "1" and so on. Ideally you'll want to build in more sophistication to be able to deal with mis-spellings and the like.
With regards to llDialog - this is really limiting in a survey sense, as you can only ask single-choice questions that way (and only with very short responses, as otherwise they'd be cut off by the viewer), and no such thing as multi-choice or open text answers would be possible. For that reason, I think the chat and IM channels still offer the most potential.
Thanks for mentioning your "buzz" project - sound interesting, and I'll take a closer look.
Nic/Richard, I agree (maybe obviously, given my interest in this) that there is potential and application for this kind of thing. Especially if you view SL as the shape of things to come, it will be useful for marketers and market researchers to start engaging with this medium, and try and find and make use of any new opportunities that come with it.
Kitten Lulu, I agree that adding some sort of natural language recognition facility would be useful. In the simplest sense, and this would be trivial to implement, I could accept "Yes" for "1" and so on. Ideally you'll want to build in more sophistication to be able to deal with mis-spellings and the like.
With regards to llDialog - this is really limiting in a survey sense, as you can only ask single-choice questions that way (and only with very short responses, as otherwise they'd be cut off by the viewer), and no such thing as multi-choice or open text answers would be possible. For that reason, I think the chat and IM channels still offer the most potential.
Thanks for mentioning your "buzz" project - sound interesting, and I'll take a closer look.
This is really cool. Perhaps she should carry a clipboard with her so that she can write down the answers.
I second the clipboard idea!
As for the llDialog thing - if the researcher were to "chat" the question, then give multiple choices, the llDialog would only need to be a simple pick A, B, C, D choice.
It might make the whole process easier for the respondent, and make them less likely to walk away. Remember, to most people it's just a game - I spend half my 1st life being hassled by researchers on the street (I have that kind of face) so to do it in SL it needs to be quick, fun and rewarding.
Just a sidenote - I came on this project by accident - but this sort of mod is why I love open source! Viva da Lindens!
As for the llDialog thing - if the researcher were to "chat" the question, then give multiple choices, the llDialog would only need to be a simple pick A, B, C, D choice.
It might make the whole process easier for the respondent, and make them less likely to walk away. Remember, to most people it's just a game - I spend half my 1st life being hassled by researchers on the street (I have that kind of face) so to do it in SL it needs to be quick, fun and rewarding.
Just a sidenote - I came on this project by accident - but this sort of mod is why I love open source! Viva da Lindens!
Very cool indeed! Are you thinking also about some follow-on app that the survey participant can activate and which would record subsequent behavior? It would be really amazing to overcome the "stated vs. derived" intent hump that lots of RL surveys can't always address: do people/avatars ultimate 'do' what they 'say?' I know the idea has scary Big Brother overtones, but it's a valid problem to try and solve, isn't it?
Post a Comment
<< Home