Where are you in this picture? Please join fellow Edgeryders in helping me identify what brings the subcommunities (identified by color) together. Take the following questionnaire, it takes only a few minutes!
Great playground
Hi Alberto … at some point of the conference it came to my mind that the Edgeryders project will probably be integrated in some way in your PhD research, and it seems I was right smilez And indeed, it’s a great playground for yoour research. Just watched the video in the post you linked … well done!
Here’s some more analytical content in addition to my questionairre response.
I gave an example of how it seems to be “by chance” that I ran into some people on the platform and interacted a lot with them, and did not meet others there even though I also could have interacted a lot (and did so, at the LOTE conference and unconference). And, the example is about one of these “induced” conversations. Not guided by staff.
Even now I still keep finding old great posts and great folks on this platform whom I haven’t interacted on the platform with so far, but would clearly like to talk to … . And to me it seems that the somewhat limited navigation capabilities of the platform promote this “element of chance” in who interacts with whom. Additionally, in my experience there’s simply a limit of personal resources (time) that makes it impossible to find all the communication partners one would want to interact with (seems a bit related to the “communities don’t scale” argument, just that they do scale, but influenced by resource limits).
Chance is of course not the best facilitator to partition online communities into sub-communities and to make them scale that way, as it does not necessarily bring the right people together for best synergy. So it might be an idea to try measuring the influence of chance, and find ways to minimize it in massively parallel online discussions.
Chance?
Matthias, first of all many many thanks for this precious help (again!).
Well, if I am right this sense of serendipity you are experience is not chance at all. The way I see it, it works with a three-layers mechanism.
The first layer is human and project-side. It is made of animation and facilitation: the Edgeryders team sets a goal, designs campaigns, promotes them online and off-. Also, once the conversation got going, engagement managers were encouraged to introduce Edgeryders to each other on the basis of common interests: “Welcome to Edgeryders, Alice! You should definitely meet Bob [link], as you two share an interest in techniques for cooking turnips while on Mars.” And finally, the think-tank ethos gives everybody explicit permission to join in each and every conversation.
The second layer is technological and project-side. It is made of design choices that (1) disables most channels made available by Drupal Commons (fora, blogs, wikis, events) “forcing” everything into the mission brief-mission report-comments cycle; (2) streams all activities onto a feed which is then very prominently displayed at the top of the home page; (3) sends people email notifications whenever a new mission report is posted. This layer serves to increase the probability that Edgeryders come into contact with other Edgeryders and information they are not looking for.
The third layer is human and user-side. It is made of your own interests, idiosyncracies and obsessions that filter out stuff and people you are not interested in and retains the ones you are.
The combined results of these three factors facilitates the “grouping” of Edgeryders by interest - which I guess is what happened in most of the cases (but not all: a couple of subcommunities seem to be determined by language choices).
By the way, congratulations on achieving the highest betweenneess centrality value not only in Subcommunity 1, but in all Edgeryders! 
Not only chance …
Ok, I should have added the obvious: the three mechanisms you mentioned do work for enabling desirable contacts on the platform. They even worked quite well for me in the sense of resulting in a bunch of connections that proved to be interesting at the conference …
But still, there’s a gap when comparing this result to what connections could be made. Because, as per my anecdotal reference, the (un)conference also resulted in interesting connections that were not at all prepared by contact on the platform. So I wondered, why is this? What would be the “perfect” set of connections that could be made, and what were the influences that this did not happen? And, to map my initial idea to your terms: it seems that the third mechanism “own interests, idiosyncracies and obsessions that filter out stuff and people” is limited by personally available time and by the platform’s navigation and filtering options so that a good deal of “chance” influences the results of this incomplete filtering process.
The question is of course, if desigining for emergence should focus on this (assumed) “perfect set of connections anyway”. It would need an awareness and filtering process that covers the platform as a whole, and might in this way contradict the concept of emergence coming from local interactions. And given enough time with the platform, the perfect (most synergistic) set of connections would probably emerge anyway. But I dunno about these; you’re the emergence designer, sir … 
Perfect?
I have no idea what the perfect set of connections would look like, so I really don’t know where to start in measuring its distance from the set of connections that we actually have. You are of course right in that, as you spend more time on the platform, your idiosyncratic filters let more stuff in, as this results into more, interesting connections. I guess this is appropriate: there is a tradeoff between time invested surfing Edgeryders and the benefits we derive from it. We adjust our time investment based on competing possible uses for that time, and the benefits they give us.
I am not sure what you mean by a filter that covers the platform as a whole. What do you have in mind? Maybe it is my emergence bias, but i have trouble seeing Edgeryders as anything other than a structure emerging from a multitude of local interactions.
linked data / graphs , and narratives
Hi Alberto.
I am interested in such approach too.
Programmer friends of mine are developing “Netention”,
which enables a graphed input of narratives, enabling users to directly benefit from various applications of the graph they contribute to.
You can find an introduction ( pdf ) on :
http://www.automenta.com/netention
It is not yet another platform - it is rather an interface ( open source ) which makes it easier to input and use relational data, using linked data / semantic web protocols.
I can also confirm that a number of conversations continued outside of the edgeryders platform,
or blended into other communities of practice.
Networks, but not social
Hey Dante, I have looked up Netention. Some stuff I understand, some I don’t. Making maps with linked data which are based on networks or influence (say, an earthquake in place A could, given a broad enough radius, affect the nuclear power station in place B clearly makes sense. The part about communities, I don’t get. Linked data use graph theory just like social network analysis: the math is the same, but the underlying dynamics are different, so that you end up applying different models.
Have you actually taken the survey?
Communities using a Graphed Approach
Hi Alberto,
glad to share graphed applications.
A way I see certain applications for communities , as detailed in some respect on http://automenta.com/netention
is by use and combination of certain vocabularies such as FOAF
http://en.wikipedia.org/wiki/FOAF_(software)
“ is a machine-readable ontology describing persons, their activities and their relations to other people and objects.”
and RDF Schema
Questionnaire: check!
I think a barrier to even try to define a perfect set of connections is the scope of the community, by design: the thematic conversations that were created shaped /narrowed down the things that participants disclosed about themselves. If you’re into fishing and you’ve met someone at Edgecamp also passionate about that, chances are you wouldn’t have had the context on Edgeryders to actually learn about your common interests… unless you’re making a living out of that or something that’s connected to transition…
Matching people’s needs and offers, and one-on-one discovery hasn’t really been a priority, rather it’s been assumed to happen to a certain extent from converging conversations. In the end it happens just like Alberto says, that the more time you spend the more you reduce serendipity.
@Alberto: Is there an empirical relevance of the really small subcommunities that were highly circumstantial and based on one time conversations, also entirely isolated from the rest? or are they a mere result of the algorhytm, that takes into account all people who have commented at least once on a report?
Structure, not idiosyncracy
Noemi and all: the results I am presenting are purely algorithimic, unfiltered by whatever I might think to be relevant. So, for example, there is a subcommunity around Tudor. It formed like this: first he commented Rossella Bargiacchi’s great Sharpest Edge mission report - still during the testing phase, and she replied. Then he left a comment to Alberto MZ’s Pstcard from the future report; JulienC commented him back, and he replied. These are just two of many interactions that happened in Edgeryders, resulting into a subcommunity - because these three users did not comment or were commented by any users other than members of the team.
Now, the question I am asking is this: is there a reason for the “grouping” of users into subcommunities? Or is it randomness? Randomness is unlikely, because the modularity value tells us the structure we see is unlikely to emerge from random interaction, but still possible - I don’t want to overstate this. However, if several people look at this and say “hey, but those communities are clearly identifiable as people discussing sets of topics!” then you are starting to have an improbable structure that fits well into a certain explanation: still not a “fact”, but worth looking into.
Thanks to you all - I now have four completed questionnaires. It would be nice to get to ten…
2 issues I see relevant
Ok, I understood better your questions now.
-
Members in some of the subcommunities indeed seem to be having shared conversations and interests, mainly those that have larger members and include more conversations that 1 in a mission report
-
Problem is that subcommunities may not be clearly cut. In my reading of Edgeryders content, in practice there is some overlapping, in that some participants should appear as members in more than 1 community if they engaged in diverse conversations… Not sure if it makes sense in the network theory, but anyway…
Data are messy
Well, of course: these are real-world data, and the real world is messy, not clear-cut. The algorithm probably has a hard time assigning some of the more curious Edgeryders, who participate in several discussion, to any one subcommunity (and the mathematical constraint is exactly that each person can only be assigned to one of them). But here’s the thing: if there are too many edges connecting two subcommunities, the algorithm merges them, because having just one subcommunity instead of two makes the partitioning better.
Also, there is nothing in the data that tells us different subcommunities are talking about different things. This is just a hypothesis. There could be other explanations: for example, one subcommunity is writing mostly in French, and non-French speakers are going to gravitate out of it. Or another one might consist of longime friends. Or it could be chance - though, as I said, it is not likely that randomness would generate such a fairly clear-cut structure.
Yep, true points
Thanks for pointing it out how the current Edgeryders portal structure and priorities influences the connections that can be expected.
Which reminded me of another point that came to my mind when I watched the first presentation about Alberto’s PhD project back in April (or so): Will people be happy with their interactions being “designed” and subtly influenced for emergence design when it’s not clearly stated how this happens and to what purpose?
It’s not an issue on Edgeryders so far as its purposes are clearly stated (still I keep an eye on what mechanisms the platform deploys and how staff behaves to guide my behavior here … I’d be uncomfortable with being unconsciously influenced for some “higher purpose” of emergence to happen.) It’s also not an issue with all the special purpose communities out there from forums to crowdmapping to online dating, which all have very obvious means to guide behavior. So we know it influences our interactions, and we agree because we share the site’s goal …
However if Alberto’s idea of emergence design works, resulting in subtle influences that go unnoticed: Does it enable to design results-of-emergence that no single participant wanted to happen? Or more directly: Could emergence design be abused by those in power to do and deploy the design?
Just food for thought though
Interaction design
Hmm, here’s again something about “perfect” connections. (Actually these are just random ideas that come to my mind without knowing all the technical backgrounds, but maybe there’s an inspiration for somebody else hidden in them.)
First, I’d like to know what you use to guide the emergence design, if you think it’s rather unfeasible to let it be informed by an “optimum” target condition? Just curious!
Then, something on how I had meant the notion of “perfect”. Of course there’s no perfect state that could be calculated exactly beforehand, so I just referred to a favourable target state that could inform the emergence / interaction design. For that purpose, could we assume that the “perfect” set of connections would be the one that serves the platform’s purposes most well, so, the one with the best synergy for figuring out transition in our case. Then further we could assume that these or similar enough connections would arise anyway, using the three existing platform mechanisms and given enough time and interaction options (including real life events etc.)
(I agree to be disproven about assuming a “target state” if, in an online community like ours, there’s no final state of connections which is quite constant over time. Yet so far, I assume from real life analogy that there are these (quite constant) better and less great positions in the social fabric for everybody, that is, where you’re more or less productive / synergistic.)
Now with this notion of a “perfect” state, one could measure what valuable connections arose late on the platform, and try to determine why this was so and how the platform could have enabled them earlier (all in a probability sense, like, what tools the platform could have offered that might have resulted in this connection happening earlier). This is the way how I’d judge “quality of emergence” on a platform like Edgeryders: emergence happens more efficiently if it happens faster, given the same amount of time investments.
Quoting Alberto:
“there is a tradeoff between time invested surfing Edgeryders and the benefits we derive from it. We adjust our time investment based on competing possible uses for that time, and the benefits they give us.”
When Edgeryders offers mechanisms for more efficient emergence of connections, it has a “competitive advantage” compared to other uses of people’s time. So people might surf it even more, positively reinforcing how well the platform and emergence on it works … . Isn’t it about such results why emergence design matters actually?
“I am not sure what you mean by a filter that covers the platform as a whole. What do you have in mind? Maybe it is my emergence bias, but i have trouble seeing Edgeryders as anything other than a structure emerging from a multitude of local interactions.”
Yes, it’s a structure that happens from local interactions. But how? As I understand it, emergence is designed by deciding what local interactions are possible and how easy they can be made. So interactions don’t just happen - the platform capabilities and staff behavior influence what interactions will happen and when. So I had the idea of this “filter” as another major platform capability.
For that, I had in mind a sophisticated algorithm to let people discover interesting content and interesting people. Like an advanced version of the already quite nice but hidden advanced search (took me some time to discover it!). Having such a tool positioned as the most prominent way to navigate would (IMHO) hopefully lead to more “efficient” emergence and thus to different sets of connections.
On a note, the filter might even use an algorithm similar to the sophisticated OkCupid matching algorithm - for non-romantic purposes here of course. ((And on another note just for the fun of it, OkCupid also has hilarious nerdy data analysis of their network, showing what can be answered with data crunching …))
Further analysis
Greetings Alberto…
This is potentially a helpful analysis but might I suggest that perhaps a ‘wordcloud’ scan of the ‘links’ between these ‘subgroups’ might give some extra focus to the nature of the subgroups.
Also, I note that your name, Nadia’s and Vinay’s do not appear in any group…??
Regards
Arthur …
PS apologies for putting more work on your ‘plate’
Great idea!
Great idea, Arthur! Actually, it it not so simple to execute. The way it works is this:
- we do a dataset dump via the internal pluming of the platform (Drupal). It results in three JSON files,
- then, a Ruby script (http://github.com/dragontrainer/edgeryders-mapper) parses the files, pulls the relevant info and reassembles them into a .net file, correctly formatted for network analysis software
- at this point, I load into my NetAn software of choice (Gephi in this case), apply the Louvain algorithm and obtain the partition into subcommunities. Notice that, though the same person is assigned to only one subcommunity, she generally interacts also with people from other subcommunitities. For example, look at the picture below: if you look to the top right, you will see that Beckery is connected to edges of different colors. So, it is not enough to take the mission reports and comments of all people in a subcommunity and do a word cloud of that: you have to assign each individual comment to the right subcommunities. And even that is difficult, because some comments (represented in the graph by edges) connect people in different subcommunities, so you have to figure out a way to parse those words.
Edgeryders team members have been dropped out of the dataset to draw this graph. The reason is this: it is the team’s job to talk to everyone, so we - by design - pull the network away from the segmentation into subcommunities (in the full graph there are many more edges, about 1500). What you are looking at is what I call the induced conversation, i.e. the pattern of interaction among community members that no one from the team is involved in.
Clouds
Greetings, Alberto…my pleasure …I did not think it would be ‘easy’ (hence the work on plate comment) but given the variety and depth of the site interactions I thought there might be enough data to get some interesting results…
Good luck…
AD
Seven in
C’mon folks, make an effort. Seven questionnaires in, some more to go…
Where is the link to Q’nnaire?
Hi Alberto,
The page appears balnk on my screen for some reason and cant seem to find the q’nnaire.