Edgesense testing - a working wiki

  • This wiki is a"work on progress" daily ( bi-daily) weekly documentation about testing edgesense.
  • The 1st set of questions are here.
  • as it is a work on progress, it wouldn't be well structured, include some thoughts, and will try not to add too much screenshots from edgesense to keep it easy to follow.
  • any remarks, questions, comments,.........etc are welcomed of course :)

Findings

  • at a first look on Edgeryders through Edgesense NOW , one can find 7 sub-communities, 3 relatively big and 4 relatively small ( according to the no of nodes).
  1. Green-light    ( 1st community )
  2. orange-light  ( UnMon and LOTE4  )
  3. Blue-light       (STF ? )
  4. Green-dark    ()
  5. Blue-dark       ()
  6. Red                (Ci2i )
  7. orange-light   ()

when removing the moderators the most affected community (from the big 3 )is the blue-light(STF ).

the small red (Ci2i) community isn’t affected by the moderators.

the 3 other small sub-communities are still undefined,

  • on nodes/ users - comments / edges statistics

edgeryders reached 2500 registered users with 634 nodes in March 10 , 2015 with a ratio of 25.4 % . the nodes to users ratio increased from 22.3 % in December 9, 2013.


  • edgeryders network includes 4272 edge from 15322 comment on March 10, 2015 with a ratio of 27.9 % . The edges to comments ratio decreased from 35.3 %  n Dec 9, 2013 (2532 edge from 7168 comment)

this decrease in the percentage could mean

  • that the new nodes aren't integrated nicely as the old ones.
  • old nodes forming the sub-communities are talking more to each other ( so no of comments increases but not new edges


-  looking through the question of :" who are the central non-moderator members? (potential moderator role in the community)"

looking at the network without the moderators one can spot some nodes taking some centrality positions , some of them are still active while others left the conversations long ago.

with the new sub-communities (after the merge) it is easier to spot who left the conversations and who is still engaged in new conversations.

basically the green-light sub-community is the old content which will show major changes unless you go with the time slider back to 2012. so any node from other sub-communities who is connected to the green-light sub-community is an old member who is still engaged.

that’s why there is no interaction between the STF blue-light community members with it.

moving with the slider from the start till now will give a very nice idea on how the community developed and how the old community “gave birth” to the smaller community active now in LOTE and UnMon. It shows also how the STF community wasn’t completly integrated with the LOTE4 for example ( again it is easy to spot the members who joined the discussions )

it would take time to spot all the nodes, so edgesense in a big network like edgeryders would be easier  for community managers more that normal users ( as community managers using edgesense will get used the network after a while of using it )

  • old members who are not active anymore
    • Alessia  ( in: 30, out: 25, bet: 0.0122 w/out mod - in : 34 out: 32 bet 0.0046 w/ mod ) ??
    • Chara.oikonomido ( last activity 3 years ago - last access 2 years 2 month ago )
    •  Lyne Robichaud  ( last activity 3 years ago - last access 2 years 7 month ago )
  •  active ones
account details without moderators details with moderators
Bembo Davies in 21 out 28 bet 0.0159 in 27 out 36 bet 0.0024
Ben in 59 out 28 bet 0.0409 in 63 out 38 bet 0.0083
k in 27 out 57 bet 0.0375 in 35 out 65 bet 0.0066
kei in 15 out 12 bet 0.0086 in 20 out 19 bet 0.0007
elf Pavlik in 42 out 37 bet 0.0459 in 53 out 47 bet 0.0103
Dorotea in 26 out 26 bet 0.0223 in 34 out 35 bet 0.004
hexayurt in 39 out 36 bet 0.0318 in 46 out 41 bet 0.0118
SamMuirhead in 25 out 18 bet 0.0234 in 32 out 27 bet 0.0073

the above members are main central nodes and are connected with the STF sub community

Zooming on LOTE4

more precisely zooming on what’s happening on the platform from May 18 to November 18.

  • No of edges increased from 3278 to 4003 creating 725 new edge (22 % increase).
  • No of comments increased from 10498 to 13921 ( 3423 new comment = 32 % increase ).
  • Modularity changed from 0.313 to 0.346


Role of moderators

As indicated above the network expanded during this period of time. This included different interactions between moderators and members. Below is a look into the activity of moderators during this period of time.

Since most community moderators are involved in different projects plus some community management tasks. Not all interactions are LOTE4 related, except for Natalia, as the node was inactive until she took the responsibility to coordinate LOTE4 which included a lot of different tasks as explained here.

account May 18 November 18
Natalia ---- in 35 out 17 bet 0.0048
Nadia in 138 out 160 bet 0.1063 in 193 out 194 bet 0.1337
Noemi in 144 out 230 bet 0.1545 in 191 out 292 bet 0.1903
Alberto in 204 out 310 bet 0.2822 in 233 out 368 bet 0.2834
Matthias in 92 out 101 bet 0.0416 in 114 out 130 bet 00.0491

general comments

// k  and Kei . 2 accounts for the same person ?

// the moderators = the accounts that has the content manager and/or site admin, not all are taking the role of community management. this should be taken in consideration.

Questions

1- about Noemi D6 node, it’s confusing since both ( Noemi and Noemi D6 ) link to the same account yet having 2 different nodes.

looking at both accounts, both are old accounts but one is not used ( the content manager one not the site admin) since 2 years.

so edgesense sees them as 2 separate accounts ( but linking them to the same url )

the question : there is no duplication in mapping Noemi ? Noemi is mapped as 2 accounts ?

2- about Alessia mentioned above : why I can’t see any activity in the profile page ? although edgsense mapped a lot of activities

Answered by Alberto in the comments below

Recommendation/ requests

  1. As the network grows and become more complicated with time, it would be cool to filter nodes with time also. to make it visible which accounts didn't contribute in 2-1 year or 6 month. it would make it easier to spot the people who were active but left the community by time ( the user life cycle ).

// the new sub-communities do that with the different colours.

  1. would be helpful also when hiding sub-communities, the calculations of no of nodes and edges could show for the only showen communities and ignores the hidden ones.
  2. showing who are the "recently active nodes" in a separate pad.

1 Like

Excellent work

This is excellent work, Hazem. Keep it up. We should probably attempt to socialize this a bit, so that this wiki becomes the result of the activity of more people. That way, developers can treat it as “the voice of the users”.

A few answers:

  • We merged the "Noemi" and "Noemi D6" users yesterday. This is done on Edgeryders itself, then Edgesense updates automatically.
  • You can't see Alessia on the user activity probably because of what follows. Up to the end of 2012 we were running on a Drupal 6 platform; when we moved out of the Council of Europe cradle, we took advantage to upgrade to Drupal 7. Matthias wrote a script to import users and content from the old database into the new one, serving the new platform. As a result, all content is there, and linked to its author. However, some of the non-core modules in the Drupal Commons distribution are only imperfectly linked to the database (maybe they store creation dates in some array somewhere). Profile management changed a lot from D6 to D7 (profiles used to be nodes!), so this is a particularly messy department. User activity is one of them, so it is not tracked against content created on the old platform and then imported. Edgesense, on the other hand, is more rigorous, because it goes right to the database, and essentially ignores all the intermediate PHP manipulation made by Drupal.

Many changes after merging Noemi with Noemi D6

So now we have a slightly different shape and less definite cutoff points between subcommunities (by the way we have 7 colours now). I agree with @Hazem that less than those computed by the application are identifiable in reality… this makes for a pretty exercise of “knowing thy community” :slight_smile:

No big deal

Subcommunities are computed by an algorithm called the Louvain algorithm, which is fast (-ish) but non-deterministic. Depending on  topology, it is perfectly possible that the same network analyzed by Louvain twice finds (slightly) different numbers of subcommunities, and assign “border” users to different subcommunities. In general, merging Noemi with Noemi D6 should decrease modularity (as a 0 to 1 index), as the network becomes slightly more similar to a star with the merged user at the center. But I don’t see much structural change, honestly. Also the modularity value is kind of the same.

changed for better - the new subcommunities

although there is no difference in the whole modularity, but the new sub-communities identified the old “dead” content to the new ones created by the old members, as mentioned above in the wiki

now it is easy to spot the old members who are still connected.

quick comment

from a quick look over edgesense now after merging, the green cluster now is mainly about the old EU content, which make it a bit easier in identifying  old members who left the community and who is still connected with the lote and unMon subcommunity …

will see if this is true or not today :slight_smile: and update the wiki.