Prior work on modeling of community activity #CoP


Lee Romero
 

Hi all - I have recently been considering some ideas about trying to model (via computer simulation) activity within a community of practice.  

There are several potential questions I would like to address in doing that.

But before I do too much, I wanted to see if there might be prior work in this.

Based on several searches in Google Scholar (and in the general Google search results) I don't find anything but it seems like something that must have been done before.

Does anyone in this community know of any prior work?  If so, do you have links to it?

Thanks for any info!

Regards
Lee Romero


Matt Moore <innotecture@...>
 

Lee,

"I have recently been considering some ideas about trying to model (via computer simulation) activity within a community of practice."

Can you say a little bit more about what you are trying to do? What sort of activities will you be modelling? Why are you doing this?

The immediate thought that comes to mind is the work of Duncan Watts.

Regards,

Matt Moore
+61 423 784 504
On Jul 31, 2017, at 4:45 AM, Lee Romero pekadad@... [sikmleaders] <sikmleaders@...> wrote:

 

Hi all - I have recently been considering some ideas about trying to model (via computer simulation) activity within a community of practice.  

There are several potential questions I would like to address in doing that.

But before I do too much, I wanted to see if there might be prior work in this.

Based on several searches in Google Scholar (and in the general Google search results) I don't find anything but it seems like something that must have been done before.

Does anyone in this community know of any prior work?  If so, do you have links to it?

Thanks for any info!

Regards
Lee Romero


Lee Romero
 

Hi Matt - For the initial work I'm keeping it simple and the "activities" I'm modeling are basically the equivalent of posting into an online community's discussion board tool (equivalent, roughly, to using a list server like SIKMLeaders uses or a Yammer Group or a FaceBook Group, etc).

Why am I doing this?  I've had a number of discussions over the years with colleagues about strategies for managing communities of practice within an organization.  The nature of the conversations is generally about the value of trying to have some control around creation of communities and also the value of trying to grow a community once it is established.  

These overlap quite a bit - I find that if you don't try to control (even lightly) what communities exist, that leads to duplication, fragmentation and ultimately small and likely not-long-lived communities.   On the other hand, if you have even some light control and try to actively encourage communities to merge when they are basically about the same topic, the result can be (I think) larger, more robust and therefore more successful communities.

While I was attending the Midwest KM Symposium in May, during one of the sessions (I forget which one - it might have been Stan Garfield's) the thought occurred to me that one assumption I make in my thinking on this is that to some extent, it reduces to math - given X people in a community if the likelihood of any person doing something at any time (whatever that "something" is - asking a question, sharing some insight, replying to someone else's question, etc.) is a relatively constant (small) percent, then as X goes up, activity goes up.

Given that kernel of an idea, I've been noodling over it for a while and thinking it seems feasible to approximately simulate this in a computer simulation.  I've been thinking about what factors are at play here (it's not as simple as thinking "any given member has P probability to start a post or to reply to another post at any time") - need to consider factors like individual member's level of interest in the topic of the community, variations in the interestingess of individual posts, how "popular" a member might be [many communities have some members whose posts will garner more replies than most members], the fact that people have a constrained amount of time, etc.

I've already started working on how to simulate this mostly because it's an interesting experiment.  As I've gotten closer to being able to (I think) somewhat realistically model communities, I wanted to see if anyone else has done anything like this (yeah, probably backwards I know but regardless, the work so far has been some interesting programming at least :) ).

Some of the specific questions I've captured that I think this could address (some are likely a stretch to say the least):

·         Can you even simulate activity in a community in a realistic manner?

·         Does the size of a community relate to the level of activity?

·         What attributes are needed to properly model a community?

·         For those attributes, which values or probabilistic models properly describe a community?

·         How do interests of community member affect activity?

·         What kinds of activities should be considered?

·         What is the effect of having multiple similar (overlapping) communities on community viability?

·         Is there a difference in behavior between a community within a constrained organization (a company) and an effectively open community (e.g., internet-based)?

·         Does the 90-9-1 rule naturally flow out of any of these models (or anything like it)?

·         Is there a size of community that is too large for the community to function properly?

·         For a community manager – is there a size of community that becomes unmanageable (or to ask a different way, can you make a decision on how many community managers might be needed to adequately manage a community based on its size)?




On Sun, Jul 30, 2017 at 3:21 PM, Matt Moore innotecture@... [sikmleaders] <sikmleaders@...> wrote:


Lee,

"I have recently been considering some ideas about trying to model (via computer simulation) activity within a community of practice."

Can you say a little bit more about what you are trying to do? What sort of activities will you be modelling? Why are you doing this?

The immediate thought that comes to mind is the work of Duncan Watts.

Regards,

Matt Moore
On Jul 31, 2017, at 4:45 AM, Lee Romero pekadad@... [sikmleaders] <sikmleaders@...> wrote:

 

Hi all - I have recently been considering some ideas about trying to model (via computer simulation) activity within a community of practice.  

There are several potential questions I would like to address in doing that.

But before I do too much, I wanted to see if there might be prior work in this.

Based on several searches in Google Scholar (and in the general Google search results) I don't find anything but it seems like something that must have been done before.

Does anyone in this community know of any prior work?  If so, do you have links to it?

Thanks for any info!

Regards
Lee Romero





 

Lee,

It is my experience that developing, nurturing and stewarding communities is both an art and a science - so that your quest for simulation is perplexing for me.   Thought I'm sure Facebook, for instance, has all kinds of modeling around communities of interest on their platform.

Productive communities of practice need an active community stewardship structure in terms of a committed community manager/facilitator, a contextual sense of mission (meaning that the community mission is relevant to some sense of connection with an organizational or business imperative or a common purpose of its members.)  There are lots of strategies and tactics to support engagement, sharing, eliciting knowledge, broadening conversations, shaping agendas and curating content.  

A few observations and references for you.
1.   Jettison the old 1-9-90 rule - it's antiquated.   The Community Roundtable, over the past few years in their "State of Community Management" report has shown a participation ladder with more meaningful distributions.  


2.  Communities thrive for different reasons, and it's important to think through the dynamics of supporting meaningful engagement and social learning - variation, consistency, content relevancy, humaneness .
https://www.slideshare.net/CatherinePaloAlto/building-communitya-conversation

Each community is going to provide different kinds of value, at different moments in time, to different stakeholders.
I commend to you this seminal piece by Etienne Wenger, Beverly Trayner and Martin de Laat (2011).   I read this once a year.

Speaking of value - there's a chronic issue with understanding communities - activity metrics are typically indicators, but are often confused with results or impact.

3.  Communities of practice are distinct from a impactful but associative network.  The collaboration framework matrix in this blog post might be of some assistance in teasing out what you might be thinking about simulating.
http://mercedgroup.com/somethings-on-overload-but-its-not-collaboration/

I looked at this article when trying to understand the cycles a community might go through
It helped me conceptualize what I was working through with a broad, global community at the time.
https://www.slideshare.net/CatherinePaloAlto/building-online-communities-strategy-launch-engagement-growth

4.   Relative to community size - sometimes the challenge is not the community size but the mis-apprehension of the community manager and a dedicated resource.  

5.   I would also recommend a work by Bradley and McDonald - we used the book in a course co-taught at Columbia's Information and KNowldge Strategy program in 2014 and 2015.. 
It's a bit dated but it situates important considerations and benefits of communities within organizations.

https://www.amazon.com/Social-Organization-Collective-Customers-Employees/dp/1422172368


Hope this is helpful.
Catherine Shinners
650-704-3889

11







Social Collaboration and Digital Transformation
Silicon Valley, USA
+1-650-704-3889

Contributor to Smarter Innovation (chapter abstracts) (Ark Group, 2014)
Blog: www.collaboration-incontext.com


On Sun, Jul 30, 2017 at 1:10 PM, Lee Romero pekadad@... [sikmleaders] <sikmleaders@...> wrote:
 

Hi Matt - For the initial work I'm keeping it simple and the "activities" I'm modeling are basically the equivalent of posting into an online community's discussion board tool (equivalent, roughly, to using a list server like SIKMLeaders uses or a Yammer Group or a FaceBook Group, etc).

Why am I doing this?  I've had a number of discussions over the years with colleagues about strategies for managing communities of practice within an organization.  The nature of the conversations is generally about the value of trying to have some control around creation of communities and also the value of trying to grow a community once it is established.  

These overlap quite a bit - I find that if you don't try to control (even lightly) what communities exist, that leads to duplication, fragmentation and ultimately small and likely not-long-lived communities.   On the other hand, if you have even some light control and try to actively encourage communities to merge when they are basically about the same topic, the result can be (I think) larger, more robust and therefore more successful communities.

While I was attending the Midwest KM Symposium in May, during one of the sessions (I forget which one - it might have been Stan Garfield's) the thought occurred to me that one assumption I make in my thinking on this is that to some extent, it reduces to math - given X people in a community if the likelihood of any person doing something at any time (whatever that "something" is - asking a question, sharing some insight, replying to someone else's question, etc.) is a relatively constant (small) percent, then as X goes up, activity goes up.

Given that kernel of an idea, I've been noodling over it for a while and thinking it seems feasible to approximately simulate this in a computer simulation.  I've been thinking about what factors are at play here (it's not as simple as thinking "any given member has P probability to start a post or to reply to another post at any time") - need to consider factors like individual member's level of interest in the topic of the community, variations in the interestingess of individual posts, how "popular" a member might be [many communities have some members whose posts will garner more replies than most members], the fact that people have a constrained amount of time, etc.

I've already started working on how to simulate this mostly because it's an interesting experiment.  As I've gotten closer to being able to (I think) somewhat realistically model communities, I wanted to see if anyone else has done anything like this (yeah, probably backwards I know but regardless, the work so far has been some interesting programming at least :) ).

Some of the specific questions I've captured that I think this could address (some are likely a stretch to say the least):

·         Can you even simulate activity in a community in a realistic manner?

·         Does the size of a community relate to the level of activity?

·         What attributes are needed to properly model a community?

·         For those attributes, which values or probabilistic models properly describe a community?

·         How do interests of community member affect activity?

·         What kinds of activities should be considered?

·         What is the effect of having multiple similar (overlapping) communities on community viability?

·         Is there a difference in behavior between a community within a constrained organization (a company) and an effectively open community (e.g., internet-based)?

·         Does the 90-9-1 rule naturally flow out of any of these models (or anything like it)?

·         Is there a size of community that is too large for the community to function properly?

·         For a community manager – is there a size of community that becomes unmanageable (or to ask a different way, can you make a decision on how many community managers might be needed to adequately manage a community based on its size)?




On Sun, Jul 30, 2017 at 3:21 PM, Matt Moore innotecture@... [sikmleaders] <sikmleaders@...> wrote:


Lee,

"I have recently been considering some ideas about trying to model (via computer simulation) activity within a community of practice."

Can you say a little bit more about what you are trying to do? What sort of activities will you be modelling? Why are you doing this?

The immediate thought that comes to mind is the work of Duncan Watts.

Regards,

Matt Moore
On Jul 31, 2017, at 4:45 AM, Lee Romero pekadad@... [sikmleaders] <sikmleaders@...> wrote:

 

Hi all - I have recently been considering some ideas about trying to model (via computer simulation) activity within a community of practice.  

There are several potential questions I would like to address in doing that.

But before I do too much, I wanted to see if there might be prior work in this.

Based on several searches in Google Scholar (and in the general Google search results) I don't find anything but it seems like something that must have been done before.

Does anyone in this community know of any prior work?  If so, do you have links to it?

Thanks for any info!

Regards
Lee Romero






Matt Moore <innotecture@...>
 

Lee,

There are thoughts that come to mind:

1. There's been a fair bit of analysis around CoPs - examples include SWOOP (Cai & Laurie) looking at network metrics, Kai Riemer doing Yammer post content analysis, etc. I would definitely look at Kai's work - one of his case studies was Deloitte Australia and he had a good relationship with Yammer.

2. This analysis is a little different to a simulation. I would imagine you could create a simple agent model of with each agent having:
- propensity to join the CoP
- propensity to post
- propensity to reply
- propensity to leave the group

Some of these variables would be dependent on each other - e.g. propensity to leave the the group may go up for an agent if their posts are not replied to.


There's a stack of research in this area from both academics and commercial organisations:
https://scholar.google.com.au/scholar?hl=en&q=success+factors+for+online+communities

My suggestion would be to keep the model as simple as possible to begin with and compare it with real world data.

As you run your simulations, you can add more and more variables and subtlety. Do not try to get too clever too quickly.

I would hit up Microsoft (Yammer), IBM, Jive, Slack & Facebook - as I imagine that they have all thought about doing this even if they haven't actually done it.

Regards,

Matt

On Monday, 31 July 2017, 6:10, "Lee Romero pekadad@gmail.com [sikmleaders]" <sikmleaders@yahoogroups.com> wrote:




Hi Matt - For the initial work I'm keeping it simple and the "activities" I'm modeling are basically the equivalent of posting into an online community's discussion board tool (equivalent, roughly, to using a list server like SIKMLeaders uses or a Yammer Group or a FaceBook Group, etc).

Why am I doing this? I've had a number of discussions over the years with colleagues about strategies for managing communities of practice within an organization. The nature of the conversations is generally about the value of trying to have some control around creation of communities and also the value of trying to grow a community once it is established.

These overlap quite a bit - I find that if you don't try to control (even lightly) what communities exist, that leads to duplication, fragmentation and ultimately small and likely not-long-lived communities. On the other hand, if you have even some light control and try to actively encourage communities to merge when they are basically about the same topic, the result can be (I think) larger, more robust and therefore more successful communities.

While I was attending the Midwest KM Symposium in May, during one of the sessions (I forget which one - it might have been Stan Garfield's) the thought occurred to me that one assumption I make in my thinking on this is that to some extent, it reduces to math - given X people in a community if the likelihood of any person doing something at any time (whatever that "something" is - asking a question, sharing some insight, replying to someone else's question, etc.) is a relatively constant (small) percent, then as X goes up, activity goes up.

Given that kernel of an idea, I've been noodling over it for a while and thinking it seems feasible to approximately simulate this in a computer simulation. I've been thinking about what factors are at play here (it's not as simple as thinking "any given member has P probability to start a post or to reply to another post at any time") - need to consider factors like individual member's level of interest in the topic of the community, variations in the interestingess of individual posts, how "popular" a member might be [many communities have some members whose posts will garner more replies than most members], the fact that people have a constrained amount of time, etc.

I've already started working on how to simulate this mostly because it's an interesting experiment. As I've gotten closer to being able to (I think) somewhat realistically model communities, I wanted to see if anyone else has done anything like this (yeah, probably backwards I know but regardless, the work so far has been some interesting programming at least :) ).

Some of the specific questions I've captured that I think this could address (some are likely a stretch to say the least):

· Can you even simulate activity in a community in
a realistic manner?
· Does the size of a community relate to the level
of activity?
· What attributes are needed to properly model a
community?
· For those attributes, which values or
probabilistic models properly describe a community?
· How do interests of community member affect
activity?
· What kinds of activities should be considered?
· What is the effect of having multiple similar
(overlapping) communities on community viability?
· Is there a difference in behavior between a
community within a constrained organization (a company) and an effectively open
community (e.g., internet-based)?
· Does the 90-9-1 rule naturally flow out of any
of these models (or anything like it)?
· Is there a size of community that is too large
for the community to function properly?
· For a community manager – is there a size of
community that becomes unmanageable (or to ask a different way, can you make a
decision on how many community managers might be needed to adequately manage a
community based on its size)?




On Sun, Jul 30, 2017 at 3:21 PM, Matt Moore innotecture@yahoo.com [sikmleaders] <sikmleaders@yahoogroups.com> wrote:




Lee,


"I have recently been considering some ideas about trying to model (via computer simulation) activity within a community of practice."

Can you say a little bit more about what you are trying to do? What sort of activities will you be modelling? Why are you doing this?


The immediate thought that comes to mind is the work of Duncan Watts.


Regards,


Matt Moore
+61 423 784 504Sent from my iPhone
On Jul 31, 2017, at 4:45 AM, Lee Romero pekadad@gmail.com [sikmleaders] <sikmleaders@yahoogroups.com> wrote:



Hi all - I have recently been considering some ideas about trying to model (via computer simulation) activity within a community of practice.


There are several potential questions I would like to address in doing that.


But before I do too much, I wanted to see if there might be prior work in this.


Based on several searches in Google Scholar (and in the general Google search results) I don't find anything but it seems like something that must have been done before.


Does anyone in this community know of any prior work? If so, do you have links to it?


Thanks for any info!


Regards
Lee Romero


Laurence Lock Lee
 


Hi Lee, Matt,

What we have in our favour now in building let’s say a predictive model for a CoP is data….at least for online CoPs. IBM had done some work in trying to classify group types (of which CoPs were just one) using online activity data….

http://perer.org/papers/adamPerer-CHI2012.pdf  …. which we built on to develop a similar characterisation based on Yammer data collected by SWOOP … http://www.swoopanalytics.com/blog/swoop-video-blog-2-yammer-groups/  which uses the diversity vs cohesion frame for characterising group types.

Now if we make the assumption that CoPs tend to be bigger online groups, with an active core and a rather larger periphery (I call this the gallery) we can start to use the network performance frame (diversity vs cohesion) to assess CoP performance. Ideally you would want the CoP to maximise the diversity and the cohesion to optimise performance (in the absence of any context specific measures like cost saving generated etc…)

Now initially in the SWOOP link mentioned above I was using group size as a rough proxy for diversity. In our benchmarking work http://www.swoopanalytics.com/wp-content/uploads/2017/06/SWOOP-2017-ESN-Benchmark-Report.pdf we have a more sophisticated measure based on Blau’s diversity formula, which basically measures diversity based on an individual’s activity spread across multiple groups. The diversity of the CoP is then simply an aggregation of the individual scores. It turns out that Group Size is not a good proxy for the Blau diversity index….in fact the opposite. Its early day’s yet, but what I am starting to see is that a good proportion of members of large groups are only members of this single group (and usually are more observers, than active participants)….and therefore have zero diversity and hence pulling the average diversity of large groups/CoPs downward. This would suggest that their is an optimum size for a CoP…

I’m sorry about the rambling nature but I’m literally working on this now ….. and yes it will be possible to build a predictive model using regression equations and the diversity/cohesion index as a dependent variable. Activity and group size is likely to be a factor as well as maybe private/public group status…..but all work in progress…

I think the reason you aren’t finding much prior work Lee is the difficulty in accessing real-time operational data to conduct the research on. I know that the Sydney Uni Digital Disruption group took over a year  to clear Deloitte’s approval processes for a one off Yammer study. But we are lucky to have a mountain of data now from our SWOOP installations, where we can access this data anonymously…..exciting stuff!

rgds


Laurence  Lock Lee, PhD

Co-Founder & Chief Scientist
Ph: +61 (0)407001628
twitter:llocklee












Tony Melendez <bamaster@...>
 

Good day all,

 

Very interesting topic and responses so far!  I’d like to add to this.  A few years ago, while I was working at Marathon Oil, I delivered a presentation at APQC’s KM Conference called “Lessons Learned from Social Network Analysis Applied to Communities of Practice”.  It’s a high-level review of using SNA to look at CoPs.  In particular, how the memberships create larger patterns and groupings.

 

I’m no longer with Marathon Oil, but you’re welcome to look at the presentation.  It’s 28 MB because there’s an embedded video, so maybe download it and view it in presentation mode.

 

https://1drv.ms/p/s!An3p6JMBPo26jocFMuRq1RKKGPd7qg

(I hope that works)

 

Anyhow, this was a neat way to look at how out CoPs were forming and gave us perspective on how to improve the CoP environment and activities to create cross-connection… to break out of echo chambers.

 

Thank,

Tony Melendez

tony@...

 

 

From: sikmleaders@... [mailto:sikmleaders@...]
Sent: Sunday, July 30, 2017 11:11 PM
To: sikmleaders@...
Subject: Re: [sikmleaders] Prior work on modeling of community activity

 

 

Hi Matt - For the initial work I'm keeping it simple and the "activities" I'm modeling are basically the equivalent of posting into an online community's discussion board tool (equivalent, roughly, to using a list server like SIKMLeaders uses or a Yammer Group or a FaceBook Group, etc).

 

Why am I doing this?  I've had a number of discussions over the years with colleagues about strategies for managing communities of practice within an organization.  The nature of the conversations is generally about the value of trying to have some control around creation of communities and also the value of trying to grow a community once it is established.  

 

These overlap quite a bit - I find that if you don't try to control (even lightly) what communities exist, that leads to duplication, fragmentation and ultimately small and likely not-long-lived communities.   On the other hand, if you have even some light control and try to actively encourage communities to merge when they are basically about the same topic, the result can be (I think) larger, more robust and therefore more successful communities.

 

While I was attending the Midwest KM Symposium in May, during one of the sessions (I forget which one - it might have been Stan Garfield's) the thought occurred to me that one assumption I make in my thinking on this is that to some extent, it reduces to math - given X people in a community if the likelihood of any person doing something at any time (whatever that "something" is - asking a question, sharing some insight, replying to someone else's question, etc.) is a relatively constant (small) percent, then as X goes up, activity goes up.

 

Given that kernel of an idea, I've been noodling over it for a while and thinking it seems feasible to approximately simulate this in a computer simulation.  I've been thinking about what factors are at play here (it's not as simple as thinking "any given member has P probability to start a post or to reply to another post at any time") - need to consider factors like individual member's level of interest in the topic of the community, variations in the interestingess of individual posts, how "popular" a member might be [many communities have some members whose posts will garner more replies than most members], the fact that people have a constrained amount of time, etc.

 

I've already started working on how to simulate this mostly because it's an interesting experiment.  As I've gotten closer to being able to (I think) somewhat realistically model communities, I wanted to see if anyone else has done anything like this (yeah, probably backwards I know but regardless, the work so far has been some interesting programming at least :) ).

 

Some of the specific questions I've captured that I think this could address (some are likely a stretch to say the least):

 

·         Can you even simulate activity in a community in a realistic manner?

·         Does the size of a community relate to the level of activity?

·         What attributes are needed to properly model a community?

·         For those attributes, which values or probabilistic models properly describe a community?

·         How do interests of community member affect activity?

·         What kinds of activities should be considered?

·         What is the effect of having multiple similar (overlapping) communities on community viability?

·         Is there a difference in behavior between a community within a constrained organization (a company) and an effectively open community (e.g., internet-based)?

·         Does the 90-9-1 rule naturally flow out of any of these models (or anything like it)?

·         Is there a size of community that is too large for the community to function properly?

·         For a community manager – is there a size of community that becomes unmanageable (or to ask a different way, can you make a decision on how many community managers might be needed to adequately manage a community based on its size)?

 

 

 

On Sun, Jul 30, 2017 at 3:21 PM, Matt Moore innotecture@... [sikmleaders] <sikmleaders@...> wrote:

 

Lee,

 

"I have recently been considering some ideas about trying to model (via computer simulation) activity within a community of practice."

 

Can you say a little bit more about what you are trying to do? What sort of activities will you be modelling? Why are you doing this?

 

The immediate thought that comes to mind is the work of Duncan Watts.

 

Regards,

 

Matt Moore

On Jul 31, 2017, at 4:45 AM, Lee Romero pekadad@... [sikmleaders] <sikmleaders@...> wrote:

 

Hi all - I have recently been considering some ideas about trying to model (via computer simulation) activity within a community of practice.  

 

There are several potential questions I would like to address in doing that.

 

But before I do too much, I wanted to see if there might be prior work in this.

 

Based on several searches in Google Scholar (and in the general Google search results) I don't find anything but it seems like something that must have been done before.

 

Does anyone in this community know of any prior work?  If so, do you have links to it?

 

Thanks for any info!

 

Regards

Lee Romero

 

 


 

Lee,
I forgot to post the link to the Wenger, Trayner, deLaat paper
http://wenger-trayner.com/wp-content/uploads/2011/12/11-04-Wenger_Trayner_DeLaat_Value_creation.pdf

Catherine


Social Collaboration and Digital Transformation
Silicon Valley, USA
+1-650-704-3889

Contributor to Smarter Innovation (chapter abstracts) (Ark Group, 2014)
Blog: www.collaboration-incontext.com


On Mon, Jul 31, 2017 at 12:17 AM, Laurence Lock Lee llocklee@... [sikmleaders] <sikmleaders@...> wrote:
 


Hi Lee, Matt,

What we have in our favour now in building let’s say a predictive model for a CoP is data….at least for online CoPs. IBM had done some work in trying to classify group types (of which CoPs were just one) using online activity data….

http://perer.org/papers/adamPerer-CHI2012.pdf  …. which we built on to develop a similar characterisation based on Yammer data collected by SWOOP … http://www.swoopanalytics.com/blog/swoop-video-blog-2-yammer-groups/  which uses the diversity vs cohesion frame for characterising group types.

Now if we make the assumption that CoPs tend to be bigger online groups, with an active core and a rather larger periphery (I call this the gallery) we can start to use the network performance frame (diversity vs cohesion) to assess CoP performance. Ideally you would want the CoP to maximise the diversity and the cohesion to optimise performance (in the absence of any context specific measures like cost saving generated etc…)

Now initially in the SWOOP link mentioned above I was using group size as a rough proxy for diversity. In our benchmarking work http://www.swoopanalytics.com/wp-content/uploads/2017/06/SWOOP-2017-ESN-Benchmark-Report.pdf we have a more sophisticated measure based on Blau’s diversity formula, which basically measures diversity based on an individual’s activity spread across multiple groups. The diversity of the CoP is then simply an aggregation of the individual scores. It turns out that Group Size is not a good proxy for the Blau diversity index….in fact the opposite. Its early day’s yet, but what I am starting to see is that a good proportion of members of large groups are only members of this single group (and usually are more observers, than active participants)….and therefore have zero diversity and hence pulling the average diversity of large groups/CoPs downward. This would suggest that their is an optimum size for a CoP…

I’m sorry about the rambling nature but I’m literally working on this now ….. and yes it will be possible to build a predictive model using regression equations and the diversity/cohesion index as a dependent variable. Activity and group size is likely to be a factor as well as maybe private/public group status…..but all work in progress…

I think the reason you aren’t finding much prior work Lee is the difficulty in accessing real-time operational data to conduct the research on. I know that the Sydney Uni Digital Disruption group took over a year  to clear Deloitte’s approval processes for a one off Yammer study. But we are lucky to have a mountain of data now from our SWOOP installations, where we can access this data anonymously…..exciting stuff!

rgds


Laurence  Lock Lee, PhD

Co-Founder & Chief Scientist
twitter:llocklee