Reputation

[Monica Keller]

What do you guys think of extending http://portablecontacts.net/draft-schema.html#account_element
with some reputation indicators.

Today we are being asked for: Number of friends/ Followers and Length
of time the account has been in existance. I am sure more metrics will
be requested but in the meantime what do you guys think of

<activity:actor>
<activity:object-type>http://activitystrea.ms/schema/1.0/person</
activity:object-type>
<id>tag:myspace.com,2009:/Person/4859568</id>
<title>Monica Keller</title>
<link rel="alternate" type="text/html" href="http://
profile.myspace.com/index.cfm?
fuseaction=user.viewprofile&amp;friendid=4859568" />
<link rel="avatar" type="image/jpeg" href="http://c1.ac-
images.myspacecdn.com/images01/81/
s_6e92c7004ff96f1295370b6254896a68.jpg" />
<poco:name>
<poco:givenName>Monica</poco:givenName>
<poco:familyName>Keller</poco:familyName>
</poco:name>
<poco:displayName>Monica Keller</poco:displayName>
<poco:preferredUsername>Monica 3.0</poco:preferredUsername>
<poco:address>
<poco:locality>Marina del Rey</poco:locality>
<poco:postalCode>90292</poco:postalCode>
<poco:country>US</poco:country>
</poco:address>
<poco:accounts>
<poco:account>
<poco:domain>myspace.com</poco:domain>
<poco:user_id>121</poco:user_id>
<poco:friends_count>10</poco:friends_count>
<poco:follower_count>2</poco:follower_count>
<poco:sign_up_date>2005-11-17T22:13:21Z</poco:sign_up_date>
</poco:account>
</poco:accounts>

For an example of this see:
http://apiwiki.twitter.com/Twitter-REST-API-Method%3A-statuses%C2%A0update

[Steven Livingstone-Perez]

I had a stab at reputation a while back and did quite a bit of work on it.

It's easy enough to add this stuff but quickly becomes of little real
value - depending on what you want to achieve of course.

As an example, plugging Twitter into this would be completely meaningless do
to various issues such as auto-follow, gaming (which is very easy to do on
twitter and other social networks) and others.

*Personally* I think reputation is a whole new area and something i'd love
to get involved in. I know Chris Messina mentioned it on a tweet a while
back (erm i think it was twitter - their search isn't great) but not sure if
he had started some thinking around it.

I do think that to start it one would need to define the parameters of what
it was intended to capture.

regards,
steven
http://livz.org

--------------------------------------------------
From: "Monica Keller" <monica...@gmail.com>
Sent: Tuesday, November 17, 2009 11:30 PM
To: "Activity Streams" <activity...@googlegroups.com>
Cc: <portable...@googlegroups.com>
Subject: Reputation

> --
>
> You received this message because you are subscribed to the Google Groups
> "Activity Streams" group.
> To post to this group, send email to activity...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/activity-streams?hl=.
>
>
>

[Peter Saint-Andre]

On 11/17/09 4:45 PM, Steven Livingstone-Perez wrote:
> I had a stab at reputation a while back and did quite a bit of work on it.
>
> It's easy enough to add this stuff but quickly becomes of little real
> value - depending on what you want to achieve of course.
>
> As an example, plugging Twitter into this would be completely meaningless do
> to various issues such as auto-follow, gaming (which is very easy to do on
> twitter and other social networks) and others.
>
> *Personally* I think reputation is a whole new area and something i'd love
> to get involved in. I know Chris Messina mentioned it on a tweet a while
> back (erm i think it was twitter - their search isn't great) but not sure if
> he had started some thinking around it.
>
> I do think that to start it one would need to define the parameters of what
> it was intended to capture.

Agreed, reputation is a huge topic and not one to be considered lightly.
There's a large volume of experience (e.g., Slashdot, Digg, eBay,
Amazon) and analysis (see the academic literature) on such systems, but
IMHO the devil really is in the details here (what kind of community
context can you assume, etc.).

Peter

--
Peter Saint-Andre
https://stpeter.im/

[Christian Crumlish]

On 11/17/09 4:45 PM, Steven Livingstone-Perez wrote:
> *Personally* I think reputation is a whole new area and something i'd love
> to get involved in. I know Chris Messina mentioned it on a tweet a while
> back (erm i think it was twitter - their search isn't great) but not sure if
> he had started some thinking around it.
>
> I do think that to start it one would need to define the parameters of what
> it was intended to capture.

On Tue, Nov 17, 2009 at 4:55 PM, Peter Saint-Andre <stp...@stpeter.im> wrote:

Agreed, reputation is a huge topic and not one to be considered lightly.
There's a large volume of experience (e.g., Slashdot, Digg, eBay,
Amazon) and analysis (see the academic literature) on such systems, but
IMHO the devil really is in the details here (what kind of community
context can you assume, etc.).

it almost seems like its own namespace

-x-

p.s.: see also http://developer.yahoo.com/ypatterns/social/people/reputation/

-- 

Christian Crumlish

MY NEW BOOK: Designing Social Interfaces. http://designingsocialinterfaces.com
Get It. Read It. Love It. Review it. on Amazon: http://www.amazon.com/dp/0596154925/

[Chris Messina]

On Tue, Nov 17, 2009 at 4:59 PM, Christian Crumlish <xi...@pobox.com> wrote:

it almost seems like its own namespace

-x-

p.s.: see also http://developer.yahoo.com/ypatterns/social/people/reputation/

Well, I think what Monica is talking about are more like "stats" than reputation.

Not all services count such things, but many could provide an aggregate count as to the number of friends or contacts someone has, or the number of fans or followers they've accrued. Twitter, Facebook, or Amazon might also provide the number of lists they've made (for Amazon, it'd be wishlists), or other stats.

I wouldn't call this "reputation" because of the weight (and near impossibility) of the subject matter — especially when applying it to two disjoint social contexts.

I think this at least bears inspection as Twitter is providing this information. For now, I've created a page on the wiki to explore this topic further:

https://activitystreams.pbworks.com/Stats

Chris

--
Chris Messina
Open Web Advocate

Personal: http://factoryjoe.com
Follow me on Twitter: http://twitter.com/chrismessina

Citizen Agency: http://citizenagency.com
Diso Project: http://diso-project.org
OpenID Foundation: http://openid.net

This email is:   [ ] shareable    [X] ask first   [ ] private

[John Panzer]

+1 to stats. And especially to life-of-account, as this is expensive
to game without a time machine.

Note that even with minimal context, it is possible to do useful
statistical analysis and classification based on these types of
features.

On Tuesday, November 17, 2009, Allen Tom <at...@yahoo-inc.com> wrote:
>
>
>
>
>
>
> On a closely related note, many sites have asked us to add "sign up
> date" as well as other reputation attributes to our OpenID service,
> mostly for anti-abuse purposes. This would be a useful AX attribute,
> especially if OPs are willing to standardize on this.
>
> Allen

> --
> You received this message because you are subscribed to the Google

> Groups "PortableContacts" group.
> To post to this group, send email to portable...@googlegroups.com.

> For more options, visit this group at

> http://groups.google.com/group/portablecontacts?hl=.
>
>
>
>
>
>
>
> --
>
> You received this message because you are subscribed to the Google Groups "PortableContacts" group.
> To post to this group, send email to portable...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/portablecontacts?hl=.
>

--
--
John Panzer / Google
jpa...@google.com / abstractioneer.org / @jpanzer

[Christian Crumlish]

On Tue, Nov 17, 2009 at 8:54 PM, John Panzer wrote:

+1 to stats.  And especially to life-of-account, as this is expensive to game without a time machine.

Note that even with minimal context, it is possible to do useful
statistical analysis and classification based on these types of
features.

ah, i follow now.

so is there such a thing as a "vocabulary" of relevant stats? (standard, or typical stats; or some canonical way of generating all possible stats, or something, based on the other elements in the xml dialect?)

-x-

 

[Steven Livingstone-Perez]

Yes - stats seems more appropriate and i think it's a useful thing to have.

 

I know an approach at reputation would be difficult but perhaps not impossible. The Yahoo link is an excellent starter here.

 

The issue IMHO is that "reputation" is quite a strong word and as such could potentially effect how you are perceived in the world. I might not use a given network very much and my reputation on that network may be classed as "unhelpful" simply because i do all my interaction on network b.

 

I do think something could be done, but only in an open way and in conjunction with the various social providers. They can help determine what defines reputation on their system and the user themselves can decide how that incorporates into their profile.

 

However, just to get something started i think a bunch of scenarios would need to be drawn up to consider possible approaches.

 

Good you've got the ball rolling on discussion anyway

 

steven

http://livz.org

--

You received this message because you are subscribed to the Google Groups "PortableContacts" group.
To post to this group, send email to portable...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/portablecontacts?hl=.

[Matthias Pfefferle]

What about the "User Labor Markup Language"?

[citation]With User Labor, we propose an open data structure, User
Labor Markup Language (ULML), to outline the metrics of user
participation in social web services. Our aim is to construct criteria
and context for determining the value of user labor, which is
currently a monetized asset for the service provider but not for the
user herself. We believe that universal, transparent, and self-
controlled user labor metrics will ultimately lead to more sustainable
social web.[/citation]

Link: http://userlabor.org/

On 18 Nov., 10:28, "Steven Livingstone-Perez" <webl...@hotmail.com>
wrote:

> Yes - stats seems more appropriate and i think it's a useful thing to have.
>
> I know an approach at reputation would be difficult but perhaps not impossible. The Yahoo link is an excellent starter here.
>
> The issue IMHO is that "reputation" is quite a strong word and as such could potentially effect how you are perceived in the world. I might not use a given network very much and my reputation on that network may be classed as "unhelpful" simply because i do all my interaction on network b.
>
> I do think something could be done, but only in an open way and in conjunction with the various social providers. They can help determine what defines reputation on their system and the user themselves can decide how that incorporates into their profile.
>
> However, just to get something started i think a bunch of scenarios would need to be drawn up to consider possible approaches.
>
> Good you've got the ball rolling on discussion anyway
>

> stevenhttp://livz.org

>
> From: Chris Messina
> Sent: Wednesday, November 18, 2009 2:17 AM
> To: activity...@googlegroups.com ; portable...@googlegroups.com
> Subject: Re: Reputation
>

> On Tue, Nov 17, 2009 at 4:59 PM, Christian Crumlish <x...@pobox.com> wrote:
>
>   it almost seems like its own namespace
>
>   -x-
>

>   p.s.: see alsohttp://developer.yahoo.com/ypatterns/social/people/reputation/

>
> Well, I think what Monica is talking about are more like "stats" than reputation.
>
> Not all services count such things, but many could provide an aggregate count as to the number of friends or contacts someone has, or the number of fans or followers they've accrued. Twitter, Facebook, or Amazon might also provide the number of lists they've made (for Amazon, it'd be wishlists), or other stats.
>
> I wouldn't call this "reputation" because of the weight (and near impossibility) of the subject matter — especially when applying it to two disjoint social contexts.
>
> I think this at least bears inspection as Twitter is providing this information. For now, I've created a page on the wiki to explore this topic further:
>
> https://activitystreams.pbworks.com/Stats
>
> Chris
>
> --
> Chris Messina
> Open Web Advocate
>
> Personal:http://factoryjoe.com
> Follow me on Twitter:http://twitter.com/chrismessina
>
> Citizen Agency:http://citizenagency.com
> Diso Project:http://diso-project.org
> OpenID Foundation:http://openid.net
>
> This email is:   [ ] shareable    [X] ask first   [ ] private
>
> --
>
> You received this message because you are subscribed to the Google Groups "PortableContacts" group.
> To post to this group, send email to portable...@googlegroups.com.
> For more options, visit this group athttp://groups.google.com/group/portablecontacts?hl=.
>

>  Emoticon1.gif
> < 1 KBAnzeigenHerunterladen

[Steven Livingstone-Perez]

Thanks Matthias - i hadn't seen this.

Yes - there is some interesting thinking in there - not sure i like the term
(maybe just me - i keep reading "UML") but at least it's distinct from
reputation.

The issue is that metrics are domain specific - 29 on one network may mean
2900 on another and so on.

So in ULML there would need to be something indicating the metric domain in
a given channel (i suspect this could be abstracted from the channel link).
I know in the FAQ it says it does nothing with metric calculation - this may
be useful in a first iteration but i don't think people would want to do
this themselves in the longer term. Guidance for domain specific metrics
would need to be given i think.

It is certainly useful in basic serialization of your domain specific social
metrics.

steven
http://livz.org

--------------------------------------------------
From: "Matthias Pfefferle" <pfef...@gmail.com>
Sent: Wednesday, November 18, 2009 10:27 AM
To: "Activity Streams" <activity...@googlegroups.com>
Subject: Re: Reputation

>> impossibility) of the subject matter � especially when applying it to two

> "Activity Streams" group.
> To post to this group, send email to activity...@googlegroups.com.

> For more options, visit this group at

> http://groups.google.com/group/activity-streams?hl=.
>
>
>

[Steven Livingstone-Perez]

I suspect there is quite a lot that could be done here if wish to extend this concept.

 

For example, as well as sign up date some may be able to provide a metric indicating a level of verification - e.g. just assigned an OpenID, verified public email account (on a public site - yahoo, hotmail etc), verified company email (on a company domain) and so on.

 

This has been at the back of my mind for a couple of years now but not sure whether any discussion has happened around it.

 

/steven

http://livz.org

From: Allen Tom

Sent: Wednesday, November 18, 2009 2:48 AM

To: portable...@googlegroups.com ; activity...@googlegroups.com ; sp...@openid.net

Subject: Re: Reputation

On a closely related note, many sites have asked us to add "sign up date" as well as other reputation attributes to our OpenID service, mostly for anti-abuse purposes. This would be a useful AX attribute, especially if OPs are willing to standardize on this.

Allen

Chris Messina wrote:

--

You received this message because you are subscribed to the Google Groups "PortableContacts" group.
To post to this group, send email to portable...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/portablecontacts?hl=.

--

You received this message because you are subscribed to the Google Groups "PortableContacts" group.
To post to this group, send email to portable...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/portablecontacts?hl=.

[Sylvain Hellegouarch]

Incidentally, Clay Shirky released an interesting note [1] regarding
how to define what could be considered authoritative. This really a
subject of importance, think of a band that want to get promoted,
finding the authoritative source of reviewers for instance is
critical.

- Sylvain

[1] http://www.shirky.com/weblog/2009/11/a-speculative-post-on-the-idea-of-algorithmic-authority/

[Monica Keller]

Thanks for the great feedback guys, this is the benefit of
brainstorming in the open :P.

So yes I agree to ditch the term reputation. Reputation implies
subjectivity and this data I am trying to convey is very specific.

It is important to share this information when we exchange user
generated content so the external parties have some idea on the value
of the content when surfaced to their users. Activities/Tweets are
fairly compact so there is not a lot to match on. Who the author is
adds value and how others respond to him/her definitely adds value. I
would love to get more feedback on what others think good metrics are

For now http://userlabor.org/ looks very aligned with what we want.
The only piece that I do not understand is network given that it seems
they do not account for bidirectional and unidirectional relationships
or how to calculate the values below for density, betweenness and
closeness... I am assuming its out of 1 ? Ha if the authors are on
this mailing list, let us know.

<network>
<item object="connection">269</item>
<item object="density">0.101</item>
<item object="betweenness">0.225</item>
<item object="closeness">0.700</item>
</network>


I should have left in my paragraph about what I meant by reputation :)
its just really a summary of the user's activity on a social network
that can be used externally to decide how to rank the user's content.
So for example if the user just signed up it may be a bot and be
publishing spammy links.



On Nov 18, 7:26 am, Sylvain Hellegouarch

> [1]http://www.shirky.com/weblog/2009/11/a-speculative-post-on-the-idea-o...- Hide quoted text -
>
> - Show quoted text -

[Chris Messina]

On Wed, Nov 18, 2009 at 8:57 AM, Monica Keller <monica...@gmail.com> wrote:

It is important to share this information when we exchange user
generated content so the external parties have some idea on the value
of the content when surfaced to their users. Activities/Tweets are
fairly compact so there is not a lot to match on. Who the author is
adds value and how others respond to him/her definitely adds value. I
would love to get more feedback on what others think good metrics are

For now http://userlabor.org/ looks very aligned with what we want.
The only piece that I do not understand is network given that it seems
they do not account for bidirectional and unidirectional relationships
or how to calculate the values below for density, betweenness and
closeness... I am assuming its out of 1 ? Ha if the authors are on
this mailing list, let us know.

<network>
       <item object="connection">269</item>
       <item object="density">0.101</item>
       <item object="betweenness">0.225</item>
       <item object="closeness">0.700</item>
     </network>

I looked at UserLabor a long time ago and had hopes that it could be the model for Activity Streams. Their model ended up being a bit too esoteric and original for my tastes, and hence our work was born.

Whether we can leverage their <network> concept is worth considering — though I don't know that it actually accurately models what we need. 

APML also tried to provide a mechanism for weighting tags some time ago, but that's not quite what we want either.

I worry a little about reaching beyond the core of what Activity Streams were designed to do, but I understand Monica's desire to meet developer's thirst for more data about the actor.

Finding a convention to convey stats certainly has merit and is worth considering.

I would be interested to know if anyone has examples of other networks providing friends/follower/contact counts in their APIs, and if so, how they represent such attributes?

[Chris Messina]

Very interesting!

I've fleshed out some more details on the wiki page:

https://activitystreams.pbworks.com/Stats

Please add more attributes as you think of them!

Chris

On Tue, Nov 17, 2009 at 6:48 PM, Allen Tom <at...@yahoo-inc.com> wrote:

On a closely related note, many sites have asked us to add "sign up date" as well as other reputation attributes to our OpenID service, mostly for anti-abuse purposes. This would be a useful AX attribute, especially if OPs are willing to standardize on this.

Allen

Chris Messina wrote:

--

You received this message because you are subscribed to the Google Groups "PortableContacts" group.
To post to this group, send email to portable...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/portablecontacts?hl=.

--

You received this message because you are subscribed to the Google Groups "PortableContacts" group.
To post to this group, send email to portable...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/portablecontacts?hl=.

[Peter H. Reiser]

wow - the ULML (uuh what an abbreviation - in German we call this a "Zungenbrecher" :-)  ) specs seem to be a  complementary fit to the Community Equity  (CeQ) concept (see specification)

e.g,

ULML  CeQ                           

actions  Contribution Equity (CQ)

reactions Participation Equity (PQ)

The  main difference is that CeQ is calculating all the value associations based on single activities (hence we love activitystrea.ms :-) ) whereby ULML provide sum of activities.

Although  there might be a few missing elements

* timeline  when the action/reactions happened (from - to )  -> maybe this could be calculated with MemberSince and PubDate ...

* aging - in the CeQ model we have an aging time per social activity  (e.g. (aging(CQ) , aging(PQ) ) . This allows us to calculate "Reputation aging". . We would need to thing about a model to provide  aging  across   

   sum(activities)

* tags  related to an activity object -  we use Tag Equity to  aggregated values across multiple sites e.g. example: you blog about "Java"  using the tag "Java" and then twitter about "Java" using #Java -  CeQ  calculates your value around "Java"  e.g TQ( PersonJava))  

so a ULML structure could look like

 <title>Facebook | Barack Oboe</title>
    <link>http://mit.facebook.com/profile.php?id=619187456</link>
    <description>Barack Oboe's profile on Facebook.</description>
    <pubDate>Thu, 06 Feb 2008 20:55:01 GMT</pubDate>    
    <user>Barack Oboe</user>
    <memberSince>Thu, 01 Jun 2005 20:00:01 GMT</memberSince>    
    <record>
      <actions>

	<item object="TotalCQ" verb="CQ">12335</item> * Total value of all contribution actions (own and received activities) of a user at this moment (aging)

        <item object="photo" verb="TQ">123</item>   * Total of all CQ activities values of this user for all content objects tagged with "photo"
      </actions>
      <reactions>

	<item object="TotalPQ" verb="PQ">12345</item> * Total value of all participation activities  of a user at this moment (aging) 

	<item object="photo" verb="TQ">18</item>   * Total of all PQ activities values of this user  for all content objects (own and others) tagged with "photo"

      </reactions>

Peter

impossibility) of the subject matter — especially when applying it to two

[Elias Bizannes]

Wow, great discussion. This has a lot of potential.

The mapping of social networks has had a huge body of study in enterprises, and people have been talking about it for a while[1]. My former employer PricewaterhouseCoopers spent a bucket on analysing internal networks. The reason being, informal networks within a company is how information flows, and so it was thought by studying it and determining who the connectors are and to what groups, there could be better organisational planning.

So a spec on mapping networks is huge. For a start, it creates a tool to start enabling better filtering on the web - and whenever the social graph is thrown around as a world changing concept, it's the ability for our graphs to be able to filter the world. This has more value than any other thing I can think of. To put it more bluntly, importing my friends' details is one thing and caters to my laziness, but leveraging it to filter my world, is a whole different level and quite transformative.

I looked at UserLabor a long time ago and had hopes that it could be the model for Activity Streams. Their model ended up being a bit too esoteric and original for my tastes, and hence our work was born.

Really? I think they've over-simplified it. There's some body of research by McKinsey on the subject and the UML spec looks very similar to it. Basically, McKinsey studied these informal networks not just how many and to whom people were connected to, but also the quality of those relationships (ie, frequency of contact). I can sympathise with people getting to academic on things and rendering something impractical, but I'm curious to think how those four measures are too esoteric?

APML also tried to provide a mechanism for weighting tags some time ago, but that's not quite what we want either.

What's the weakness with APML?
 

I worry a little about reaching beyond the core of what Activity Streams were designed to do, but I understand Monica's desire to meet developer's thirst for more data about the actor.

I agree and think the above discussion should make its way into a new effort. The more you dig into this, the more we'll realise how many ways it can go.

[1] http://semanticstudios.com/publications/semantics/000006.php

[adrian chan]

I just want to toss my hat in the ring and offer to pitch in to any future work this or similar efforts might make towards a network/status metric, as well as other attributes and actions that exceed the scope of activity streams. 

I agree that network position, as status or rank, will vary from social network to social network. Clearly the distinctions between open/closed, a/symmetrical, not to mention the cultures, themes, and primary activities of different social networks contribute to the relative importance of social and network rank attributes. 

Which makes me wonder, if we were able to codify a set of attributes, as well as use 'reactions' as a means of confirming/validating them, could one have a sort of independent authority by which to broker and validate status/rank? Combine that with topicality and you have something pretty powerful -- expertise, credibility, reputation, by domain, activity, and so on. Add to that 'recent activity' and other metrics of user activity and you have a distributed social data on the user, his/her social position, in what domains of interest, validated by his/her network audiences plus brokered independently, and activity history.... 

Would seem possible to qualify network stats from major socnets. I dont know. I'm probably being naive. 

Either way, count me in if such an effort were to kick off. 

Adrian Chan

--

You received this message because you are subscribed to the Google Groups "Activity Streams" group.
To post to this group, send email to activity...@googlegroups.com.

To unsubscribe from this group, send email to activity-strea...@googlegroups.com.

For more options, visit this group at http://groups.google.com/group/activity-streams?hl=.

--

415 516 4442     Twitter: @gravity7
Social Interaction Design, Expertise, Consulting (gravity7) (gravity7 blog) (slideshare)
Sr Fellow, Society for New Communications Research (SNCR)
LinkedIn (www.linkedin.com/in/adrianchan)
Facebook  (www.facebook.com/adrianchan)

[Kaliya *]

On Wed, Nov 18, 2009 at 8:57 AM, Monica Keller <monica...@gmail.com> wrote:

Thanks for the great feedback guys, this is the benefit of
brainstorming in the open :P.

So yes I agree to ditch the term reputation. Reputation implies
subjectivity and this data I am trying to convey is very specific.

This is great news.

I have been concerned for some time with the use of the word reputation used to describe "numbers" when it is not really possible to capture it in numbers.

Reputation is a subjective judgment we make in our heads with a range of information we have about people based on our own personal interactions and information we can find out about people/organizations/sites.

I was calling what people did call "reputation" "personal transaction history" this could of course be shortened to "stats" 

 

[Steve Ivy]

"Whuffie"? :-P

> To unsubscribe from this group, send email to
> activity-strea...@googlegroups.com.

> For more options, visit this group at
> http://groups.google.com/group/activity-streams?hl=.
>

--
Steve Ivy
http://redmonk.net // http://diso-project.org
This email is: [ ] bloggable [x] ask first [ ] private

[Melvin Carvalho]

On Mon, Nov 23, 2009 at 5:50 AM, Kaliya * <identi...@gmail.com> wrote:
>
>
> On Wed, Nov 18, 2009 at 8:57 AM, Monica Keller <monica...@gmail.com>
> wrote:
>>
>> Thanks for the great feedback guys, this is the benefit of
>> brainstorming in the open :P.
>>
>> So yes I agree to ditch the term reputation. Reputation implies
>> subjectivity and this data I am trying to convey is very specific.
>
> This is great news.
>
> I have been concerned for some time with the use of the word reputation used
> to describe "numbers" when it is not really possible to capture it in
> numbers.
>
> Reputation is a subjective judgment we make in our heads with a range of
> information we have about people based on our own personal interactions and
> information we can find out about people/organizations/sites.

I think there's a subtle difference between trust and reputation.
Technically trust can be mapped to a number [0,1] which is the
probability that a certain (future) action will be performed as
expected.

Reputation is the set of data that is used to calculate a trust score.
Data is of various types. Observed and Reported is one axis.
Individual and group is another. The mapping of reputation events
onto a trust store can be extremely complex, and will indeed involve
subjectivity or "value system". That's not to say you cant produce
useful metrics depending on the subsystem you are building. Ideal is
to put the data out there and allow people to work out the best
algorithms that suit.

> To unsubscribe from this group, send email to
> activity-strea...@googlegroups.com.

[Kaliya *]

On Mon, Nov 23, 2009 at 6:04 AM, Melvin Carvalho <melvinc...@gmail.com> wrote:

On Mon, Nov 23, 2009 at 5:50 AM, Kaliya * <identi...@gmail.com> wrote:
>
>
> On Wed, Nov 18, 2009 at 8:57 AM, Monica Keller <monica...@gmail.com>
> wrote:
>>
>> Thanks for the great feedback guys, this is the benefit of
>> brainstorming in the open :P.
>>
>> So yes I agree to ditch the term reputation. Reputation implies
>> subjectivity and this data I am trying to convey is very specific.
>
> This is great news.
>
> I have been concerned for some time with the use of the word reputation used
> to describe "numbers" when it is not really possible to capture it in
> numbers.
>
> Reputation is a subjective judgment we make in our heads with a range of
> information we have about people based on our own personal interactions and
> information we can find out about people/organizations/sites.

I think there's a subtle difference between trust and reputation.
Technically trust can be mapped to a number [0,1] which is the
probability that a certain (future) action will be performed as
expected.

Trust is a feeling that is generated by a whole complex of human social interactions.
 People are designed to read and manage face to face human social interactions - we have evolved for millions of years to do this.

Reputation is the set of data that is used to calculate a trust score.
 Data is of various types.  Observed and Reported is one axis.
Individual and group is another.  The mapping of reputation events
onto a trust store can be extremely complex, and will indeed involve
subjectivity or "value system".

You can score all kinds of things. Trust I don't believe is not one of them. It is dangerous for us to lable or call any "number" no mater how complexly we make its algorithm.
 

 That's not to say you cant produce
useful metrics depending on the subsystem you are building.  Ideal is
to put the data out there and allow people to work out the best
algorithms that suit.

Agreed - people need a range of measures that they themselves can mix and match.

-Kaliya
 

[John Panzer]

On Mon, Nov 23, 2009 at 9:57 AM, Kaliya * <identi...@gmail.com> wrote:

On Mon, Nov 23, 2009 at 6:04 AM, Melvin Carvalho <melvinc...@gmail.com> wrote:

On Mon, Nov 23, 2009 at 5:50 AM, Kaliya * <identi...@gmail.com> wrote:
>
>
> On Wed, Nov 18, 2009 at 8:57 AM, Monica Keller <monica...@gmail.com>
> wrote:
>>
>> Thanks for the great feedback guys, this is the benefit of
>> brainstorming in the open :P.
>>
>> So yes I agree to ditch the term reputation. Reputation implies
>> subjectivity and this data I am trying to convey is very specific.
>
> This is great news.
>
> I have been concerned for some time with the use of the word reputation used
> to describe "numbers" when it is not really possible to capture it in
> numbers.
>
> Reputation is a subjective judgment we make in our heads with a range of
> information we have about people based on our own personal interactions and
> information we can find out about people/organizations/sites.

I think there's a subtle difference between trust and reputation.
Technically trust can be mapped to a number [0,1] which is the
probability that a certain (future) action will be performed as
expected.

Trust is a feeling that is generated by a whole complex of human social interactions.
 People are designed to read and manage face to face human social interactions - we have evolved for millions of years to do this.

Absolutely true.  People's brains have evolved over millions of years to manage social relationships which are largely about trust decisions.  In fact this may have been the primary driver in brain development over the past 3M years.  This is the the context of a relatively small number of familiar people (a few dozen, up to Dunbar's number) with additional mechanisms layered on to help us scale past that number (all of modern human society).  

We should provide data that will let human brains do what they do really well, and let computers do what they do really well.  Human brains do really well when they can see faces, when they make trust decisions about a relatively small number (<~150 or so) of familiar people, and when they have history to fall back on and future relationships to use for retribution/reward.

Human brains don't do very well when dealing with other things, like lots of numbers, low probability but high impact events, instantaneous transmission of private data worldwide, or people who can change their faces/identities at will.  Computers should focus on helping to deal with these things wherever possible.

[Chris Messina]

On this subject, I find this post on Robo.to's new "relationship
strength" feature quite interesting:

http://massive.ly/how-to-set-up-relationship-strength

Clearly there's a desire, even if subconscious, to make our robots and
computers more empathic and able to quantify intimacy along a
proximity gradient.

This'll get worse before it gets better, or weirder before it gets
more normal. Funny how we change the computers, and they change us.

Chris

[Melvin Carvalho]

On Tue, Nov 24, 2009 at 1:51 AM, Chris Messina <chris....@gmail.com> wrote:
> On this subject, I find this post on Robo.to's new "relationship
> strength" feature quite interesting:
>
> http://massive.ly/how-to-set-up-relationship-strength

An interesting link. I think relationship strength was discussed a
while back on Social Network Portability. I think it's an important
point, though can be hard to model. In fact in mathematical terms the
difference between a social graph and a social network is that the
network has relationship strength. The theory and algorithms of graph
theory differ widely from that of network theory. So there is a lot
of depth there.

For me, the classic paper on Trust and Reputation is still:

http://groups.csail.mit.edu/medg/ftp/lmui/computational%20models%20of%20trust%20and%20reputation.pdf

Quite a long paper, and obviously very old, but if you are interested
in the field and background, I think could can do a lot worse than
reading the first couple of chapters...

> --
>
> You received this message because you are subscribed to the Google Groups "Activity Streams" group.
> To post to this group, send email to activity...@googlegroups.com.
> To unsubscribe from this group, send email to activity-strea...@googlegroups.com.

> For more options, visit this group at http://groups.google.com/group/activity-streams?hl=en.
>
>
>

[Jonathan Dugan]

We implemented stats (basically activity counting) in the Rails
library with model ActivityStreamTotal defaults to an Fixnum (an int)
or can be overwritten with a reference to a variable or a method at
install.

http://github.com/face/activity_streams/blob/master/lib/models/activity_stream_total.rb

"""
options: A hash of options. Currently the options can take:
:status for non display-able, debug, or internal activities
:total for keeping total counts across grouped activities in the
activity_stream_totals table. The :total can be either a Fixnum
(positive or negative)or the symbol name of an instance variable
or method.
"""

Start page for the Rail code here http://www.matsonsystems.com/activity-streams/

Would love feedback. There have been minor upgrades since late 2008
release, but not a major overhaul to conform the spec yet.


Regards,
Jonathan

--

Jonathan Dugan, PhD
Matson Systems, Inc.
(650) 799-5369 @jmdugan
http://www.matsonsystems.com

> > impossibility) of the subject matter ? especially when applying it to two

> You received this message because you are subscribed to the Google Groups "Activity Streams" group.
> To post to this group, send email to activity...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/activity-streams?hl=.
>
>