All posts in the topic Participation Inequality (Short link)
Summary
- There are 17 posts — by 5 authors — in this topic.
- Latest post made by Michael JasonSmith at 2007 Aug 14 05:53 UTC
| From | File | Date |
|---|---|---|
| John S Veitch | jsveitch.vcf | 2007 Aug 12 09:37 UTC |
A few thoughts: 1. The more GS helps us measure participation and see it in summary views the easier it will be to test new ideas. 2. I think web views on topics and posts over say the last week would be a useful way to allow people to navigate some posts/topics. 3. A "recommend" or four/five star rating option next to posts would be a nice option even if it is primarily done via the web and a recommed click through in the e-mail footer. (I wonder if we could automatically Digg and Redit posts recommended from out site? 4. I guesstimate that E-Democracy's use of the two post in 24 hours tool expands our 1 percent of most active to 3 percent and our 9 percent up to 20 in some cases. It would be interesting to document our real numbers, compare forums, and see if we can broaden participation through facilitation/tools toward specific improvement goals. See: http://www.useit.com/alertbox/participation_inequality.html End bit ... Your only real choice here is in how you shape the inequality curve's angle Are you going to have the "usual" 90-9-1 distribution, or the more radical 99-1-0.1 distribution common in some social websites? Can you achieve a more equitable distribution of, say, 80-16-4? (That is, only 80% lurkers, with 16% contributing some and 4% contributing the most.) Although participation will always be somewhat unequal, there are ways to better equalize it, including: Make it easier to contribute. The lower the overhead, the more people will jump through the hoop. For example, Netflix lets users rate movies by clicking a star rating, which is much easier than writing a natural-language review. Make participation a side effect. Even better, let users participate with zero effort by making their contributions a side effect of something else they're doing. For example, Amazon's "people who bought this book, bought these other books" recommendations are a side effect of people buying books. You don't have to do anything special to have your book preferences entered into the system. Will Hill coined the term read wear for this type of effect: the simple activity of reading (or using) something will "wear" it down and thus leave its marks -- just like a cookbook will automatically fall open to the recipe you prepare the most. Edit, don't create. Let users build their contributions by modifying existing templates rather than creating complete entities from scratch. Editing a template is more enticing and has a gentler learning curve than facing the horror of a blank page. In avatar-based systems like Second Life, for example, most users modify standard-issue avatars rather than create their own. Reward -- but don't over-reward -- participants. Rewarding people for contributing will help motivate users who have lives outside the Internet, and thus will broaden your participant base. Although money is always good, you can also give contributors preferential treatment (such as discounts or advance notice of new stuff), or even just put gold stars on their profiles. But don't give too much to the most active participants, or you'll simply encourage them to dominate the system even more. Promote quality contributors. If you display all contributions equally, then people who post only when they have something important to say will be drowned out by the torrent of material from the hyperactive 1%. Instead, give extra prominence to good contributions and to contributions from people who've proven their value, as indicated by their reputation ranking.Your website's design undoubtedly influences participation inequality for better or worse. Being aware of the problem is the first step to alleviating it, and finding ways to broaden participation will become even more important as the Web's social networking services continue to grow.
On Wed, 2007-08-01 at 17:58 -0500, Steven Clift wrote:
> 3. A "recommend" or four/five star rating option next to posts would be a
> nice option even if it is primarily done via the web and a recommed click
> through in the e-mail footer. (I wonder if we could automatically Digg and
> Redit posts recommended from out site?
Recommending I think is a good idea. I'm not yet convinced on the star
rating (despite being a fan of the idea earlier) -- my reasoning is
this:
It works well at, eg. Netflix or Amazon. This is because you get visits
to the same subject (a video, or book) over, and over again, with
millions of users.
The same is *not* true for a post, typically. The occasional post might
hit the big time news, but typically it is seen for a day or two, and
fades from popularity.
Google implements star ratings on posts. I have only *very* occasionally
seen them used -- and they have millions of users.
I'm not necessarily saying it's a bad idea, but I think perhaps it would
best be integrated as part of the 'quality contributors' idea that you
discuss. A simple thumbs up/thumbs down, which contributes to the user's
overall 'karma' would be one way of doing this, I think. Then highlight
entries in the topic/post according to their 'quality'.
I agree that participation information is important, Steve. One of our [my?] goals for GroupServer is to have the best participation statistics of any message-based system: including keywords, and post statistics. We have plans for adding more participation information to groups and profiles. Recently, we added the ability to view the participation statistics in a group http://groupserver.org/groups/development/stats.html but there is still some way to go. In particular, I would dearly love to plot the number of members in the group over time. In addition to the simple membership statistics, Dan and I are quite keen to add “sociograms”, which would visualise who interacts with whom. If someone asks me about sociograms in a new topic I will explain them more fully, and tell you how I was inspired by the Wattle Tree visualisation of Trac that Judy Kay help develop http://www.it.usyd.edu.au/research/tr/tr582.pdf I also agree with your second point, Steve: the user should be able to see what has happened in the last week, which is possible now. Using email, the messages that the user has or has not read are marked read. With the Web, topics that have new posts are marked "unvisited", as are posts that the user has not seen. Dates are displayed against all topics and posts. Various Web Feed readers also indicate if a post is read or unread. Is there something else that we do not do, Steve? I am against grouping anything by a "week" because it gets into all sorts of religious and cultural arguments about what a week *is*. Someone else can start that fight… I could indicate when Kalends, Nones, and Ides occur, but that would be too obscure for most people ☺ Besides, not all groups would benefit from the weekly grouping: http://groupserver.org/groups/development/messages/topics.html Addressing your third point, any modification to a post (such as adding tags or rating) would have to wait until we have a mechanism for tracking all changes to the posts, and have a mechanism to back out of those changes. This *major* item of infrastructure has been dubbed the “audit trail” project; some other items also rely in it being done, such as allowing managers to edit aspects of a user's profile. (Once again, open an “Audit Trail” topic if you want to know more ☺) The technical challenge of rating posts aside, we also have the problem of the usefulness of such a mechanism. As Richard says, the similar feature in Google has very little use. From my own experience, if users cannot be bothered using topics correctly, then I cannot expect them to take time out to add extra metadata, which is only available in one of or three interfaces! Adding tags to posts is a different story, as I can see the participation coach (alias forum manager) using them to highlight important posts, such as those written by a guest speaker in a public issues forum. Finally, you are right that there are many unanswered research questions relation to participation in online groups. I am especially interested in how invitation-only forums differ from more open groups. However, research is very expensive, and fraught with difficulties. For example, controlling external factors (promotion of the forum, external events that may make the forum popular, the weather, the time of year, the changing membership of the forum…) make experimentation a nightmare. I actually have a small research project on this month. I will keep this group informed with how it goes.
To Richard's comments ... one reason I like "Recommend" or positive votes is
that with agree/disagree I am concerned with the political majority pushing out
the minority in our forums ... however, for most sites this would be a nifty
feature.
On Thu, 2007-08-09 at 05:27 +1200, Steven Clift wrote:
> To Richard's comments ... one reason I like "Recommend" or positive
> votes is that with agree/disagree I am concerned with the political
> majority pushing out the minority in our forums ... however, for most
> sites this would be a nifty feature.
I actually don't think it is too much of a concern unless the political
minority is *very* small. It would be easy enough to come up with an
implementation that was loaded against 'ballot stuffing' for example.
One suggestion off the top of my head would be something like using the
natural log of the number of 'thumbs up' or 'thumbs down' votes (so
1000x more votes would not result in anything close to 1000x the
rating), and then use other vectors to weight the 'goodness' of the
post.
For example, 'guest speakers' (as used by Dan in the Canterbury Issues
Forum) might get a bonus points, newcomers might get a bonus points,
people who accumulated many overall thumbs down ratings might lose
points.
Another more sophisticated way would be to track all votes against all
individuals, and down-bias votes from people who regularly vote always
down or always up, and severely down-bias votes from people who always
vote down or up particular people (and up-bias votes from people who
allocate a vote against the trend for a particular individual).
Unfortunately, Steve, there is no simple way to make democracy less democratic; by definition the majority view always pushes out the minority. With Richard's logarithms system, the majority would still always push out the majority, but it would look close :) With ranking a post, it is easy to game by always giving out the top or bottom rank, or give out the highest and lowest values as much as possible. There are complex ways, to limit the power of democracy, but they are complex. For example, the election for the Doge of Venice, which I was reading about this week http://www.hpl.hp.com/techreports/2007/HPL-2007-28R1.html linked from Schneier on Security http://www.schneier.com/blog/archives/2007/07/security_analys_1.html I would rather allow the UI to highlight the user's social role (guest speaker, participation coach) than have a voting system. Readers can then give weight to the user's post accordingly. Posters can already state that the agree or disagree with a poster, and do so quite readily (here and in the Canterbury Issues Forum, at least).
On Thu, 2007-08-09 at 05:27 +1200, Steven Clift wrote:
> To Richard's comments ... one reason I like "Recommend" or positive votes
> is that with agree/disagree I am concerned with the political majority
pushing
> out the minority in our forums ... however, for most sites this would be a
nifty
> feature.
BTW, I just realised I don't think I said anything about agreeing or
disagreeing with the post. My suggestion was that you would indicate
whether you thought a post had interest value or not, which isn't
anything like agreeing or disagreeing.
To Mike's comments: to my mind this entire idea is about adding
relevance to the system -- at any given time there might be 600 posts
that could be considered to be 'new'. I have 5 minutes. Which posts
should I read? It has nothing to do with whether people agree with a
particular poster or not.
Things that don't help me: topics index (because it just tells me what
is new, not what is 'important'), posts index (even worse than a topic
index, it makes me skim through everything).
This is why things like Digg and Slashdot excel at satisfying the desire
to 'fill a few minutes with something interesting' -- things of interest
float to the top. I don't believe the same can be said of GS ... yet.
I would suggest that the 'interest' level of a particular post is almost
*always* tied to the poster -- whether it is because they have
expertise, write well or happen to be the person of the moment.
So yes -- I agree that things like the social status of the user (eg.
guest speaker) are strongly tied to this. Here's a basic proposal:
1. each post has a 'interest rating',
2. the 'interest rating' consists of various weighted values.
For example:
25% of the interest rating could be made up from the 'thumbs up/thumbs
down' rating.
25% from the number of posts made by the user who made the post in
topics *you* have participated in -- "Do I like to debate with this
individual"?
25% from a rating of the number of thumbs up/thumbs down *you* have
given the user who made the post
25% from the 'karma' value of the user who made the post.
The idea is that you get posts *you* are more likely to be interested in
highlighted, so that you're encouraged to participate (you haven't spent
all your time wading through finding something of interest, so you'll
have more time to reply).
On Thu, 2007-08-09 at 05:27 +1200, Steven Clift wrote:
> To Richard's comments ... one reason I like "Recommend" or positive votes
> is that with agree/disagree I am concerned with the political majority
pushing
> out the minority in our forums ... however, for most sites this would be a
nifty
> feature.
I'm against a "voting system" for posts. The only one I know of that
almost works is at slashdot. There is a deliberately developed system
there, and volunteers can be invited to participate as moderators. I
don't read slashdot now, but long ago I was an active participant.
Even with their system, it didn't work very well. Any post that could be
"read" in 20 seconds tended to get a good mark if it was funny or
interesting. Early posts tended to get good marks. Anyone who wrote
something considered and thoughtful can't get a look in. In a geeks
forum any letter longer than six lines is an essay, and short essays are
preferred.
I have another way to find the best stuff. Del.icio.us is useful, but
even more, I depend on my friends finding stuff for me and giving me a
heads up. I use Ryze a lot for exactly that reason. Most of what I need
to know has a way of getting to me.
The following file was added to this topic:
> …at any given time there might be 600 posts that could be considered > to be 'new'. I have 5 minutes. Which posts should I read? It has > nothing to do with whether people agree with a particular poster or > not. Excellent, a task ☺ Unfortunately, I cannot see how a voting system, or the voting component of a rating system, will ever do anything other than tell you what people agree with. However, adding a blog facility to GroupServer would allow someone to write a commentary on what has happened in the group, similar to the Linux Weekly News Kernel Development page http://lwn.net/Articles/243052/ Now it only comes down to whether you agree with the blogger, and whether you can find someone to blog for the group! Blogging has other benefits, so I would prefer to focus development effort there, rather than a system for ranking posts. I cannot think of a less equatable interface than a blog, as it only lets one (or a few) people post. However, if the problem of too many posts is caused by participation equality then maybe an inequitable system is the solution ☺
If we adopt a rating system, I think it should be based on individual ratings,
with some ability to aggregate them, and some kind of default aggregation
(perhaps none at all) for users who are anonymous or have not configured their
preferences.
To me "what to read" when in a hurry is based on my interests, not everyone
else's. Of course, I can't be bothered describing my interests in much depth,
so I'd like to piggyback on the effort of people who I know share my interests
in certain areas (my "taste buddies").
The relevance index has to determined before I read a post, so rating posts is
going to be pretty ineffective. I might rate the odd post for the benefit of my
taste buddies, but I'd like to be able to a whole topic too, as there are fewer
of them and they are of higher value.
I would like to be able to rate groups, people and occasionally posts with my
own rating for "relevance". Then I want to be able to create a feed where I
prioritise posts from groups I've rated and by people I've rated, posts rated
by people I rate, and posts by people rated highly by people I rate highly.
Perhaps, I'd throw in a few keywords to look out for. I'd like to be able to do
this in a "channel" for each interest. Some of my channels would be shared with
others. I'd use channels shared by people I rate to decide the naming of some
of mine, so the sharing of channels would limit the proliferation of new ones.
Maybe I'd add some of my taste-buddies' channels as inputs into mine.
Obviously, negative ratings would have the opposite effects.
The system would also track my reading patterns while I'm logged in, and learn
my preferences based on what I click on and how long I dwell.
On Mon, 2007-08-13 at 10:01 +1200, Michael JasonSmith wrote:
> I cannot think of a less equatable interface than a blog, as it only
> lets one (or a few) people post. However, if the problem of too many
> posts is caused by participation equality then maybe an inequitable
> system is the solution ☺
Well to address what I think were your two points:
1) A voting system would be skewed towards people voting for posts they
agree with,
2) A blogging system would be better because it would allow highlighting
of the good posts
Rating system:
A voting system may well be skewed towards the posts that people agree
or disagree with. The *volume* of votes will always be higher on the
most controversial posts though (which in itself is a way of
highlighting worthy posts). That could in fact be the way of doing it, a
single 'recommend' button (as I think Steve said).
The rating system proposed isn't a voting system. It has a voting
component. The idea is to try and provide some idea of what might be
interesting; to help speed the cream floating to the top.
Blogging system:
Well firstly -- blogging systems depend on someone wanting to blog. It
seems *incredibly* unlikely to me that someone will be willing to blog
for even the majority of groups, let alone all of them.
Now, having said that, blogging systems certainly have their place. For
a group that has a blogger, that would *almost* always be better than an
autonomous system. Unless the blogger goes on holiday. Or gets sick. Or
bored. Or sucks. Or is just plain biased.
It also has the big disadvantage that I have to *read* it through in
order to glean anything useful from it. I don't want to read much -- the
more time I spend reading the meta information, the less time I spend
reading the information.
This is a little like the Topic vs Post feed conversation that never
quite made it into this group (unfortunately). You prefer the posts
feed, because you read it like a newspaper. I prefer the topics feed
because I read it like a table of contents (and read the things that
look interesting in the correct place, in my email or on the web ;)).
You like syndicated content, I like syndicated summaries.
Dan:
You think rating posts is likely to be ineffective because what you want read
is likely to be based on what you like to read, not what others like to read. I
propose that what you like to read is often (but not always) the same as what a
number of others like to read. I think a voting system is likely to be
ineffective if people *perceive* it as ineffective, which is why I don't think
a voting system should standalone, but rather needs to be a component of an
overall ranking system.
I go further and propose that what you like to read, is actually determined
partially by who you like to converse with. If there are 5 people posting in a
topic, and 3 of them are people that you have held a number of conversations
with in the past, then you are more likely to want to be involved in that
conversation -- this is individual relevance. This may contain people that are
counter to your personal worldview -- this is a *good* thing, I'd suggest that
the basis of a stimulating conversation is counter viewpoint, not similar
viewpoints.
I like the idea of being able to setup keywords you're interested in. That
could quite easily be just another vector in the system. I really like the idea
of automatic rating based on 'dwell' factors -- how many people have dwelled on
reading this post. You'd probably have to offset dwell by number of words ...
The 'next big thing' is relevance. There is simply too much information to
digest. We could blog it, but then we'd still need a relevance engine to help
us figure out which blogs to read (which is why I don't read particular blogs,
I tend to search for information in blogs) :)
Richard,
I entirely agree that the next big thing is relevance, and am therefore
glad that we are having this conversation.
> You think rating posts is likely to be ineffective because what you
> want read is likely to be based on what you like to read, not what
> others like to read.
No, I said that basing it on what _everyone_ thinks is ineffective.
Using _others'_ ratings is where the opportunity is, partly because it's
easier to rate people than posts, and partly because the rating is
already on the post when it arrives.
Basing the ratings of people who post to the same topics as me is a
great idea. We'll have our first interface to that soon, judging by the
progress Michael is making.
Actually, I think that whether something is relevant and interesting is
correlated with whether I agree or not - but negatively. By definition,
stuff I already know can't be interesting or relevant.
On Mon, 2007-08-13 at 11:45 +1200, Dan Randow wrote:
> I entirely agree that the next big thing is relevance, and am therefore
> glad that we are having this conversation.
>
> > You think rating posts is likely to be ineffective because what you
> > want read is likely to be based on what you like to read, not what
> > others like to read.
>
> No, I said that basing it on what _everyone_ thinks is ineffective.
> Using _others'_ ratings is where the opportunity is, partly because it's
> easier to rate people than posts, and partly because the rating is
> already on the post when it arrives.
I absolutely agree, which is why I said very early on that I thought the
'votes' should attach to the user, not so much the post. They *do* need
to attach somewhat to the post as well, IMO, otherwise a new user with a
brilliant post would never be highlighted.
Perhaps we actually deliberately bias the system towards new users, and
as they post more they (exponentially) get less bias.
> Basing the ratings of people who post to the same topics as me is a
> great idea. We'll have our first interface to that soon, judging by the
progress Michael is making.
Yes, we shall. It will still be a fairly one-dimensional view though,
much as keywords are currently -- I can't see how it would expose
brilliance as a standalone system.
I'm not convinced that there is much *singular* benefit in rating people
directly, primarily because the net effect will just be a trust network.
Trust networks have been done, and done again. Trust networks are good,
but as you say below -- the things that are often *most* relevant (or at
least interesting) are counter-view (which is what I said in my previous
post). I don't think I need to trust someone, or even trust their POV,
for what they say to be relevant (or interesting ... they aren't
necessarily the same thing).
> Actually, I think that whether something is relevant and interesting is
> correlated with whether I agree or not - but negatively. By definition,
> stuff I already know can't be interesting or relevant.
Yes, I agree with that.
I would actually prefer that *nothing* was manually rated. The main
reason is that *maybe* 5-10% of people will use the 'voting' system, or
whatever. And worse, those people will almost always be the same 5-10%.
Relevance is both a collective and an individual thing. The 'collective'
sorts the wheat from the chaff. Whether the collective agrees or
disagrees with a particular post, people generally won't click
'recommend' on a post that says 'me too'. They may not click 'recommend'
on a post that they disagree with either (but that doesn't matter,
because someone may, and the system would be logarithmically biased).
Even 'special' participants (like a guest speaker) might reply with say,
'thank you' or something. Which is hardly worthy of recommendation.
The keywords involved in a topic are somewhere between collective and
individual. If you setup a list of keywords you're interested in, then
they could contribute to providing the basic filtering at this stage.
Once you get beyond this basic sorting, the individual relevance kicks
in. This is where the real magic would be -- is this post/topic likely
to be interesting because I regularly converse with the participants? Do
the participants have a good history of 'recommended' posts? Is the
participant a special participant (guest speaker for instance)?
On the subject of rating topics ... I think this is more a case of the
whole being the sum of the parts. I can see *some* benefit to rating
topics, but I just think it would be clunky and not particularly well
used.
I'd like to focus on the Useit article's advice to essential turn site use into a form of participation. The user task - help less frequent web visitor quickly find out what is hot this week/month. Simple Ideas 1. Right Now - Topic list ordered by most recent post. Innovation complete. :-) 2. Posts - Make "Posts" clickable and sort last 30 days of posts by number of replies, add option to expand time frame to one year/all time. Next Level 3. Most Viewed Topics - Convert web stats into "democratized navigation" 4. Most Viewed Posts - Topic views for seven? days from posting plus individual post views 5. Most E-mailed Posts/Topics - We don't have this option, but I find the results at the StarTribune quite interesting: http://startribune.com - they are often not the same, e-mailed posts connect more passionately with readers Ultimate 6. Most Recommended - I vote for a simple "recommend" button next to individual posts and a click-trough in the footer (something else may need to be condensed) and simply let crap/things people disagree with not rise up. May be it is a "Good" or "Like it!" button to make it snappy. If folks end up not using the feature, they I would recommend turning it off. 7. Popular Topics - Here is my rough formula - assign points based on: - Webviews (weighted by %/#registered web members so we can have cross-site popular topics too) - Total posts to topic (volume = interest) - Total number of authors (diversity = good, usual suspects bad) - Posts within topic recommended - Hours of sustained posting Bonus points: - First time participant post - Less frequent participant post Deductions: - One sentence posts - Two authors make up more than 50 percent of posts I would default popular to the last two days and give an option for the most popular this week, month, this year, specific years, and all-time. Cheers, Steve P.S. Say, check this out: http://news.bbc.co.uk/2/shared/bsp/hi/live_stats/html/map.stm
Richard noted:
I like the idea of being able to setup keywords you're interested in.
My response:
Moving from search results to web feeds for search (which is built it in and I
assume displays based on time and not relevancy) to saving your search words
for a menu of self-generated topics would be useful.
What would be really useful is setting group or site specific "Search Alert",
daily e-mail digest with links to posts that contain your word(s) that day.
Imagine that I am the Mayor of X city. I'm on digest mode because I can't
handle the e-mail, but I want to know when I am mentioned. I set an alert on my
name.
I would guess that almost all elected officials in the communities where we
have Issues Forums would set alerts on their names (and other terms) if we make
it really easy to turn it on.
Thanks for your posts, Richard and Steve! I really appreciate the level of thought that has gone into this topic! Richard nicely summarised both my points http://groupserver.org/r/post/6gPpYFMezpW6vxzIOmVXTo 1. I fear that any recommendation system will become an agreement system. 2. I think blogs will overcome this situation. There were three methods proposed to overcome my apprehension about recommendation systems: 1. Limiting choice, 2. Rating individuals, and 3. Downplaying the users' voting effort. I favour 3 the most, as it comes closest to my ideal of dropping the idea ☺ The first two options still have the problem of votes being the same as agreement rather than interest; even in the simplest option the vote is either agree or disagree (ignore). The problem with getting someone to blog is a major one, but it is one that (should) fix itself. The groups that will have the problem of too many posts are large ones, where there is a lot of activity. (We are all assuming that activity is correlated to interest). As the group is large, there is a large number of people who could potentially write the blog. Because there is a lot of interest in the group, the writer will get a heap of attention, so we can rely on Adam Smith to pull us out of the mess. (I do not have the right Smith quote to hand.) Trust networks were suggested by Dan, and criticised by Richard http://groupserver.org/r/post/74D0D5PjZfQ9mIyC5vZpn3 I would add another nail into the coffin: most users do not post. Steve suggested counting the number of replies to a post. http://groupserver.org/r/post/4U1xGnhiZ7AO96G5XNnenI There are huge problems with this: * People do not understand threading, * People do not understand threading, and * People do not understand threading. For this reason, GroupServer does not have a threading view of a topic, and does not use threads to determine if posts belong to a topic! The rest of Steve's suggestions *mostly* focus on one thing: giving the user a better idea of what is going on. That is a *M*A*J*O*R* goal of mine ☺ The recent changes to search and the topic lists http://groupserver.org/r/topic/6CDpVzg8RLNt9nFFVdRRfu were made so this was possible. By providing information, we feed the best pattern-recognition and summation engine we have available to us: the user. There is plenty of information that we could provide but either do not collect, or do not display. In the former category is view-information. While this would only be applicable to Web views, it would still be worth displaying. In the second category is the number of unique authors in a topic: it is computationally expensive to calculate this for every topic-list.