Lists Home |
Date Index |
Am 29.01.2006 um 00:57 schrieb Greg Hunt:
> Crowds have different effects on the task. For one thing, they
> average error, not just in the case of jelly beans, but also as a
> software project estimation technique (its a bit sad that a
> technical mailing list would have picked the jellybean example).
> In this case, a crowd is an expert.
See below please.
> Depending on the social structure, a crowd can enforce simple
> ideologies, the Wikipedia NPOV (and its associated assumption that
> neutrality is possible) is a good example of this, and in that
> context they also enforce filtering (see the discussion pages
> associated with the Wikipedia article on the Holocaust - there is a
> great deal of argument about who is an expert and who is not). The
> experts in that case are assumed to have looked at primary sources
> and in some sense have vasluable opinions. The participants in
> that discussion tend to be talking about limited ranges of
> secondary sources and they look to me more liked editors with
> opinions, than experts themselves. Perhaps this is another
> instance of the crowd averaging error? Certainly the discussion of
> Irving and the other revisionsists reads that way.
... really sad, and you are true with your opinion about the
participants, I think.
> In the case of technical topics the quality of the content is
> variable and dependent on the editors and editing process, the XML
> article is far more comprehensive than the Mainframe entry for
> example, which has a few eccentric statements in it (Speed and
> Performance), but there is not much debate about the technical
> facts, so the averaging process is easier. The mainframe article
> probably looks the way it does because the crowd is smaller.
> What I hope to get from an expert is fast coherence and quotable
> opinions. The question about the choice of a self-selected group
> or the individual expert seems to be whether you want your expert
> to be an editor/filter or a creator. Crowds seem not to create,
> but can filter very well within the limits of their social belief
> systems (the Wikipedia NPOV for example constrains what can be
> said). An individual expert has a clearer response time.
> An interesting question is, what happens as the crowd gets smaller
> - what is the threshold size for effective debate and filtering?
If you would like to look at my posting in answering Vladimir
Gapeyev, you will find something about bad experience with crowds I
had in my business, and about some causes of the problems (I've
learned a lot out of this crashed projects, which I won't discount).
Interestingly these have all been project teams with the task to
create something (organisation, data, and software).
May be filtering versus creating makes the difference.
About the software project estimation technique:
About 10 years ago, I read about several computer science projects.
They found a certain difficulty. Many software experts have had
similar or even the same lessons, read the same books, partly talked
to each other, used the same techniques, rules, patterns and
paradigmes, etc. Because of this the estimates tended to be biased in
the same direction in several case studies. These crowds didn't
average the error, but on average produced one in the same direction.
I think, this again is counter evidence against wisdom of the crowd
without a lot of precautions - even when filtering information. And
the causes are more social again.
I hope it's easier today, when we have better access to software
experts all over the world.
Well, I'm not thrilled by this wisdom-of-the-crowd thing. But I'm no
opponent. I hope it will work with the metadata dictionary wiki.