Re: [apsa_itp] Top-down vs. bottom-up

["David Bray" <David_Bray@xxxxxxxxxxxxx>]

Hi Peter, 

Sounds fascinating. Off-hand, I can recommend research by James March, Jay
Galbraith, and Thomas Malone -- they're all Organizational and Information
Systems researchers, but definitely applicable to the research you
reference below. They're also quite seminal in the field, so they provide
solid grounding. A lot of my research is focused on trying to link their
research to trends in government, particularly inter-organizational
knowledge sharing efforts. 

As for my own thoughts on the proposal, I can safely say that my empirical
modeling (extending March's original model) shows that for multi-tier
hierarchies, bottom-up approach is superior to a top-down approach if the
goal is to keep up with changing environments or a turbulent external
reality. That said, possible snags include: 

(1) there is a small (but statistically significant) negative impact of
adding breadth -- so as you expand the size of an organization, you will
experience a performance loss, in part because of delays incurred with
knowledge flows up (and then back down) the different parts of the
hierarchy

(2) how does the organization identify its experts (which may change/shift
with time, as the environment itself changes?) More often than not, the
experts are not the majority consensus of the organization -- but are
nevertheless "right" compared to a known external reality/truth. They also
may have been wrong in the past, but are now right. Thus, the trick in
practice is knowing who's right when we don't really know the absolute
truth for any given time? 

... hope this helps. Perhaps we could chat more one day? 

-d. 
http://userwww.service.emory.edu/~dbray/

"Peter Muhlberger" <pmuhl1848@xxxxxxxxx> on Sunday, August 19, 2007 at
1:50 PM -0500 wrote:
>Hi David:  You make some fascinating points about the value of
>different types of organizational structure and the role of knowledge
>elites.  I'm wondering how you would react to another quasi-utopian
>citizen engagement organizational concept, one that has some loose
>similarity to the idea of cascading delegates.  Some colleagues of
>mine and I will be testing this model out in an NSF project beginning,
>in earnest, next year.
>
>What I have in mind is something I call Multi-Layer Deliberation (MLD)
>and was based on models used, apparently with some success, in the
>Porto Alegre Brazil citizen participatory budgeting process and
>real-world experimentation with something called 'sociocracy.'  The
>idea for MLD is that you have citizens discuss some policy issue in
>small groups (say 10 participants) and then select a member or two to
>the next level group (perhaps by some kind of proportional
>representation).  Say the next level group contains 10 members, each
>one from a different lower-level group.   Levels can be added so that
>ultimately all groups are represented by one group at the top of the
>hierarchy.
>
>This model would seem to have some benefits:  Because of
>exponentiation, you could literally represent every adult on the
>planet in about eight group levels.  Good ideas would hopefully be
>filtered in and travel up the hierarchy and, if we have
>representatives go back to their original groups, back down to 1st
>level groups.  This would allow one person's good idea to potentially
>spread to an entire population without millions in advertising.
>Presumably people would elect people who were more capable than
>themselves, allowing good and capable people to filter up the
>hierarchy.  The top level group could potentially interact with, say,
>political elites as bonafide representatives of the public.  Also,
>upper level groups could organize and utilize lower level groups for
>various kinds of work--citizen action, information processing, etc.
>There is a problem of bottom up control--it may be that
>representatives promptly represent only themselves and not their own
>groups.  This might be aided by natural language processing
>technologies (which we are also testing) that would allow lower-level
>group members to efficiently determine how well their representatives
>are representing them.
>
>Any thoughts on the ins and outs of such a model?  Suggestions for
>literature for me to examine?
>
>Cheers,
>Peter
>

--
apsa_itp mailing list served by Harvard-MIT Data Center
List Address: apsa_itp@xxxxxxxxxxxxxxxxxxxxxx
Subscribe/Unsubscribe: http://lists.hmdc.harvard.edu/?info=apsa_itp