Owen, In fact, there is a component of TRANSIMS that is somewhat similar to what is described below. If you recall the "simulated annealing" step of a TRANSIMS analysis, a certain percent of unsatisfied traveler demand is iteratively rerouted until either all the travel demand for a given urban region can be realistically satisfied, or it is determined that the travel demand is unrealistic. However, the design of this component was carefully developed such that the individual agents are only in possession of local knowledge of the system, not global knowledge.
The analyst, or course, can see the overall effects on the system of the indiviuals' trip replanning, but the agents in the simulation only see their local effects. --Doug -- Doug Roberts, RTI International [EMAIL PROTECTED] [EMAIL PROTECTED] 505-455-7333 - Office 505-670-8195 - Cell On 1/21/07, Owen Densmore <[EMAIL PROTECTED]> wrote:
On Jan 20, 2007, at 5:58 PM, John Hellier wrote: > Is anyone working on Real Time Organizational Modeling where the > model continually evolves based on changes in the organization. All > members of the organization contribute to the changes even down to > the creation of an email, how the email contents affect the > organization and how the recipients respond to the email. Well, this sounds almost like TranSims in its completeness and depth! Doug might have a suggestion how to approach something quite this detailed and ambitious. Sounds like LOTS of fun too! One problem in this approach is that it is susceptive to the Butterfly effect .. extreme dependency on initial conditions. http://en.wikipedia.org/wiki/Butterfly_effect This is not a huge problem, but does mean that parameter scans, design of experiments, and the like are needed to make sure your predictions are stable enough for your purpose. Possibly computing a Lyapunov exponent would be a useful tool, but I confess to never doing so with my models, blush! http://en.wikipedia.org/wiki/Lyapunov_exponent > What I am looking for is the encoding of an organization such that > as someone creates an email, an observer can watch this happening > in the model and see the effect. Maybe the email has little or no > impact or maybe it has a growing ripple effect. I like the word "encoding" here. We've generally built behavior via algorithms, with a certain amount of stochasticity, but have not, in my mind, been quite formal enough. Carl: do you think policy modeling, and category theory in general, could handle encoding an organization? > This model should have a view of the entire organization including > tracking all actions performed. I realize that trying to capture > everything is a bit daunting but if possible it could yield > incredible insight into how organizations work. I'm curious: what is prompting this? Is it a possible project you may be working on? I ask because that might let you do *some* narrowing. > I generally feel that most decisions made in organizations are made > with such limited information that it is amazing that most > organizations don't fail. Or is that they are a lot less brittle > than one might imagine. No doubt about that! That said, one successful narrowing I know of is Steve's visualization of the pharmaceutical industry. Rather than look at the entire organization, the model looked at projects and their life cycle. Its a very interesting viz and maybe you could drop by the office for a show & tell. A second stunt Steve pulled off was actually a multi-organizational simulation of the entire British criminal justice system, including the police, courts and more. Not sure if this would apply in your case. > I know that there is quite a bit of work done in more bit size > pieces. I'm mainly interested in the much larger task of taking a > company of 40K and tracking every action and interaction. And then > by extension, actions connected outside of the organization. I > know, huge, maybe impossible. Is there a way to adapt social > networking concepts to an organization to help model it? > > Any ideas? I'd propose a WedTech meeting .. the lunch chats we have at Redfish on Wednesdays. They often are pretty unformed and brown baggy. It'd give you a way to talk through the modeling effort, and get good feedback from at least those that have tried such a thing. I'd sure love to think about this a bit more. For example, one approach might be to accept the bit sized pieces, but then have them interact. That would make the problem more approachable by decomposition. > Thanks > > John Hellier -- Owen Owen Densmore http://backspaces.net ============================================================ FRIAM Applied Complexity Group listserv Meets Fridays 9a-11:30 at cafe at St. John's College lectures, archives, unsubscribe, maps at http://www.friam.org
-- Doug Roberts, RTI International [EMAIL PROTECTED] [EMAIL PROTECTED] 505-455-7333 - Office 505-670-8195 - Cell
============================================================ FRIAM Applied Complexity Group listserv Meets Fridays 9a-11:30 at cafe at St. John's College lectures, archives, unsubscribe, maps at http://www.friam.org
