Skip to Content.
Sympa Menu

pcdb - [pcdb] A PCDB can't serve as a predictive model of reality -- it only serves design process.

pcdb@lists.ibiblio.org

Subject: Permaculture Database

List archive

Chronological Thread  
  • From: Paul Cereghino <paul.cereghino@comcast.net>
  • To: pcdb <pcdb@lists.ibiblio.org>
  • Subject: [pcdb] A PCDB can't serve as a predictive model of reality -- it only serves design process.
  • Date: Fri, 30 Mar 2007 22:49:41 -0700

Jedd - I think you line of thinking around playing out scenarios is right on target. I don't know if this is where your going, but I want to steer us away...

I am not convinced that a PCDB can be designed, or should be designed to be a predictive model.

- regional variation in diseases and pests may cause members to drop in and out of guilds.
- interaction dynamics will change depending on what environmental conditions are causing stress or disturbance
- different installation and maintenance structure can affect the outcome of any design.
- subtlties of place will make the purported beneficial interactions of a guild meaningless.

after a lifetime of research you might be able to model the response of hackberry accross soils, climates, densities, stressors, disturbance regime, maintenance regime, browse pressure etc... just one species (I don't know anyone that grows hackberry around my neighborhood...). Now take four woody species and throw them together with 15 groundcovers, in 15 different climates, in 15 different soils, with 15 different stewards!! I am not saying you will have more then a thousand permutations, but you are going to have clusters of outcomes depending on context. Now we are talking about hundreds of species central to human diet, and a thousand of strong interest... many of which we have only a vague understanding of their autecology.

You cannot model when you don't have accurate parameters. Every parameter has error (for example your -10 to +10 scale). When you include increasing numbers of parameters each with error into a complex model.. (and we are talking about an astoundingly complex model!) the error propagates unexpectedly. The model fails to predict reality. We don't have the information to feed the model. We will not have that information... too much interaction between parameters, too much substantial variation. A database cannot discriminate.. the discriminatory capacity must be imbedded in the data. Our ability to provide the the data required to make a database discriminate is inadequate. It is the million monkeys on typewriters trying to write shakespeare. Or alternately we keep our relationships general, and the model can't discriminate at all. It misses obvious flaws and generates thousands of 'beneficial' combinations without a 'second thought', without a 'gut reaction'.

In my mind the goal is to create more effective human designers so they can create systems in place and report back... my goal would be to 1) support human agro-ecological designers by creating a self-regulating information system that creates beneficial information linkages between individuals (or books and individuals). The function of the system is not to design, it is to facilitate information flow between human designers. My sister worked in DB design in a massive social services DB in California. When they were developing the datastructure, they spent weeks traveling with social workers in cars from contact to contact, studying how they thought about their clients, how the classified information, the sequence of their thinking...

I doubt the function of the database will never be more than a repository of information that allows for dynamic queries. All the information will be variably accurate. I think you are right in trying to understand who will be approaching a data set, what questions they will ask, and what kind of information would benefit them, and identifying sources of error.

In fact one of the critical questions is how does the database gather information. As they say, shit in, shit out. As recently mentioned (I think D'Aoust in response to London) The nature and limitations of the 'observations' going into the system will totally constrain the ability to access useful information.

Paul Cereghino







Archive powered by MHonArc 2.6.24.

Top of Page