Systems Design and HumanML (RE: RE: markup race)

["Bullard, Claude L (Len)" <clbullar@ingr.com>]

Copied to the list to document some points..

-----Original Message-----
From: arthide@magicvillage.de [mailto:arthide@magicvillage.de]


>A label applied for simplistic identification using 
>surface characteristics is useful in systems, but 
>becomes problematic given the type of interpretation 
>applied.

Yes.  It is important in any case to use the datatype 
in an appropriate process, similar to the way business 
rules work for a requested service.  It is true of many 
kinds of "facts" one might gather.  For example, 
geography is a limiting influence on human behavior. 
It permits an activity but does not dictate it except 
for example, where say survival is involved.  But 
in cases such as access to metals or woods, the resources 
are more important than the geography.  The geography 
may permit the resources and therefore, influence the 
local socio-culture.  The cultural influence is a much 
stronger influence on behavioral choices.  The effects 
of urban anonymity (due to spatial mobility and population 
density) include a weakening of the cultural 
influence and this enables more diverse behaviors, both 
good and bad.  So this is a complex relationship and 
most facts in isolation are not compelling, one might 
say, using the expert system approach, they have weak 
certainty values and therefore, weak correlation.  Semantic 
incoherence affects almost all cross-domain analysis.

One discovers in the field of criminology that race 
features have been held in low regard for most of this 
century.  Again, they are used primarily not as an 
"identifier" but as an "observable" in the process 
of identification.   I think it important to make 
both the distinction of type (race is not an 
"identifier", it is a name for a set of observable 
characteristics), and to point this out as a property 
being used in a process.  That makes it possible to 
then declare controls over process (eg, a policy 
for interpretation or actions).

>The most primitive i imagined was a tag
>with the values "emic" and "etic", but you got it 
>precisely in the wider sense for the design of the language, 
>which i really meant. This way results can be appropriate; 
>no matter, on which level, whether simple or complex.

The simplest models for this are the ones from 
systems theory, that of nested or recursive 
process:  input, output, control and feedback. 
View dimensions can be modeled thus because this 
model fits the requirements for chaotic systems 
nicely.  

>Before i found humanML i thought privately
>a bit about The Parenthesis in html
>(and relatives) and comparable constructions
>in spoken languages, and which sense it can 
>have for creation and communication in images. 

>For that reason, I cite the analogical 
>reasoning where a method from a member of the same 
>class is applied to another member to test if a 
>predetermined outcome results. 

Consider for example, the HumanML concept 
of using genre-languages.  These are simply high-level 
authoring languages designed to enable the author to 
work with names and concepts with which they are 
familiar, but with some structures annotated as 
HumanML types to enable a more human-oriented 
interpretation or set of rules to be applied.  A 
HLAL, is in effect, a point of view, an observer 
type that enables easier authoring, but which has 
the HumanML capabilities for interpretation.  The 
HumanML abstractions provide class members to test.

>Has some double-meaning. The contradictions 
>betwen the selfdefinitions of the
>individuals, the definition of individuals by 
>legal and illegal authorities, and for example
>scientific or religious systems of definition 
>(or classifications or descriptions) is obvious, 
>a daily experience of millions of people. 

>It seems to me, codes which can work in such settings, 
>need something like a "Straight universal Kernel / Grammar"
>(please don?t nail me on the professional definitions), 
>as well as a "Flexible Configuration / Interface", easy to use in 
>endless variations.

Yes, but I doubt the universality.  The best we get is 
yet another viewpoint, in this case, HumanML.

>It is one of the great cultural inventions (with all its problems) 
>to tolerate these contradicitons in definitions, and 
>even, for the protection of the individual and the society, 
>in certain situations, not always to allow to mix those definitions
>and descriptions. For example before court, 
>when known human facts are not allowed to be used for a decision,
>because the state had no right, to collect them.

Policy.  It is a control in a protocol or formal communication.  
Semiotics introduces the notions of codes and we find this 
implemented in almost any database system as some form of 
select.  Each state, for example, has its legal codes and 
these vary widely making statistical collections of crimes 
or other behaviors a very difficult job.  Federal codes 
exist (eg, NIBRS) and these are collected in large volume, 
but they are policy-controlled (for example, even though 
they collect a lot of behavioral data, they don't collect 
names and therefore, the system is worthless for investigation). 
On the other hand, names are collected by the NCIC.  All of the 
data has the same sources (your local public safety agencies) 
but is going down different pipes to different processes with 
different controls.   Early in HumanML, we discussed the 
ROUTE structures of VRML as a means to model an event/gesture/message 
being received by a person and then routed through different 
processes with each process in some way modifying the message 
before passing it on to the next process.   If we considered 
as having both external processes (environment (everything not self))
and internal processes, then the routing of the messages is 
a simple model.   

It is easy to take a systems approach but we have to consider 
the message(s), the choices made at each node, the authoritative 
process that declares which choices a subnode has, and the route.

len