notational engineering and the search for new intellectual primitives
DESCRIPTION
TRANSCRIPT
Cover Page
Uploaded June 27, 2011
Notational Engineering and the Search for New Intellectual Primitives
Author: Jeffrey G. Long ([email protected])
Date: September 25, 2002
Forum: Talk presented at the Lawrence Livermore National Laboratory.
Contents
Pages 1‐2: Proposal and Bio
Pages 3‐31: Slides (but no text) for presentation
License
This work is licensed under the Creative Commons Attribution‐NonCommercial
3.0 Unported License. To view a copy of this license, visit
http://creativecommons.org/licenses/by‐nc/3.0/ or send a letter to Creative
Commons, 444 Castro Street, Suite 900, Mountain View, California, 94041, USA.
Title: The Notation is the Limitation: Notational Engineering and the Search for New Intellectual Primitives
Speaker: Jeffrey G. Long
Date: September 25, 2002
Estimated time: 60 minutes (45 for talk, 15 for Q&A)
The abstractions we use enable our perception, thought and communication, but they can also limit it. This talk will first present the thesis that in order to understand complex systems, and to adequately respond to many of the other challenges facing society today, we will need to develop wholly new abstractions ‐ new intellectual primitives ‐ with which to see and describe nature. It will argue that such an effort would be greatly accelerated, and made much more likely to succeed, by the creation of a proposed new discipline called "notational engineering," which will be described.
As an example of notational engineering, the talk will then present a theory of representation which is based on a new intellectual primitive called a "ruleform". The theory, called "Ultra‐Structure Theory," sees entities, structures and relationships as by‐products of complex processes, and postulates that any process can be represented by a finite but possibly large set of rules. It further hypothesizes that rules in any format can be converted into an If/Then format, and can be placed into a series of tables based on the particular "form" of the rules, i.e. how many "If" conditions there are, and how many "Then" statements there are. These place‐value tables are called "ruleforms", and they offer a practical and formal, yet highly abstract and concise way of organizing and representing myriad numbers of rules.
Lastly, as an example of a recent application of Ultra‐Structure, the talk will briefly discuss a project that was done for the Department of Energy to describe the rules of English and the rules of DOE classification guidance such that a computer could determine the classification level and category of a text document. The resulting knowledgebase consisted of tens of thousands of rules and was maintained directly by subject experts (in this case, certified document classifiers).
Further information:
Civilizations have traditionally developed notational systems by accident rather than systematically, so the hunt for new abstractions could be greatly facilitated by the systematic study of the history and evolution of a variety of types of notational system, e.g. the branches of mathematics, language and writing, musical notation, chemical notation, movement and dance notation, and money. In particular this search would be helped by a good general theory of the structure of notational revolutions such as occurred with Hindu‐Arabic numerals or the infinitesimal calculus. This proposed new subject of "notational engineering" would have as a primary goal the development and systematic testing of new abstractions in many areas, including (e.g.) new ways of representing value besides money, and new ways of representing complex systems besides the current tools of mathematics, computer science and natural language.
Ultra‐Structure Theory represents all knowledge of the world in tables of data rather than in the software of the system, so that the remaining software is "merely" an inference engine that has very little subject‐specific knowledge. This makes the knowledge (rules) easy to modify and liberates subject experts to directly manage the knowledge, rather than needing to communicate through a programmer to change program code.
Ultra‐Structure Theory constitutes a merger of expert system and relational database theories which minimizes the need for software maintenance and maximizes system flexibility. One prediction resulting from the theory is that all the members of each broad class of systems (e.g. all corporations, all games, all legal systems, all biological systems) differ from each other in terms of the specific rules governing their behavior, but not in the form of these rules. In other words, families of systems share the same "deep structure" or collection of ruleforms.
Biographical Information:
Mr. Long is a Systems Scientist. From 1995‐2002 he worked for DynCorp Systems and Solutions, a Washington consulting and services firm, on a contract for DOE. Prior to that he worked at The George Washington University as a Senior Research Scientist, first as director of the Notational Engineering Laboratory and then also as Deputy Director of the Declassification Productivity Research Center. He holds a BA degree in Psychology from the University of California at Berkeley.
The Notation is The Notation is the Limitationthe Limitationthe Limitationthe Limitation
Notational Engineering and theNotational Engineering and the Search for New Intellectual Primitives
Jeffrey G. LongSeptember 25, [email protected]
P d liProposed outline
1: Background on the general problem: representation and notational systems
2: Overview of Ultra-Structure: an approach to complex systems using a new abstraction
3: Example: The Reviewers Assistance System
September 25, 2002 Copyright 2002 Jeff Long 2
1: The Problem1: The Problem
September 25, 2002 Copyright 2002 Jeff Long 3
Many, if not most, of our current problems arise from
We may have pragmatic competence in using certain kinds
y, , pthe way we represent them
We may have pragmatic competence in using certain kinds of complex systems but we still don’t really understandthem theoretically– economics, finance, markets– medicine, physiology, biology, ecology
This is not because of the nature of the systems, but rather because our analytical tools – our notational systems and the abstractions they reify are inadequatethe abstractions they reify -- are inadequate
September 25, 2002 Copyright 2002 Jeff Long 4
Complexity is not a property of systems; rather,
Systems appear complex under certain conditions; when
perplexity is a property of the observer
Systems appear complex under certain conditions; when better understood they may still be “complicated” but they are tractable to explanation
Using the wrong, or too-limited, an analytical toolset creates these “complexity barriers”; they cannot be b h d i h i lbreached without a new notational system
These problems cannot be solved by working harder,These problems cannot be solved by working harder, using faster computers, or moving to OO techniques; they do not arise due to lack of effort or lack of factual information
September 25, 2002 Copyright 2002 Jeff Long 5
information
So far we have explored maybe 12 major abstraction spaces
September 25, 2002 Copyright 2002 Jeff Long 6
Notational systems facilitate perception, cognition and communication
Each primary notational system maps a different “abstraction space”– Abstraction spaces are incommensurableAbstraction spaces are incommensurable– Perceiving these is a uniquely human ability
Acquiring literacy in a notation is learning how to see a new abstraction space
Having acquired such literacy, we see the world differently and can think about it differentlydifferently and can think about it differently
September 25, 2002 Copyright 2002 Jeff Long 7
Notational Theory Offers a New Intellectual SynthesisNotational Theory Offers a New Intellectual Synthesis
Broadened to include all notational systems (not just l ) i h d li h d ilanguage), it sheds light on, and integrates:– Whorf’s notion of linguistic relativity, – Chomsky’s notion of an innate linguistic capabilityy g p y– Toynbee’s notion of the evolution of civilizations by challenge
and response– parts of numerous other theories in many areasparts of numerous other theories in many areas
September 25, 2002 Copyright 2002 Jeff Long 8
Conclusions From Section 1
Every set of intellectual primitives, reified in a
Conclusions From Section 1
y p ,notational system, has limitations: these appear to us in the form of a “complexity barrier”
Many of the problems we face now as a civilization are fundamentally representational or notationala e u da e ta y ep ese tat o a o otat o a
We need a more systematic way to develop and settle abstraction spaces: notational engineering
September 25, 2002 Copyright 2002 Jeff Long 9
2 O N A h2: One New Approach
September 25, 2002 Copyright 2002 Jeff Long 10
Current engineering methods work well only underCurrent engineering methods work well only under certain conditions
September 25, 2002 Copyright 2002 Jeff Long 11
This is the area addressed by Ultra-Structure Theory
Ultra-Structure Theory is a general theory of systems representation, developed/tested starting in 1985F i l i f l Focuses on optimal computer representation of complex, conditional and changing rules
Based on a new abstraction called ruleformsBased on a new abstraction called ruleforms
The breakthrough was to find the unchanging features of changing systems
September 25, 2002 Copyright 2002 Jeff Long 12
Unfortunately Complex and Changing Needs Exist inUnfortunately, Complex and Changing Needs Exist in Every Organization
Needs
SW & DB
Needs
time 1 time 2
SW & DB
time 3...time 1 time 2
September 25, 2002 Copyright 2002 Jeff Long 13
The theory is based upon a different way of describingThe theory is based upon a different way of describing complex systems and processes
observablebehaviors surface structure
generatesrules
f f l
middle structure
constrainsform of rules deep structure
September 25, 2002 Copyright 2002 Jeff Long 14
As Wolfram has recently argued, rules are a very y g , ypowerful way of describing things
Multi-notational: can include all other notational systems
Explicitly contingent Describe both behavior and mechanism H d d f th d b t d d Hundreds of thousands can be represented and
executed by a desktop computer
September 25, 2002 Copyright 2002 Jeff Long 15
Hypothesis: Any type of assertion can be
Natural language statements
reformulated into one or more If-Then rules
Musical scores Logical arguments Business processes Architectural drawings M th ti l t t t Mathematical statements
But often several “atomic” rules are needed to createBut often several atomic rules are needed to create one “molecular” rule, e.g. “3 strikes and you’re out”
September 25, 2002 Copyright 2002 Jeff Long 16
If/Then Rules are Best Represented as Data (records) O i d i t T bl i R l ti l D t bOrganized into Tables in a Relational Database
If A and B then consider C, D, E, F...
A B C D E F1
Rule #
234 1 Ruleform}5n
}
September 25, 2002 Copyright 2002 Jeff Long 17
Structured and Ultra-Structured data are semantically yquite different
Structured data separates algorithms and data, and is good for data processing and information retrieval tasks,e.g. reports, queries, data entry
Ultra-Structured data has only “rules”, formatted in a manner that allows a very small inference engine to reason with them using standard deductive logic
Th i f i (“ i i l ”) f The inference engine (“animation rules”) software has little or no knowledge of the external world
September 25, 2002 Copyright 2002 Jeff Long 18
The Ruleform HypothesisThe Ruleform Hypothesis
Complex system structures are created by not-il l d hnecessarily complex processes; and these
processes are created by the animation of operating rules. Operating rules can be grouped i ll b f l h f iinto a small number of classes whose form is prescribed by "ruleforms". While the operating rules of a system change over time, the ruleforms remain constant. A well-designed collection of ruleforms can anticipate all logically possible operating rules that might apply to the system, and constitutes the deep structure of the system.
September 25, 2002 Copyright 2002 Jeff Long 19
Th C RE H th iThe CoRE Hypothesis
We can create “Competency Rule Engines”, or C RE i ti f 50 l f th tCoREs, consisting of <50 ruleforms, that are sufficient to represent all rules found among systems sharing broad family resemblances, e.g. ll ti Th i d fi iti d t tall corporations. Their definitive deep structure
will be permanent, unchanging, and robust for all members of the family, whose differences in
if d b h i ill bmanifest structures and behaviors will be represented entirely as differences in operating rules. The animation procedures for each engine will be relatively simple compared to current applications, requiring less than 100,000 lines of code in a third generation language.
September 25, 2002 Copyright 2002 Jeff Long 20
The deep structure of a system specifies its ontologyThe deep structure of a system specifies its ontology
What is common among all systems of type X? What is the fundamental nature of type X systems? What are the primary processes and entities involved
in type X systems?in type X systems? What makes systems of type X different from
systems of type Y?
If we can answer these questions about a system,If we can answer these questions about a system, then we have achieved real understanding
September 25, 2002 Copyright 2002 Jeff Long 21
Conclusions From Section 2 One example of a new abstraction is ruleforms To One example of a new abstraction is ruleforms. To
truly understand complex systems such as biological systems, we must get beyond appearances (surface structure) and rules (middle structure) to the stable ruleforms (deep structure).
This is the goal of Ultra-Structure Theory.
September 25, 2002 Copyright 2002 Jeff Long 22
3: Application Example: the3: Application Example: the Reviewer’s Assistance System
September 25, 2002 Copyright 2002 Jeff Long 23
DOE Reviewer’s Assistance System Requirements
650 guides defining 65,000 topics that are or may be classifiedE i b k d k l d i d i Extensive background knowledge required to interpret guidance
Guidance changes over timeGuidance changes over time Terminology in documents changes over time The objective is advanced concept spotting, not document
understanding
September 25, 2002 Copyright 2002 Jeff Long 24
Normally This Would be Done Using an ExpertNormally This Would be Done Using an Expert System Shell
ES often have trouble with >1,000 rules; RAS has >100,000 rulesK i i h i i bili f l b Key issue is the maintainability of rules by experts
There are many benefits from using relational database to store rules as data, including:store rules as data, including:– Built-in referential integrity– Easy report-writing and queries
S bj t t i t i k l d b di tl ith t– Subject experts can maintain knowledgebase directly, without relying on KE or Programmers
September 25, 2002 Copyright 2002 Jeff Long 25
RAS D fi G id C t d All P iblRAS Defines Guidance Concepts and All PossibleLexical Expressions of Those Concepts
DefineDefineSystemSystemConvert GuidesConvert Guides
DefineInterpretations
DefineInterpretations
SystemReadySystemReady
Apply Guidance
Apply Guidance
ReadDocument
ReadDocument
DocumentReviewedDocumentReviewed
September 25, 2002 Copyright 2002 Jeff Long 26
Rules Specify Relations Between Topics, Concepts, and T kTokens
September 25, 2002 Copyright 2002 Jeff Long 27
C l i F S i 3Conclusions From Section 3
A rule-based system can provide precise and rigorous interpretation of key DOE terms and concepts
A rule-based system stored as tables in a relational database allows creation of a knowledgebase which candatabase allows creation of a knowledgebase which can become as large as necessary
Such a knowledgebase is very easy to specify, change and review directly by subject experts
September 25, 2002 Copyright 2002 Jeff Long 28
References
Long, J., and Denning, D., “Ultra-Structure: A design theory for complex systems and processes ” In Communications of the ACMcomplex systems and processes. In Communications of the ACM(January 1995)
Long, J., “A new notation for representing business and other rules.” In Long, J. (guest editor), Semiotica Special Issue on Notational Engineering, Volume 125-1/3 (1999)
Long, J., “How could the notation be the limitation?” In Long, J. (guest editor), Semiotica Special Issue on Notational Engineering, Volume 125 1/3 (1999)Volume 125-1/3 (1999)
Long, J., "Automated Identification of Sensitive Information in Documents Using Ultra-Structure". In Proceedings of the 20th Annual ASEM Conference, American Society for Engineering ManagementASEM Conference, American Society for Engineering Management (October 1999)
September 25, 2002 Copyright 2002 Jeff Long 29