ORB Visualization
Notes on declarative and procedural formalism and routes to induction
A review of the publications of the Internet using “formal methods and routs to induction” and “declarative and procedural knowledge representation” will bring one to the stark reality that 98% of scholarship on knowledge representation is deeply influenced by a school of thought that claims”
“ Inquiry in order to be
inquiry in the complete sense has to satisfy certain demands that are capable
of formal statement.” [1]
(page 16, Logic: The
Theory of Inquiry, John Dewey 1938)
The literature on non-monotonic logics is perhaps where the limitations of formalism are most often acknowledged. But in most scholarly publications the authors simply never question the assumption that somehow, somewhere it is possible to define a formalism that can run on a computer and is able to conduct any inquiry.
How does one get away from the
authority of past scholars like John Dewey?
How does one point out that formalism has not made any type of
noticeable inroads on computational sense making? Where can one show computers that make sense of the world?
John McCarthy has focused a lot of his attention on developing extensions to logic to express inferences that are part of commonsense. This has led to important innovations like non-monotonic logic. Although I think this is an enormously valuable research development, I personally prefer to focus on these foundational domains of knowledge. Doug Lenat, with the CYC project, concentrates on accumulating the sheer bulk of knowledge that will make it possible to understand a story or explanation. [2]
(Benjamin Kuipers - from an interview with the IT Magazine Ubiquity 2004)
One answer is to set the deep philosophical claims aside and look at strong evidence implying the need for human-centric information processing, and to accept the astute observation that 98% of scholarship on knowledge representation seems to be misleading and philosophically unsound. The quality of the professional work and the difficulty of the scholarship is not questioned. What is questioned are new expenditures in CYC ontology and related AI projects. One should add up the investment in both time (decades) and expenditures and ask whether the scholarship on knowledge representation has yielded all of the value that can be obtained from the school of thought expressed by Dewey.
“There are many explanations for the failure of the
U. S. Central Intelligence Agency, the Federal Bureau of Investigations, and
their fellow intelligence agencies to “connect the dots” in time to stop the
terrorist attacks. The list of reasons
could start with the well-known inability of these organizations to
communicate. But their analysts’
out-of-date tool kit surely didn’t help.
Over the past decade, the business market has seen extraordinary
advances in data mining, information visualization, and many other tools for
“sensemaking”, a broad-brush term that covers all the ways people bring meaning
to the huge volumes of data that flood the modern world. [3]
(From MIT’s Technology
Review – March 2003 – M. Mitchell Waldrop)
The article by Waldrop focused on the term “sensemaking”, a term most closely associated with Karl Weick. But can anyone make sense of why the business market can make these extraordinary advances and yet nowhere do we see real sense making technology deployed?
Where is the objective reasoning for accepting the concept that the business market has something that the intelligence community does not have? Where is the objective reasoning that a group of scholars should be ignored.
Prueitt made a presentation on the issue of sensemaking at the 2001 e-Government Conference. He indicates that sense making should be far simpler that what the artificial intelligence community imagines. Prueitt’s work has demonstrated a new technology based on the separation of syntax from semantics and on the principles of Human-centric Information Production (HIP). He has made proposals, which received excellent peer review but no funding, designed to measure massive data streams and detect novelty in patterns.
What does Weick point out?
Karl Weick states his research interests as:
Effects of stress on thinking and imagination; consequences of indeterminacy in social systems; high reliability organizations; management of professionals; narrative rationality. (2003) [4]
The point that he makes is that instrumentation and mechanically produced reasoning may assist the human into making sense of something that is in fact false.
On the face of Weick’s notion of sense making there is a warning. The common sense humans have in everyday experience, can be mislead by the illusion that John Dewey would have us accept as an absolute truth.
The door opens into various literatures that have extensive scholarship but have received limited funding when contrasted to the funding spent on the knowledge representation literatures, and the many academic positions in AI that now exist in the universities.
In-Q-Tel has made an interesting start on something new by focusing on the concept of sense making, rather than the concept of knowledge representation. After all, the human needs only data regularity in context information to make sense of, not a dialog with an all-knowing computer system?
Our history shows a small community of scientists who have over and over again followed the rules and made the grant and contract submissions only to see our proposals deemed fundable but not funded.
All we ask for is the minimal funding to call a three-day workshop. We feel that at this time, such a workshop could gather together a group of 20 leading scholars who might properly propose how a new science of knowledge systems be created.
References:
[1] http://www.erzwiss.uni-hamburg.de/sonstiges/dewey/DewLog38.pdf
[2] http://www.acm.org/ubiquity/interviews/v4i45_kuipers.html
[3] http://www.stratify.com/infocenter/download/MIT.pdf
[4] http://www.umich.edu/~psycdept/Faculty/kweick.html
http://www.isi.edu/expect/projects/agents/rosetta.html