User talk:Jon Awbrey
Contents 
Work Area
Notes & Queries
John, send me the jpg at drwhite@uci.edu and I'll see what the problem is. Doug
Bookmarks
 Thanks, Jon, for this bookmark Doug 10:43, 16 September 2010 (PDT)
 Glad you liked it. Sadly, the OpenStudy site, that I thought might have some potential, doesn't seem to be catching on, at least, not for anything beyond undergrad kibitzing. Jon Awbrey 13:14, 2 October 2010 (PDT)
 We're moving ahead with our Causalities project and a new R package, not much fanfare yet but lots of interest. You invested in 56 sites I recall, which was the best for you? Doug 07:28, 3 October 2010 (PDT)
 Glad to see you back at it. I noted that one of your sources, Logical Implication, Wikipedia has been replaced recently with a disputed page. Have you had a chance to evaluate one against the other? Doug 07:11, 15 May 2012 (PDT)
 Visiting Wikipedia on subjects I know anything about usually brings me to tears, so I don't go there anymore, except perhaps to look up the occasional fact about bands or songs from the 60s, where they do okay, unless of course there's a dispute about what musicians were also Scientologists. Jon Awbrey 07:42, 15 May 2012 (PDT)
Discussion
Jon,
A pleasure to have you join us. "Recent changes" has what is being worked on currently. Presently my son, Scott D. White and I are working on exploratory causal analysis for databases used in an educational and research context. Edit and add any pages you would like! Doug 08:11, 25 June 2010 (PDT)
Doug,
Thanks, it looks very interesting here. I did some thinking on causal analysis from a logical point of view in the late 80's early 90's, finding papers by H.A. Simon and a book by A.W. Burks very helpful. Did a Master's in Quant Psych writing a program (in Turbo Pascal !!!) for the "Exploratory Analysis of Sequential Observations" (EASO). C.U. Around, Jon Awbrey 08:26, 25 June 2010 (PDT)
Jon,
We are working with Judea Pearl's 2000 (2nd edition) Causality: Models, Reasoning. and Inference using the SCCS database and an Exploratory causal analysis for networks of ethnographically wellstudied populations, using Instrumental variables (IVs) and ChalakHalbert White Extended IVs (XIV). It goes beyond logic and probablistic reasoning, in an interesting Bayesian way, to a calculus of causality. Scott and I are working on streamlining and adding to the Anthon Eff and Dow (2009) software that I now used in my UCI courses.
You are working with sequences which adds more to causal inference and is treated in the XIV approach. Doug 13:11, 25 June 2010 (PDT)
Doug,
The problems in combinatorics, graph theory, and group theory that I was working on in the late 70's started exploding in their combinatorial way, leading me to think I would need to program a theorem prover as a helpmate. I soon decided that I needed to have it prove theorems the way mathematicians usually do, by remembering previous theorems and proofs, and this led me to work on sequential learning algorithms. I prototyped a very simple sequential learner based on classical ideas of Thorndike and Guthrie that was flexible enough to use for exploring the regularities in sequential observation data from a variety of ongoing social psych experiments at my site. And that pretty much used up the 80's. Jon Awbrey 20:02, 25 June 2010 (PDT)
Jon,
We're on the same page in graph theory, group theory. Glad to hear your motivation for theorem provers and sequential learning. I understand how that could absorb a decade. I have followed Peter Turchin in terms of social science dynamics, lots there in his three books and algorithms which I explored in several articles of my own. http://ccsl.mae.cornell.edu/sites/default/files/Science09_Schmidt.pdf is a physical lawextractor of great interest, found and posted in the "newsworthy events" on a Main Page 2009 subheading. I think the Judea Pearl]] approach supersedes most of the formal logics approach however. You might want to explore the (google: Statistical Entailment Analysis) or Galois lattice approaches to empirical logics derived from data. My colleague Carter Butts has programmed SEA in [R] but the program is in pieces and I haven't tried piecing it together for use in data analysis as yet. Do our interests overlap or not? Doug 18:07, 26 June 2010 (PDT)
Doug,
I kept up with the AI/CogSci literature pretty well through the 80's and 90's, with special reference to the work of Ken Forbus, Stephen Grossberg, John Holland (and the whole HoHoNiTh group), Ben Kuipers, and Judea Pearl that was relevant to Qualitative Dynamics. My quantitative psych studies at MSU and UIUC were mostly applied statistics and systems theory, so I did a few projects with DYNAMO (before STELLA's time).
 J.H. Holland, K.J. Holyoak, R.E. Nisbett, and P.R. Thagard (1986), Induction : Processes of Inference, Learning, and Discovery, MIT Press.
I'm still working on a way to summarize the main trends without bogging down in the full retrospective, but it's doing me some good to write this out. Jon Awbrey 13:16, 27 June 2010 (PDT)
Current Interests
Hierarchy of Systems
 Inquiry Driven Systems
 An inquiry driven system is a system that maintains a state of information about an object domain and that develops its state of information over time in a direction of increasingly useful information. An inquiry driven system is a special type of sign process, one whose time evolution is positively directed over the long haul in accord with a dimension of value or a measure of utility affecting its state of information. Informally speaking, it frequently makes sense to think of this utility as the "alacrity" or the "clarity" of the information.
 Sign Process
 A sign process is a sign relation plus a dimension of time and a mapping of times to signs. (This is very rough — it will take more work to make it precise.)
 Sign Relation
 A sign relation is a triadic relation among three domains, called the object domain , the sign domain , and the interpretant sign domain , that is subject to the following definition:
 A sign is something, , that brings something, , its interpretant sign determined or created by it, into the same sort of correspondence with something, , its object, as that in which itself stands to .
 Triadic Relation
 A triadic relation (or ternary relation) is a 4tuple consisting of three sets and a subset of their cartesian product. Naming the constituent objects in full formality, a 3adic relation is a 4tuple where , but the name of the subset is often used to denote the whole relation and causes no confusion so long as the context is constant throughout the discussion.
Related Readings


Discussion
 DW comment {@ Inquiry Driven Systems}. There is no way this can be monotonically increasing. See Pearl 2001. Say you have a working burglar alarm and are prepared to rush home from work when you get an arranged call from your security company. A call comes, but before you rush out a colleague says: did you feel the earthquake? It's a 5.8. You reassess. The causal network is iteratively changing. Pearl offers a critique of your approach, and a very new and better understanding of causality.
 JA comment. I was experimenting with the phrase "over the long haul", but worried about it — maybe "in the long run" would be better, but it's hardly ever monotonic in any case. This was just the skinniest sketch, so there is much that needs to be filled out here. A few points that come to mind right off:
 We are talking about control systems — cybernetic, goaloriented, intentional, or purposive systems, as various people call them — so the main thing is that we can usefully explain the behavior of the system by attributing a goal, objective, or purpose to it. That brings us to the threshold of complexity that separates 2adic relations and 3adic relations, and I've been mostly focused on the latter for a while. I'll look into Pearl again over the summer, as I think I went tangent to that whole literature stream somewhere in the mid 90's. (But I've got two family crises eventuating right now, so I may be out of this loop for a while.)
 The "object system" may be something initially described as nondescriptly as "the whole universe" or "the environment". What it means is that object systems are typically complex structures of object subsystems existing in various relations with each other, only the "immediate objects" of which occupy an agent's attention at any given time — and discovering that structure is one of the jobs of the inquiry driven system.
 That was the high end of the generality and vagueness spectrum, so maybe a few KICAS (keep it concrete and simple) examples would be a good antidote to airsickness at this point. Computations and proofs are familiar activities that fall under the above definition of a sign process. For example, we pass in time from the obscure sign "5 + 7" to the clear sign "12" for the same object, to wit, the number 12. Examples like that illustrate the fact that the measure of information in the successive signs need not increase, only the clarity of the information. Many questions and quibbles arise here, of course, as to how you quantify a thing like "clarity". Historically speaking, this is where "confusion matrices" and "receiver operating characteristics" came in.