Sunday, August 06, 2006

A New MarshalPlan!

There is now available a standalone version of MarshalPlan. This computer-generated gizmo -- and the words that follow are carefully chosen -- is a software representation of an approach to evidential inference that emerged out of my collaboration with David Schum in the late 1980s and early 1990s. This software -- one might call it MarshalPlan 2.0 -- is not exactly a "concept" of a software tool. It is a bit closer to being a "proof of concept"; i.e., MarshalPlan illustrates how my approach to inference works -- and that it does work. In any event, MarshalPlan is very clunky technologically speaking -- for example, some of the "buttons" in the various "stacks" that comprise my evidence marshaling system do not work; some or all of the color schemes I use are, at best, aesthetically unappealing; and the layout of fields, buttons, and other objects on the visual fields is inefficient.
  • The original version of MarshalPlan was created with HyperCard. The current version of MarshalPlan was created with Revolution Enterprise, a sophisticated successor to HyperCard. The sophistication of Revolution, however, is not matched by any sophistication that Tillers has in programming or scripting. Indeed, even though Revolution makes it easy for even an idiot to develop software, the author of MarshalPlan 2.0 is less than an idiot.
  • This clunky software, moreover, comes without commentary. You will not even find many explanatory notes. The only software manual you will find takes the form of published literature that lays out some (but not all) of the theoretical underpinnings for MarshalPlan. See, first, P. Tillers & D. Schum, A Theory of Preliminary Fact Investigation; and, second, David Schum, The Evidential Foundations of Probabilistic Reasoning 491-504 (Wiley & Sons, 1994; paperback reprint, Northwestern University Press, 2001). If you wish to have explanations, you will have to invite me to give a talk (and you will have to pay my expenses) or you will have to attend Professor Edward Cheng's Evidence Colloquium at Brooklyn Law School on November 21, 2006 -- my general topic will be "Reasoning about Evidence." Alternatively, come to the Cardozo Law School conference on Graphic and Visual Representations of Evidence and Inference in Legal Settings. It is possible that I will talk a bit about MarshalPlan 2.0 then.

    If you would like to download MarshalPlan from the web, please send me an e-mail request (at peter@tillers.net) and, if your purposes are benign, I will send you the URL for the download page and a username and a password.

    Today I added one (but just one) word of explanation in MarshalPlan 2.0. I inserted the following comment in a field in an otherwise empty stack called "Probative Value & Credibility Assessment":

    MarshalPlan has no stack that lays out a strategy for assessing the probative value of evidence. This is only because I have nothing much to add to the extensive and exciting work that has been done and continues to be done on various kinds of inference networks. The leaders in this field are people such as Henry Prakken, Douglas Walton, Timothy van Gelder, and David Schum. (Judea Pearl has a very important theory of inference networks. But his theoretical commitment to the notion that there is no good inference without judgments about causal links sharply limits the utility of his theory for present purposes.)

    Credibility assessment is a special form of weight-of-evidence-assessment; i.e., it involves a special form of inference network logic; i.e., reasoning about testimonial credibility involves a special form of hierarchical evidential inference. The most elaborate and sophisticated theory of witness credibility assessment is the one developed by David Schum. (A gap in his theory -- the only significant gap I can find -- is the absence of much of anything about "linguistic uncertainty," the uncertainty one may have about a testimonial report because of the limitations in the speaker's ability to use language and, more generally, because of the inherent imprecision and ambiguity of language.)

    The absence of a stack [in MarshalPlan] for weighing the probative value of evidence (including testimonial evidence) proves that weighing evidence involves more than weighing evidence: The weighing of evidence requires discovering evidence, imagining hypotheses, developing arguments, and a host of other mental activities, some of which are identified by the stacks in MarshalPlan. All of these activities involve imagination and constructive mental activity. (But these mental processes are not limited to imagination and mental fabrication: They also involve the EVIDENCE, which is an essential constraining force on human imagination and judgment.)

    A Very Import Acknowledgment: If MarshalPlan 2.0 has any virtues, the credit belongs much more to David Schum than it does to me. Correlatively, if MarshalPlan 2.0 has any vices (and I think it has many), the fault is literally entirely mine.
    Post a Comment