Overview of the Visual Logic tool

This chapter gives an overview of the operations of Visual Logic. For detailed information on using the tool, consult The Visual Logic User Guide (see insert).

Introduction

VL is a software tool designed to explore the logical concepts presented in the preceding discussion. It provides various means for the visual display and manipulation of associated clause trees. The user interacts with the tool primarily in one of two 'modes.' 'Query' mode allows the user to find evident, supported, and conflicted terms from the clause database (program). It also provides 'explanations' in the form of clause trees. 'Explore' mode on the other hand allows the user to essentially trace the execution of logic programs by expanding and retracting arbitrary branches of the clause trees.

Visual Logic was implemented using SWI-Prolog and the graphical package XPCE. Both SWI-Prolog and XPCE were developed at the University of Amsterdam and have been ported to various Unix/X-Windows environments as well as Microsoft Windows (Win32s, 95 and NT). VL itself has been written with portability in mind and should run without modification on any platform for which SWI-Prolog and XPCE are available.

The tool is composed of several components:

• Main module: integrates the various modules and provides the main user interface.
• Computation module: computes derivations for Query mode.
• Explore module: lets the user explore the logic program's space in a non-linear fashion.
• Query module: lets the user query the system with specific goals such as "show all supported propositions" and provides an 'explanation' in the form of annotated clause trees.
• Results module: accumulates and shows all known results.
• Built-in editor and program lister: lets the user edit the input source file (i.e., the program clauses) and view the current clause database. The source file is a plain text file consisting of program clauses, the exact syntax of which will be described in chapter 5.

Below is a screen shot of the tool. Due to limited screen space, not all windows are fully visible.



Each of the graphical windows corresponds roughly to a module. The main window (at top) is the command center for the tool. Here, the user can load the source file, quickly switch between the various windows, launch the built-in editor to modify the source file, see a listing of the program clauses, invoke the help system, or quit the tool. The heart of the tool lies in the Query and Explore windows, which will be described in detail next. For now, note that the Query window shows the annotated clause tree for a, from the example program in chapter 3.

Query window

In this window, the user can query the tool with specific goals. For example, in a program relating to a trial, the user may ask whether guilty is evident. This is done by typing in guilty then clicking the Evdnt button. Likewise, the user may ask whether forensic_evidence is supported / conflicted. It's also possible to do a 'general query' on a term (Query button). In this case, the tool will first check to see if the proposition is evident; if that is the case, it then checks whether the proposition is supported and/or conflicted. Goals may contain variables, e.g., knows( defendant, Y). In fact, typing in X and clicking Query gives all possible derivations from the current program. (Depending on the size of the database, this may take a long time.)

The tool responses to a query in one of two ways. If the goal fails, it simply returns a 'failed' message. Otherwise, it lists the results on the left pane. A proposition may be listed more than once if there are more than one way to derive it. Clicking on a listed proposition will display its associated clause tree on the right pane. In effect, this provides an 'explanation' for the answer. Each node is displayed with its contingency set (shown in list notation). The nodes are color coded as followed:

• Blue: Evident.
• Purple: Supported.
• Bright Red: Conflicted.
• Black (bold font): Supported and conflicted (i.e., moot).
• Green: either TRUE or ancestry resolved (a node whose ancestor is its negation).
• Black (regular font): none of the above.

Each tree is display on its own white, raised area, called a cell. Various ways are provided for changing the display of the trees. A node can be collapsed, an entire cell can be collapsed, or all cells can be collapsed to give a broad view of the entire display.



In contrast to Explore, Query computes the entire tree and displays it at once. The user does not see how the trees are built, only the finished result. On the other hand, Query computes all values for any given node, whereas Explore only shows whether a node is evident. It is important to realize that a node's attribute shows its calculated global value which is not necessarily borne out by the tree it is in. That is, the fact that a node is colored blue indicates that it can be shown to be evident; however, it may not be the case that this particular tree shows that the node is evident. Also, the value of a node as shown is the most general value that can be computed for it. Therefore, if a node is blue, then its term can be shown to be evident and no more, i.e., it is neither conflicted nor supported. In the sample screen above, guilty is evident but not supported since one of its interior nodes, forensic_evidence, is conflicted. forensic_evidence is in fact moot since it is both supported and conflicted.

In some instances, it may be preferable to show each node's value pertaining only to the tree it is in, that is, its local value. In Explore mode, the user can toggle between showing global and local values by clicking on a cell, as shown below. In the cell on the right, the parentheses around the label a indicate this cell shows only local values.