This paper, part of a special issue devoted to qualitative reasoning about physical systems, is an important contribution to the field. The author explores the issue of how the function of a circuit (its purpose) is related to its structure (the schematic diagram). Causal reasoning analyzes how disturbances from an operating point propagate through a circuit, thereby determining the qualitative behavior of a circuit from its structure. In addition, unlike quantitative predictions, this method produces intuitive, causal explanations for the behavior predictions. These behavioral predictions, combined with their explanations, form the basis for reasoning which explains how the purposes of the circuit are achieved by that behavior.

The author makes clear the difference between the general approach that he has taken and the methodology used in expert systems:

Causal reasoning encounters severe complexities which can usually be resolved by incorporating teleological or geometrical knowledge. If I were interested in building a performance program, the temptation for including this extra knowledge would be overwhelming. However, that would be shortsighted. To understand what causal reasoning, or teleological reasoning is, one must study it in isolation uncorrupted by other forms of reasoning. Otherwise one has merged two types of reasoning without ever identifying either one individually. In addition, little scientific progress is made and we are not much closer to the ultimate goal. . . . To achieve robust performance, the underlying theories must be identified. This methodology stands in sharp contradistinction with the popular expert-systems methodology. Expert systems are aimed at producing what performance is possible in the short term without consideration of the longer term. . . .

Both causal reasoning and reasoning about purpose (teleological reasoning) have been implemented in the EQUAL program. EQUAL takes as input the schematic for a circuit, and produces a qualitative prediction of the behavior of the circuit, an explanation of that behavior, and a teleological parse which relates every component to the purposes of the overall circuit. In performing its causal analysis, EQUAL utilizes a component model library which describes the behavior of every type of electrical component. The teleological reasoning, on the other hand, utilizes a grammar of basic mechanisms to parse the circuit’s behavior. An example of the lowest level causal analysis of an amplifier circuit derived by EQUAL is:

The increasing voltage at node IN is applied to the base-emitter junction of Q1. By the transistor law, this causes an increase in current from the node OUT into the collector of Q1. By Kirchoff’s current law, this same current is flowing out of the bottom of R1 into node OUT. . . .

Building on this description, EQUAL obtains a graph describing the mechanism by which a circuit achieves its input/output behavior, and the purpose of an individual component is defined by how it contributes to this mechanism. By using a taxonomy of component configurations, this “mechanism graph” can be parsed into higher-level descriptions such as:

Resistor RC2 is functioning in I-LOAD configuration, for transistor Q1 functioning in CE configuration, which is STAGE1 of CASCADE, which is BASIC-AMPLIFIER of FEEDBACK. . . .

There are many, many subtleties and complexities involved in the theory, such as the nature of causality, causal heuristics, detection and analysis of feedback, dealing with ambiguity, and the analysis of operating regions. This key paper in qualitative reasoning is not an easy paper to read; but then again, the author is solving an extremely difficult problem. The reader must be familiar with electronic circuits to follow this paper in all of its depth and sophistication. It might have been useful for those not familiar with the field of qualitative analysis if the author had compared his approach with that of others solving similar problems, such as [1].