Arguments: Part Two:
Explaining Stuff

An argument, as opposed to a narrative or a script for example, serves as an explanation. A simple way to start thinking about this is to consider an argument as a sort of investigation.

When you investigate something, your task is to collect evidence and tell the story of what happened. For example, an insurance adjustor might be called to investigate an electrical fire. The adjustor would have to collect data from the scene, like the source or location of the fire, the extent of the fire damage, lingering odors or residue, and so on. Also, the adjustor might interview the policy-holders to determine what kind of people they are — what is their claim history? What is their financial history? What is their police record? And so on.

The first kind of data is the stuff that we need to explain in the investigation. Let's call that "trace data". The second kind of data is stuff that helps or hurts our various reasonable explanations. Let's call that "non-trace data". One's competence at identifying and one's ability to distinguish what's what in the argument is paramount. One thing that aids both of these skills is starting from in the right place.

When you start conducting an investigation, what do you do first? Accuse someone? Threaten the bystanders? Eat a sandwich? Likely none of these. First, you want to frame a good question that your investigation will attempt to answer. In lots of regular cases, that question takes the general form "what happened?"

It can be daunting to figure out what the best first question is. In really complicated cases, like debates about hot political issues, it's virtually impossible to get a productive discussion started without first figuring out what you're arguing about. Since it's such a complicated task, let's focus on the simple insurance adjustor case.

The adjustor might first investigate "what did the policy holders eat for lunch?" But that would be silly. More reasonably, the adjustor might investigate "what caused the fire?" That's more like it. Let's agree that it was already determined to be an electrical fire — this helps to narrow the range of answers. For example, we do not consider answers like "a plane dropped napalm on the house." (Though one could cook up an exotic case where this is actually what caused the electrical fire, but it's far-fetched enough to remain off the list of reasonable explanations.)

Reasonable explanations to the slightly-modified question "what caused the electrical fire?" could be:

• A failed circuit breaker
• A failed GFI
• Faulty wiring or a short
• Old or worn out wires
• The fire was started deliberately

These are all pretty reasonable answers to the basic question. But we can't really decide which one is the best answer without further investigation.

In the course of the investigation, the adjustor discovers wires that were not touched by the fire, but whose insulation is cracked and brittle. The adjustor also discovers that the electrical work was done without a permit by an uncertified contractor.

The first of these pieces of data is what we agreed to call "trace data". It is trace data because the conclusion needs to explain it. That is, whatever answer to the initial question sounds best will have to be able to explain this data. The second piece of data is what we agreed to call "non-trace data". It is non-trace data because it is relevant to the question at hand, but it does not need to be explained by whatever conclusion we choose.

To evaluate which answer is best, we consider which best explains the data. The obvious case is that the wires were old. If that is not obvious, then we could add the following background information — also non-trace data: badly insulated wires can arc, which can cause a fire. Again, this piece of data does not need to be explained, but it significantly helps the "worn out wires" explanation. We might say that it "underwrites" the explanation.

With this established — assuming that all reasonable parties agree that it's the best explanation of the traces — the investigation can continue. Probably the next most obvious question is "to what extent should insurance cover the damage?" This is a much more subtle sort of investigation — one that will require a heap of non-trace data.

One could be tempted to ask what appears, on the surface, to be the same question: "should insurance cover the damage?" This is not the same question. In fact, it's not a particularly interesting question, for the appropriate response is either "yes" or "no".

On the other hand, "to what extend should insurance cover the damage?" includes both "entirely" and "not at all" (yes and no), as well as shades in between. More often than not, the best answer lies here.

Coming soon: the dangers of the false dilemma.

© 2005 Sorrell
October