I've done both.
Law teaches how to bring rigour to fuzziness. In software we can set boundaries and push the fuzziness back into the problem domain. Lawyers can't do that, they must address the world as it is.
Let me explain what I mean.
Suppose you use constants with values for HOT, WARM, COOL and COLD. In a normal computing system you will need to define these precisely -- you'll need to give float or decimal values to the thresholds between them.
By setting those thresholds, we simply make the complexity go away. Users wind up carrying the bag for any paradoxes that arise.
Lawyers can't do that, they deal with linguistic variables:
The reasonable person similarly circumstanced.
The buyer at arm's length without notice.
Offer and acceptance.
Reasonably foreseeable.
You can't give any of these a hard value (though the general features of each are pretty clear for almost all cases). They are fuzzy sets.
Dealing with fuzzy sets without destructively reducing it to ordinary logic is almost unique to law. About the only other people doing it are some control engineers and the attendant theoreticians.
edit:
I'd also add that the experience of extremely careful examination of facts and teasing out subtle distinctions are useful skills. So too the learned ability to do that with buckets of documents, quickly, and then having to draft reasonably straightforward briefs or case notes.
I find that skill particularly useful when discussing requirements / user stories.
At the top of the list I'd put "the learned ability to do that with buckets of documents, quickly, and then having to draft reasonably straightforward briefs or case notes."
One's goggles in going through the rooms of buckets of documents are tuned to those "fuzzy sets" of legal issues. An ancient lawyer once said that lawyering is the art of finding the relevant.
Or as Edmund Burke put it: "Law sharpens the mind, by narrowing it."
But don't lawyers deal with this fuzziness to some extent by putting the question to a jury? For instance, suppose we were writing a software application to decide when to shut down a machine because it got "too hot". We could go ahead and assign a specific temperature, or perhaps set up an equation with a few different readings.
Now suppose there was some litigation because a tennis player collapsed during a tournament that was "too hot". Would the lawyer define "too hot", or would the jury? You could have a situation where all parties agree on the law, and that it turns on whether it was "too hot", and the legal system would treat the jury's definition almost as definitively as a temperature reading input into an algorithm, right?
On the other hand, if the player were to sue for negligence, then the question of whether it was "too hot" is a factual question that the jury would decide, after the lawyers present evidence for their preferred answer (and possibly evidence against the undesired answer).
The jury's finding would be a finding of "fact" as to whether it was "too hot" would apply only to that case, because their finding would necessarily depend on the specific facts of that case. Juries cannot make findings of "law", which are general principles/rules that can be applied to other cases.
The jury.
> the legal system would treat the jury's definition almost as definitively as a temperature reading input into an algorithm, right?
For this case, yes. For all time, no. The fuzzy set of "too hot" would become the legal precedent (the stare decisis). The actual temperature would remain a question of fact and not law.
If you read up on how fuzzy logic is applied in control systems, you'll see that there's always steps to take non-fuzzy inputs and "fuzzify" them before performing the fuzzy logic itself. Afterwards you de-fuzzify to get a crisp output.
Courts work very much like that: the law retains fuzziness because no crisply algorithmic system can encompass the total complexity of the human world. That's how equity arose ... which is a law history lesson for another day.
There are a bunch of papers on what counts. My take away was that many courses of study say they have a way of reasoning, but at the end of 4 years (or an additional 3), most students perform no better on abstract tests of whatever metric was selected for that major (e.g. law, analogic reasoning