Thursday, November 06, 2008

Trespassers Will

David McCoy complains about a sign reading "No Dumping - Violators May Be Prosecuted". He objects to the permissive and non-threatening implications of the modal verb "May", and is nostalgic for the unequivocal days of right and wrong, of firm laws and firm enforcement of the laws, when cardboard signs read "Trespassers Will Be Shot". [Modal Verbs: The Words Behind the Loopholes ]

I want to make two points here. The first relates to the purpose of the sign - what is it trying to achieve, and what kind of language is appropriate to this purpose. The second point relates the nature of ambiguity - is May really more ambiguous than Will?


The purpose of signs like these is of course admonitory - they are designed to dissuade people from doing something. You might think that an uncertain sanction (may be prosecuted) is less of a disincentive than a certain sanction (will be shot).

But then which of the following warning signs do you find most persuasive? Perhaps it depends on your national culture. We British are a nation of accountants, so we like to be told what the maximum fine is.
  • British: If you climb on this electricity pylon you will be fined £100
  • German/American: It is a Federal Offence to climb on this electricity pylon.
  • Italian: If you climb on this pylon you may die.
(I inherited this joke from my father, but I don't know whether it was his own observation or he got it from somewhere else. I think there may have been a French one as well. Please comment if you have a source, or any other nationalities.)

Based on the simple linguistic analysis David recommends, the possibility of death is not as strong a sanction as the certainty of a fine. Like Hermione's joke in the first Harry Potter book - "We could all have been killed - or worse, expelled."

David likes warnings to be proper threatening. When I was growing up, parents and teachers said things like "Don't do that or else": that was regarded as proper threatening in the days of "unequivocal days of right and wrong, of firm laws and firm enforcement of the laws". So we usually did what we were told and didn't find out what the punishment would have been.

Following David's logic, parents should say things like "Don't do that or else you will be grounded for three days and fined 2 weeks pocket money." But is precise specification of the sanction really more effective in encouraging good behaviour? Or does it merely encourage calculated risk-taking?


Meanwhile, “Trespassers Will” is still pretty indefinite. Does it mean ALL trespassers or SOME? Immediately or later? If I trespass and I am not shot, is that a counter-example or a stay of execution? If no trespassers have ever been shot, what should I conclude from that? Do I want to be the first?

Perhaps it would be better to have a sign, like road safety signs, saying “341 trespassers have been shot this year. Some of them survived. Please take your litter home.”

Or perhaps this …

Next to his house was a piece of broken board which had: “TRESPASSERS W” on it. When Christopher Robin asked the Piglet what it meant, he said it was his grandfather’s name…it was short for Trespassers Will, which was short for Trespassers William. And his grandfather had had two names in case he lost one—Trespassers after an uncle, and William after Trespassers.


I agree with David that the English language is tricky. But I think he is wrong to blame modal verbs - may, can, shall, might, etc. - which he identifies as "the words that power loopholes." These words tend to be used consciously, not to create inadvertent loopholes but to deliberately underspecify some outcome. In David's example, the rule is pretty clear ("No Dumping") but the sanction is underspecified ("Violators May").

Meanwhile, "Trespassers Will" is a false universal, with no modal word to draw your attention to the ambiguity. In my opinion, this kind of ambiguity is much more dangerous - both when used deliberately, to mislead or persuade, or when used inadvertently.

Some people find the NLP metamodel useful for identifying structural ambiguities in English speech. (Other people think NLP is the work of the devil, but of course the devil is in the detail.) All structural ambiguities, or just some? I suspect there is no assured method for eliminating all possible ambiguity.



Update: Here's a great example, found by @ShawnCallahan

Embedded image permalink
Updated 8 June 2014

3 comments:

david.mccoy said...

Richard - I believe we are at odds on two of my views.

(1) On the power of WILL vs. MAY:

You say “Meanwhile, “Trespassers Will” is still pretty indefinite. Does it mean ALL trespassers or SOME? Immediately or later? If I trespass and I am not shot, is that a counter-example or a stay of execution? If no trespassers have ever been shot, what should I conclude from that? Do I want to be the first?”

To examine my point, we should compare two sentences:

"Smokers will be denied insurance coverage"
"Smokers may be denied insurance coverage"

Of the two, I argue that "will" is by far the more powerful and further, that it is rigid in its power. Specifically, if you implement that construct in a business rule engine (or even in any coded system) it is always definitive.

If [applicant] = “Smoker” THEN [policy_status] = “DENIED”

There is no need to ask if ALL or SOME will be denied – the end result is 100% denial, as defined in this rule. If we stick to mushy actions as implemented by humans, then certainly I can accept a fuzzy interpretation. Rules are broken all the time – hence the mess we are in right now. But, as implemented in silicon, the LOAD, COMPARE, BRANCH NOT EQUAL compiled logic is anything but fuzzy. The outcome is always the same: DENIAL. Will is definitive in this case.

(2) On comparing WILL vs MAY in alternative scenarios:

On your blog, you state, “Based on the simple linguistic analysis David recommends, the possibility of death is not as strong a sanction as the certainty of a fine.”

No. I believe two equivalent constructs that differ only by MAY or WILL are easily compared. I always favor the WILL construct.

BLAH BLAH WILL BLAH >= BLAH BLAH MAY BLAH

Your scenario is a bit unfair as you are cross-comparing two vastly different statements. I would reword it to say: “Based on the simple linguistic analysis David recommends, the possibility of death is not as strong a sanction as the certainty of death.”

Consider:

I MAY feed you to a crocodile
I WILL give you a penny

All I would say about these sentences is:

I MAY feed you to a crocodile is not as strong as I WILL feed you to a crocodile
I WILL give you a penny is stronger than I MAY give you a penny

I will not (and did not) assert that the possibility of being fed to a crocodile (the possibility of death) is of lesser interest to me than the assurance of gaining a measly penny.

The later argument reminds me of Pascal’s Wager. The infinitely small possibility of an infinitely important outcome is the ultimate “may.” But, it is a “may” that only a fool would ignore.

Richard Veryard said...

Of course there is a linguistic difference between WILL and MAY, but if your agenda is to find ambiguity, then I don't think MAY is the chief culprit.

Let's look at your example. What is the real difference between the two sentences?

smokers will ...
smokers may ...

In this particular case, you are probably safe to assume that "smokers will" means "all smokers will always". However, a pedantic systems analyst would need to verify this assumption.

But putting this pedantry aside, in both cases the insurance company has the power to deny insurance coverage. As far as I can see, the main differences are (i) where in the process the insurance company makes this decision and (ii) how visible and predictable this decision is to the smoking customer.

So I can't see that MAY confers any more power to the customer. Arguably, it is the visibility and predictability of WILL that gives slightly more power to the customer.

So I prefer to discuss questions of power in WHO/WHOM terms, within the context of a distributed system with multiple stakeholders, rather than purely linguistic analysis.

Systems thinking aside, there is another notion of power implicit in your comment, which has to do with the propositional strength of a given rule. I agree that WILL (with ambiguities removed) has greater propositional strength than MAY.

Of course I might choose values of P and Q so that MAY P is more persuasive than WILL Q, but this is nothing to do with propositional strength.

However, propositional strength is not always the same as logical precision (lack of ambiguity). If you are looking for an "iron-clad business rule that has no loopholes", then it is precision you need, not propositional strength. And because WILL is vulnerable to greater ambiguity than MAY, it is potentially less precise.

Richard Veryard said...

And we should not forget the ambiguity of all the other words. Who counts as a smoker? What about someone who is currently trying to kick the habit, and hasn't had a cigarette for three weeks? Who counts as a trespasser?