solicited review for AI and Law

	Review of REASONING WITH RULES:
	An Essay on Legal Reasoning and Its Underlying Logic


Jaap Hage is one of the mysterious members of the AI and Law community:
someone capable of significant insights bathed in inexplicable prose,
someone who delivers scholarship and erudition through burbling lines of
symbols.  It is therefore with eagerness and fear that one approaches
his major work to date, REASONING WITH RULES (Kluwer, 1997).  A good
book is an image of its author, and this book is in fact a faithful
portrait of Jaap Hage's thought.  It is an important addition to the Law
and Philosophy Library series because it addresses important issues in an
accessible way.  The surprise is that the book is not really written for
the AI and Law community, which has been the cauldron for the themes
that emerge in the book.  It is a book for the wider philosophical legal
readership, an exporting of the ideas of AI to the philosophy of law.
It is a book that I am happy to say AI and Law researchers will be quick
to recommend to outsiders.

The book is a meditation on the formal modeling of rule application.  It
begins by arguing for defeasibility, then brushes through Raz
(exclusions), Toulmin (argumentation), and Naess (balancing reasons),
while commenting on some philosophy of rules (Anscombe and Searle).  The
book is preoccupied with distinctions:  between rules that explain and
rules that justify, brute facts and institutional facts, constitutive
and normative rules, rules and reasons, social rules and social reasons,
social reasons and personal reasons, goals and reasons, principles and
rules, and so forth.  The resulting map is the kind of thing that would
result from a survey course in the philosophy of rules.  It is
unfortunate that Hage chooses to be glib in his treatment of these
distinctions, taking a common language philosophy approach, rather than
delving into the formal distinctions that modelers might make that would
reflect philosophical differences.

One important topic for Hage is "deontic collapse," which refers to
the way 

		P --> OQ

in deontic logic is replaced by something like

		P >-- Q

in a system of defeasible reasons.  On this topic, Hage's common
language approach is effective.  Having made enough room for all kinds
of reasons, Hage has us wondering why some would insist that all norms
be expressed as deontic facts.  That is, why one would insist on the
traditional deontic approach, when there are so many legitimate
alternative ways of making facts, norms, rules, and reasons interact.
More interestingly, Hage's explanation is both arresting for its
plausibility and frustrating for its twists:

	in all cases of deontic collape ..., we have reasons to perform
	some mental action that would lead to a particular result, and
	these reasons do not only make that (sic; [make it the case
	that]?) we should perform this action, but also to (sic; [but
	also make]?) the results of performing this action.  The reasons
	why we should let a goal, principle or rule generate reasons
	function [they function] as reasons why the goal, principle or
 	rule actually generates reasons, [...] and the reasons for
	assuming the presence of a state function [they function] as
	reasons why this state is already present.  (p. 126)

It is a mouthful, no matter how it is edited.  While this is not the
kind of philosophy that I appreciate best, and it almost certainly needs
some deciphering, I find Hage here no more convoluted than a Stanley
Fish-Ronald Dworkin debate, and much more rewarding.  Hage probably has
in mind something like the difference between quoting and disquoting in
logic ('Boston is a nice city' versus '"Boston" has six letters'), or
the difference between evaluation and protection from evaluation in an
environment such as an operating system's command-line interpreter ('rm
*' versus 'echo rm *').  "P >-- Q" is a substitute for "P --> OQ"
because the latter is asserted within a context of evaluating reasons
normatively.  This has to be the right idea about deontic collapse (and
one suspects that if more deontic logicians used UNIX, there would be
less animosity toward those who do not take normative statements to have
truth values).  Of course, one wishes Jaap Hage could deliver the
knockout punch rather than merely float like a butterfly.

The second half of the book launches into RBL, Hage's "Reason-Based
Logic" that weighs reasons and permits an interaction of rules and
principles.  RBL possesses enough ways to express reasons for applying
rules to satisfy someone of even Jaap Hage's broad interests:  cases,
inrterpretations, rationales, exclusions and priorities all fall within
the scope of RBL, at least in Hage's expert hands.  

The book ends with some comparison of technical approaches and a curious
recollection of the main metaphors used in the book.  It is clear that
these metaphors impress Hage deeply and drive his work, though it is
unfortunate that these metaphors will not amaze anyone, inside or
outside of the related technical fields.  Those metaphors are:  (1) the
container metaphor for logic (as a representation), and (2) the 
balance-of-forces metaphor for (judicial) reasoning.  The first idea is
that logic is less useful than many logicians would have us believe.
Deduction is essentially noninformative.

	Each possible world can be considered as a container of
	information.  ... Any sentence that contains [only] information
	that is ... contained in the theory will be true in all models
	of this theory.  (p. 246)  

Hage probably owns a lot of containers that can be stored one within the
another, since that is really the picture he seems to have in mind.  He
should add that there is a way in logic to combine containers (to
conjoin theories), not just to look inside a container to see what other
containers might be contained therein (deduction).

The second idea is that reasons push against each other, cancelling as
well as reinforcing, and the result of reasoning may hide the complexity
of the reasons that figured in the result.

	Just as the combination of the forces determines in which 
	direction the body will actually be moved (accelerated),
	the combination of the adduced reasons determines whether
	a rational audience will accept the conclusion or its negation,
	or refrain from judgement.  (p. 252)

The beauty of this metaphor is the disconnection between the existence
of a force and its manifestation as acceleration.  Just as an
acceleration vector can be explained by a single resultant force, rather
than a balance of forces, reasoning can be explained as a deduction from
premises (a deductive reconstruction), rather than an adjudication of
disagreement.  The picture is so clear that the important moral might be
lost:  those who criticize defeasible reasons from deductivist stances
(by holding that a deductive account could always be reconstructed post
hoc) should have to face the fact that there are legitimate alternative
views that may in fact be more useful.

Hage's gentle and encompassing introduction to the problem of applying
rules is a useful review of the arguments that have brought many
researchers in AI and Law to defeasibility.  It is by no means an
exhaustive review of the arguments, and it provides no answers to the
usual rebuttals.  It is particularly silent on the questions raised by
Carlos Alchourron and Thorne McCarty, which to be fair, were also in
progress in 1997.  But, like Bart Verheij's thesis supervised by Hage in
this same time, it starts from the basic and most naive logical position
and sees what considerations motivate deviation.  This is the main
feature that makes it a useful bridge to the communities that have not
yet faced this line of thought.

Hage's attempt to go deeply into the philosophical issues sometimes
gives dubious results.  One finds the following classification (p. 69):

	The fact that 
	taking Mary's purse
	would be theft		makes it 
				the case that
						John ought not to 
						take the purse.

	The fact that taking
	Mary's purse would be
	theft			is a reason
				to accept that
						John ought not to 
						take the purse.
						
	The fact that taking
	Mary's purse would be
	theft			is a reason
						for John not to
						take the purse.

	The fact that taking
	Mary's purse would be
	theft			is a reason
				to assume that
						John ought not
						to take the purse.

Surely no one will disagree with this clarification of the interaction
of fact, obligation, and epistemics.  But will anyone feel enlightened
by it?  The problem seems to be that Hage wants to engage the legends --
Raz, von Wright, Ross, Kelsen, Toulmin -- in serious debate on issues
that are no longer puzzling.  Hage is trying to look back on early work
that frames a problem simply, then compare it to the varieties of
alternative framings that have occurred.  Most contenders with the past
would instead have taxonomized the historical variations or conceded the
minutiae in favor of the main themes.  Hage is like a fighter who has
dissected the champion's style and is ready to challenge for the title,
but the old timers are dead, the ring has been torn down, and boxing has
been banned.  The one nice thing about Hage's testiness is that he
invites comparisons of long-fetid bodies of thought with the
contemporary themes of AI and Law.  Those who hide among dubious or
crude distinctions, anankastic versus deontic, regulative versus
constitutive, principle versus reason, have been served notice that
Hage is itching for a fight.

The book stands on higher ground when it turns to RBL.  Here, the main
innovation is a willingness to introduce predicates whenever they are
needed:  applicable, excluded, against_purpose, interpretation_of,
correct_interpretation_of, prevails_over, incompatible, within_purpose, 
etc.  RBL is something like a pre-PROLOG theory which does not presume to
provide a foundation for every predicate whenever it is introduced.
Need to use the concept of against_purpose before you have listed all
possible ways of deriving against_purpose?  This is not a problem for
Hage.  Either someone will assert an against_purpose relation, or else
someone will theorize about how to derive against_purpose from something
more basic.  Or both.  To those raised on mathematics, this is an
unacceptable top-down approach.  One should define new concepts in terms
of those things already defined.  One should definitely not leave
important concepts undefined.  But philosophical logicians have often
been top-down in their formalizations (for example, von Wright in his
delivery of deontic logic, and Rescher in his treatment on dialectics).
And many computer scientists prefer top-down design as a way to increase
organization and clarity of a symbol system.  In an age when philosophy
can be illuminated by different formal systems, when necessary and
sufficient conditions are already suspect as formal criteria of
intellectual adequacy, who can blame Hage for his PROLOG-like formal
manner?  We should probably all be asking ourselves whether our formal
systems are unnecessarily rigid.

Perhaps more of us should be writing RBL, instead of worrying that
mathematicians will brush such analysis aside.  Just by our raising
this question, Hage's book hits its mark.

The final thing to note about "Reasoning with Rules" is that it is very
little concerned with the technology and application of AI and Law.
Fortunately, few legal practitioners who enjoy jurisprudence will find
the discussion unmotivated.  RBL may even be suggestive to anyone
building a legal expert system.  For these two reasons, recommending
Jaap Hage's book will be one of the very best ways to bring people to AI
and Law in larger numbers.