Software Engineering methods should be applied to legal practice

[arendt]

DU Lawyers - I could use some feedback here.

Feel free to flame, although I would hope that is because I am
unaware that what I am proposing has already happened

------

Software Engineering methods should be applied to legal practice
by arendt

1. The origins of Software Engineering

Software engineering (SWE) arose to improve not only the control over
the software production process and its costs, but also to improve the efficiency
and reusability of the code produced. One of the earliest successes of SWE
was to determined that certain computer-language constructs, such as "go to"
statements, almost always led to problems (e.g., "spaghetti code"); so, their
usage was strongly deprecated.

SWE was able to make progress because it operates in a tightly constrained
linguistic universe. This universe contains only the instructions of a particular
computer architecture, and the instructions of operating system functions
built to manipulate that architecture. Later, higher level languages abstracted
these instructions across a large number of hardware platforms and operating
systems, but the fundamental finite-ness of the language remained in place.

To economize on precious hardware, early computer architects ruthlessly
limited language constructs. Later, as hardware became cheaper,
instructions sets became more complex. But, this was proven to
be inefficient in the well-studied "CISC/RISC"* architecture wars.
The end result of the co-evolution of these two architectures was to
have a RISC core surrounded by a CISC instruction translator,
where the relative complexity of these final CISC instructions was
much lower than 1970s era CISC instructions.

SWE's seminal contribution was to move computer programming out of
the era of pure text, and into the era of graphical representations, graphical
interfaces, and, eventually, automatic checking of SWE constructs and references.
Although this is obvious in retrospect, at the time, it took some serious
argumentation to make the case:


"Can you imagine spending five years' salary on a custom-built house
on the basis of an exhaustive narrative description of how the home
will be built? No pictures, no plans, no visits to a similar house - just
a 150 page narrative. 'The living room, which faces south-southeast,
will be 27' x 16' at its greatest width, with the western half taking a
trapezoidal form, the west wall being 13'4" long (abutting the northern
portion of the east wall of the kitchen)...'

"If you use English to describe a complex system, the result takes up
so much space that its hard for the reader to grasp how the parts fit
together. Worse than that, English has some built-in problems that make
it very difficult to use where precision is needed..."

"Not only but notwithstanding, and/or unless...

"In trying to understand narrative in POLICY DOCUMENTS, memos,
and specifications, we continually run up against the variety of possible
forms that English allows...

"...as frequently happens, the POLICY STATEMENT involves combinations
of conditions..." (these can be expressed precisely via tools such as decision
trees, decision tables, if-then-else constructions, etc.)

C. Gane & T. Sarson
"Structured Systems Analysis: tools and techniques"
pp. 4, 76-7, & 80

SWE created the potential for vastly more reliable software, although the majority
of programming organizations today still operate under the myth that really good
programmers are only slowed down by SWE. However, with the end of the
SWE language wars, and the adoption of the UML standard from Rational Software,
SWE seems to have found a standard formalism, and that formalism is being
forced upon engineers by management.

2. The legal profession is the ultimate CISC architecture

With the above background, one must ask why the same evolution has not occurred
in the legal profession.

By and large, the legal profession has used computers to merely automate a
spaghetti code process. The vast literature of precedent has been put into searchable
databases; but legal language is still a rats nest of "party of the first part" archaic
constructs. Today, these constructs are mainly a barrier to entry for interloping outsiders.
One only becomes a legal expert by rote study of fossilized phraseology, previous
cases, and glosses of those cases by expert commentators - sort of like Talmud.

The teaching of law still is done by the "case study" method, thereby directly
acknowledging the lack of formal rigor in the corpus of legal concepts. This
is the downside of English law's over-reliance on precedent, as opposed to
statute law.

It is largely due to the precedent system that the judiciary is so clogged with cases.
In any action, each side can bring a stack of precedents to court. It is often the
case that the side with deeper pockets can overwhelm the system with so many
precedents to be studied and considered that the other side's legal team gets
beaten by a bad argument they don't have resources to refute. This is no small
matter. The 2000 Florida election was decided by applying a law meant to
enfranchise minority voters in a manner that effectively disenfranchised them
by refusing a recount. This kind of gamesmanship is quite unremarkable in the
precedent-based legal system. That makes it no less egregious.

Of course, the whole point of having human judges is to recognize the unique
features of each case and to apply the many conflicting laws in a wise manner.
I personally think of the border line between legal and illegal behavior as being
a fractal. Each case is testing an unknown point of the fractal space, asking
whether that new point is inside or outside the fractal boundary of legality.
This analogy is appropriate because it was through the study of fractals that
mathematicians became aware of how much computing was going on in
natural processes, such as mountain formation. The formation of the law
is a process that has taken immense amounts of legal computation. A good
judge, despite the guild-like methods by which legal knowledge is gained,
can apply the results of that computation to correctly place a case vis-a-vis
the fractal boundary.

Mathematicians know that creating fractals is a compute-intensive process when
you are dealing with a well-defined mathematical space, such as the field of
real numbers. Imagine the difficulty of finding a fractal border when you don't
have an equation (i.e., statute), but only the status of a couple of neighboring
points (i.e., precedents) . Unless you are a good judge, its like stumbling around
in a mine-field.

3. The legal profession's backwardness cripples the legislative process

The majority of legislators are lawyers. Yet the Constitution allows legislators
to introduce any law they care to draft, whether or not it is obviously bad law.
Any bill can become law without any kind of judicial review before the fact.
The Constitution grants the power of deciding the legality of these laws to the
Judiciary, which decides after the fact of the law existing. The process of finding
test cases and appealing to the Supreme Court can take decades. This is
like waiting a year to find out the results of one compilation of your computer
program. Only instead of merely slowing one programmer down, you are wasting
the whole country's time.

This legislative situation has arisen because the precedent law is so complicated
that no one can predict the goodness or badness of a law. In fact, obviously
bad law, like the Dred Scott decision, has been approved by partisan Supreme
Courts.

4. What is to be done?

I doubt that we can dismantle the self-serving chaos, delays, and one-of-a-kind
costs of the legal system in the face of the entrenched lawyers' lobby. This lobby
fought the lay practice and then the computerization of trivial matters, such as wills
and testaments, for decades. They see every clarifying or simplifying improvement
to the legal process as a threat to their pocketbooks and priest-like authority. Tort
reform would merely limit the scope of the chaos, without changing its nature.

What we can do, however, is rationalize the legislative process, which as already
noted can allow even the most specious "law" to be passed.

There is nothing to stop a progressive lawmaker from creating a sensible
object-oriented representation for Congressional bills and using it inside his
office and staff. Its certainly possible to automatically translate computer-friendly
decision tables and the like back into legal boilerplate for submission. My guess
is that such back-translations would be as good as anything a lawyer would produce.
Except, when the intent of that lawyer is to deliberately introduce loopholes, or bias,
or legal booby-traps, or unwarranted advantage into a bill. Unfortunately, such
intents abound in today's lobbyist-driven legislative process.

Therefore, the introduction of object-oriented representations for laws must be
managed by first-rate legal minds. These lawyers must have as their goals to streamline
and speed up the legal process and to give transparency to the legal language.
Essentially, they are fighting for a different kind of Sunshine Law. The
conventional Sunshine Laws in government merely open discussion. But
legal mumbo-jumbo can create a fog so dense as to be impenetrable to mere
sunshine.

It is long past time to reduce the complexity of the legal system. Rather than sacrificing
the performance of the system by changing it, a correctly engineered object-oriented
legal system will improve the performance. As in the RISC/CISC war, complexity
will not be completely eliminated; but it will be greatly reduced. Pursuing the analogy,
old legal precedents would be able to be translated by a CISC front-end into clear,
object-oriented RISC precedents. Such translation would be relatively painless, much
as instruction set emulators allow old code to run on new computers.

Finally, lest lawyers fear for their livelihoods, who do you think will eventually be
writing the code? The computer profession has had "language lawyers" for decades.
These are the guys who argue over the minutiae of what constructs are *legal* in
a programming language. Furthermore, it is a tossup whether the public at large
finds programming language or legal language more opaque; so no need to worry
about deskilling in a RISCified legal profession.

Hello, lawyers, cluephone calling! My personal opinion is that the legal profession is
missing a major revenue source here. I thought that might get your attention.

----

* Complex Instruction Set Computer / Reduced Instruction Set Computer

[htuttle]

1. Interesting comparison

I hadn't thought about it before, but it leads me to all sorts of metaphors.

Researching case law precedents (or trying to read the Patriot Act) is like tracing someone else's spaghetti code, for example.

spaghetti code     n.     Code with a complex and tangled control structure, esp. one using many GOTOs, exceptions, or other `unstructured' branching constructs. Pejorative. The synonym `kangaroo code' has been reported, doubtless because such code has so many jumps in it.
http://info.astrian.net/jargon/terms/s/spaghetti_code.html

[sweetheart]

2. RISG - Reduced Instruction Set Governance

The problem with all this. I proposed similarly in 2 articles published in banking journals. I'll see if i can find them for this thread.

I proposed the natural evolution of all banking systems as an end-state in a legal-financial-architecture. Clearly assets and "orders" no longer "move" in the markets like they did during paper times... the concept that an order originates in chicago and gets routed to new york is still a concept today as paper has not been replaced, only mimic'ed.

To replace paper involves re-conceiving the transaction cycle as real-time and then re-conceiving the settlement process similarly. What does this do, trade-and-settle in .1 nanoseconds? between 2 sovereign nations? What does it do to contract law and international trade? The natural evolution of legal-computational systems would completely and radically restructure the court structure and the common accessability of an informed judiciary. It is a brilliant concept, but you face the same hurdles i once faced in financial services... a corrupt, global, wealthy, politically entrenched community of bankers that had no interest whatsoever in making cheques clear immediately without charges and the implicit loans of the "transaction cycle", as who has your uncleared bank cheque loaned but the bank.

The principal that a legal recorded transaction of multiple currencies could be initiated through settlement between 2 individuals/corporations/sovereign states is a deep and complex legal conundrum as the checkpoints being considered to make trade "fair" are themselves contingent on a longer settlement cycle.

Whilst i agree arendt, the corporatocratic elite like EDS (ignorant army of clones they are) have blowjobbed their way through bids like this selling documentum like stuff... big repositories of documents... legal workflow. I think myself i would charter a project in one court area and then re-write the code for a county, and then rewrite it again for a state and once more for a nation. It would take at least 4 zero-to-production rewrites to get the frame right for what you propose... as by far the most intensive changes would be in property law.

... as in property law, if the record is kept computationally, transaction times can be smushed to zero. That and the unnatural monoplies of money, petrol transit and finance for social matters.

The instructions are:

create citizen
delete citizen
Propose referendum
vote

[arendt]

8. A few more details, please, if you have them.

> the corporatocratic elite like EDS (ignorant army of clones they are) have
> blowjobbed their way through bids like this selling documentum like stuff...
> big repositories of documents... legal workflow.

Could you expand this a little? What is in these big repositories? What
does a legal workflow look like?

thanks,

arendt

P.S. Now that you mention it, I recall your unfortunate interaction with
the international banking system.

[sweetheart]

17. workflow databases

Consider a package of documents, photographs and other materials. This "folder" object is stored in a database and indexed via any number of indicies in a RDBMS. Then there is designated a series of queues and "roles" within the workflow system virtually shovelling this object around between "inboxes" based on decision criteria taken at the central workflow server.

All the inboxes are maintained in a huge superkey index hash: queuename: datetime: priority: folder-object: state-on-arrival Then select from this table depending on various restrictions in SQL gets you a queue management system... crude, but does the trick. I think the entire technology of this can be achieved by polling and RDBMS, you don't even need an application server arendt. The performance would be dreadful, but it would not be noticed until well after proof of concept... as it can be jacked up on big hardware.

i had a great interaction with the international banking system. I have been collectively fucked in my career in financial integration by carlisle group and bushishta election thieves inc. I accept my lot as just another fucked citizen, one who gets MORE flack as an american abroad as all yanks are presumed to be bushta's once the accent is heard.

I know how to design a much more stable international depository and exchange framework than they use today, by integrating the monopoly of the "identity" function to the regulator. Then the legal authority is part of the legal transaction to shift 100, billion pounds.... the soeverign needs to be involved in the transaction, as the ultimate repository and guarantor of the reference data. Companies act taken to its IT extreme... the online management of the monoplies chartered as legally sacrosanct public data... and their transactions.

Databases are soooo large now, that all the transactions of all the businesses in the world would fit on one mainframe computer. By legistlating this computer to be legally absolute software, a new evolution of legally absolute knowledge can be evolved much more effective than current paper systems.

I'm a believer arendt... just people do not realize what software is capable of.

[rogerashton]

3. Software engineering is a myth.

None of its objectives have actually been accomplished. But it does provide employment opportunities for a number of my colleagues.

[arendt]

4. so is a balanced budget, but both are desirable.

I already dealt with that in my original post:

> SWE created the potential for vastly more reliable software, although the
> majority
of programming organizations today still operate under the myth that really good
> programmers are only slowed down by SWE.

The fact that it is not used due to "real men don't eat quiche"
programmers, organizations too cheap to train their people,
and managers afraid of losing arbitrary control, in no way
invalidates the concepts used in SWE.

I'm assuming that only a programmer would make such a statement.
Then, you must understand the entrenched culture that SWE is up
against.

Have you ever read "Design Patterns"? Do you use any Object
Oriented languages? If yes, then you are already using some SWE
ideas, whether you admit it or not.

Besides, I asked for responses from lawyers, not programmers.

arendt

[GAspnes]

5. bookmarked for later (n/t)

[calm_blue_ocean]

6. Response from a DU (Patent) Lawyer

I am not sure that I understand all of what you are saying, but here are some comments:

1. VAGUE ON PURPOSE -- JUDICIAL DISCRETION

In some cases the law is purposely vague to some extent because human situations contain more potential relevant dimensions or axes than we could ever deal with ahead of time.

Example: sentencing guidelines for convicts

Some people (like Ashcroft) want sentencing guidelines to be firm and determinative, like an equation with a limited number of inputs.

Other people (like Rehnquist) want a given judge to have lots of discretion to sentence within a wide range on a case by case basis. These people argue that no sentencing statute could ever fairly capture the variety of relevant circumstances that a judge would consider in her cases. Of course, a judge may be unfair, but the question is: which is generally fairer: (1) a judge who can hear all the evidence of a convict's goodness and badness; or (2) a predetermined statute with a limited number of predetermined inputs that may or may not fit well with a given case?

Another example, from copyright law: "substantial similarity"
Let's say that I want to remake Star Wars, but change just enough things in my movie to avoid copyright infringement. (For example, they did this with the successful movie "SpaceBalls.")

Copyright law would handle these potential Star Wars lawsuits (there was no suit over SpaceBalls that I know of) by saying, rather vaguely, that the question is "substantial similarity" and then letting the judge and/or jury decide on substantial similarity.

I cannot imagine a statute (object oriented or otherwise) that could do this job better. How would you even write a statute that could take into account all the similarities and differences between two movies and provide a determinative answer about exactly how similar you must be before you have infringed Star Wars.

A statute might list some factors to consider, such as similarities in characters, similarities in plot, differences in tone (eg, parodies, like Spaceballs). In fact, cases do build copyright infringement law in this limited way. Still, I don't think we could come anywhere near the precision of a fractal here and I don't think it would even help to try (but be my guest, if you would like to).

This kind of purposeful vagueness is all over the law, and it is probably unavoidable when you start considering that the law deals with human interactions, which are not generally predictable by or reducible to algorithmic certainty (sorry Eliza).

2. CONTRACTUAL GOBBLEDY-GOOK

Contracts can be written in plain, understandable language and they often are. I don't think anybody would have trouble understanding a typical apartment lease. Warranties (think of a power train warranty on a new car) are pretty boring to read, but I think the average person usually understands them if she bothers to read.

Insurance contracts have complex language, but they don't let you read those and they are too long to read anyway.

I have come across surprisingly few people who have run into real-life problems with contractual gobbledy-gook language. Usually the problem is more like:

(1) "I bought this new car and it is a lemon -- what can I do?"

rather than

(2) "I bought this new car and thought this language in the warranty meant one thing and now that there is a problem the car company argues that it means something else and now we have a dispute because the language is so legally arcane."

I know the type (2) problem can and does happen, but I think it is kind of rare. If you want to see an actual case where the type (2) problem really happened, then GOOGLE "Tasini v. New York Times." I just don't think the type (2) problem happens often enough to get too steamed up about it.


3. STORY FROM LAW SCHOOL

On a law school exam, I once suggested a revision to the way juries operate. I suggested that instead of voting "guilty" or "not guilty," that each juror would give a confidence level. For example, "I am 85% confident that Mr. Simpson is guilty." Under my proposal, legal guilt or innocence would be determined based on the aggregate confidence levels of the 12 jurors, rather than on unanimous yes or no voting.

IF YOU EVER GO TO LAW SCHOOL DO NOT SUGGEST THIS ON AN EXAM! YOU WILL GET A BAD GRADE! I found out the hard way.

[arendt]

7. Thanks for the feedback

> 1. VAGUE ON PURPOSE -- JUDICIAL DISCRETION

I think I made space for judicial discretion in my discussion.
It is the job of the judge to balance conflicting laws. You point
out that the law could be settled, but not the punishment. Fine,
I will settle for the clear law.

You have pointed out that sentencing/penalties is an
integral part of the legal process.

> 2. CONTRACTUAL GOBBLEDY-GOOK

It is useful to know this does not come up in court very often.
But, I think it might come up in legislation all the time.

Finally, the fact that you offered no indication that any kind
of SWE process is even being considered is evidence that
my proposal is not already under consideration.

Thanks

arendt

[calm_blue_ocean]

9. You might want to take a crack at . . .

writing the copyright infringement statute according to your methods. I think that could be a useful vehicle for further discussion.

Like basically a computer program that takes two text samples and decides if there is copyright infringement or not. I think that small bit of statutory coding might show you how difficult it is to encode even a small, simple piece of the law so that the results are fair.

Without a sample "SWE" program, it is kind of difficult to visualize what you are planning.

Suggested further reading: CODE and Other Laws of Cyberspace by Law Professor Lawrence Lessig

[arendt]

10. Good example, but I need some help

My point about the law being fractal is that legal
structure builds up by iteration.

So, initially, there was some vague statement about
copyright. Then there was a case which refined that
statement. Then another case, another refinement...

If I had someone expert in copyright infringement,
he should have some knowledge of the iterative
branching structure of the law, as it evolved; sort of
a phylogenetic tree of copyright infringement law.

Using such knowledge, I could take a crack at an
inheritance hierarchy and some compositional objects
for this field. Perhaps all the relationships among
objects in this field could be codified into an ontology.
(E.g., copyright life depends on author's date of death;
satire is an exception to copyright).

With such a detailed hierarchy of precedents, I can
try to fit the features of the case into the hierarchy.
Of course, if the case is unique, it won't fit cleanly
into the existing hierarchy. In that case, I'll have to
split a class or add another component to the composite
object. But, those kinds of enhancements to the strucutre
are exactly the kind of things judges are there to do,
when they create new precedents that do not violate,
but simply extend old precedents.

I'm not proposing anything new. I'm just proposing
formalizing what is already done and doing it in
a modern language with automatic type-checking
to flag troublesome areas.

The point is, there is a lot of domain specific syntax
and semantics to formalize. Since lawyers can do
this in their language, it must be possible to restate this
in another language.

----

thanks for the specific lessig reference. I am certainly
aware of Lessig's good work on intellectual property
in cyberspace.

arendt

[calm_blue_ocean]

13. Oh, that's very different

1. Court Cases do build iteratively on statutes to fill in the interstices of the law. The indexing of cases by points of law contained therein was numerically formalized by the early 1900s by the West Company Key Numbering system. I am glad that our (pre-computer) legal research methods have proved to be so helpful in your nascent field :)

2. You seem to labor under the impression that it is hard to find precedential court cases. It is not. Some people still use the West Key Numbers (on computer these days).

I generally prefer full text searches. This is because West's key numbers are only as good as the indexers themselves and the system of limited accountability in which they toil.

Full text searching actually works a lot better in a database of legal case than in it works for, say, searching the whole web, with Google. This is because of the special vocabulary used in the law. In my SpaceBalls example, I would search:

copyright and "substantial similarity" and (movie or film)

In this example, the legalistic phrase "substantial similarity" gets rid of most of the crap from my search results, while keeping the good cases in.


The problem:

The databases, like West's and LEXIS, are very expensive to use. This is not because judicial opinions are copyrighted -- they are not. Rather, this is because LEXIS and other database maintainers require that extremely high prices be charged for access because the databases are expensive to maintain. I do not think that your fancy SWE methods would make databases less expensive to generate and/or maintain. They would probably do the opposite.



FINAL NOTE
==========

I am not trying to discourage you from thinking about these problems. As a patent attorney, I believe in technology and I know that somebody, someday will pioneer better ways of getting people the relevant case law they need for cheaper. I hope its you!

I just want you to have some idea of the prior art of legal databases and of the number of variables potentially involved in a typical legal decision. When you think about it, there are a lot fewer decisions than there are potentially relevant variables!

[arendt]

14. Exactly the kind of informative response I wanted, thanks

> I just want you to have some idea of the prior art of legal databases and of
> the number of variables potentially involved in a typical legal decision.
> When you think about it, there are a lot fewer decisions than there are
> potentially relevant variables!

That's exactly why I asked DU lawyers to bring me up to speed.
Now I have West's as a start point for googling.

And, I completely agree about #vars >> #decisions. That is why I
characterized the situation as fractal.

> The databases, like West's and LEXIS, are very expensive to use. This is not
> because judicial opinions are copyrighted -- they are not. Rather, this is
> because LEXIS and other database maintainers require that extremely high
> prices be charged for access because the databases are expensive to
> maintain.

Everyone knows that LEXIS is ridiculously expensive, no matter what
you use it for. Unless the data is private, Google can't be beat.

I wonder how much it would cost to de-privatize the legal citation system?
I don't mean steal it, but rather take it over as eminent domain, and pay
the owners something like 10% over the current stock price. This is
really legal infrastructure. Of course, we no longer believe in "common
carriers" or "regulated monopolies" or any of that disredited old "good
government" stuff, so this is just a pipe dream.

> You seem to labor under the impression that it is hard to find precedential
> court cases.

Where do you get that? I said its easy for both sides to flood the judge
with precedents.

----

Anyway, thank you for playing

arendt

[calm_blue_ocean]

16. Answers

1. De-Privatizing cases ("opinions"):

a. There are free databases that include limited sets of cases. Sometimes these are word searchable and sometimes not. If the cases are word searchable, then GOOGLE will get you to the case. The problem with these free databases is that they seldom include in a database all the courts and/or time periods that you need to find before you go to court (courts have specific rules about which opinions have precedential value and which opinions do not -- this varies by court). In the law, a partial search can actually be worse than no search at all!

b. as far as I can know, the main obstacle to having a new company set up an inexpensive database with access to a meaningful universe of cases is the fact that WEST has historically claimed copyright in the page numbers in the database. You heard me correctly, the PAGE NUMBERS (but not the text) is copyrighted (See, LEXIS v. West). More specifically, West claims the association between each piece of (non-copyrighted) text and its page number is copyrighted. As far as I know, this is why no private company has simply scanned all the law books into a single database.

c. despite the above mentioned difficulties, I am shocked that there seems to be no comprehensive, inexpensive case search tool. Shocked and appalled!!!!

Even the free US Gov't patent database works pretty good and is pretty comprehensive and very searchable at this point! Why is case law database practice so far behind?


2. Difficulty finding precedents

I did not speak precisely. Sorry about that. I meant that you seem to labor under the impression that it is difficult to isolate the most relevant and authoritative cases from the large universe of court opinions.

This finding a couple "needles in a haystack" problem is part of legal research and part of legal work. However, it is not a particularly big or costly (other than LEXIS charges) part of the cost of litigation. Law firms often have clerks or new lawyers (with lower billing rates) do this work of isolating the most relevant cases. The logic here is quite fuzzy, but intelligent, young law students will exhibit startling convergence when given similar research tasks independently of each other.


3. Ninety Degrees Off Topic

BTW, Your big costs in litigation relate more to the discovery process (deposition, requests for documents, motions for confidentiality and privilege, expert witnesses, interrogatories). This discovery does not have very much to do with case law. Basically, case law provides the legal standards and discovery provides the facts for the current case at issue.

Think about the legal proceedings surrounding Bill Clinton. We all knew the impeachment legal standard ("High Crimes and Misdemeanors") -- that legal research was trivial! What costed all the time and money was figuring out whether there was a blue dress with his DNA on it. That took quite a few "Motions For Depositions" (and oppositions thereto). Most modern litigation has similar "discovery" costs and they are huge.

[arendt]

18. Very interesting

> (courts have specific rules about which opinions have precedential value
> and which opinions do not -- this varies by court). In the law, a partial
> search can actually be worse than no search at all!

Very useful fact.

> the PAGE NUMBERS (but not the text) is copyrighted (See, LEXIS v. West).
> More specifically, West claims the association between each piece of (non-
> copyrighted) text and its page number is copyrighted. As far as I know, this
> is why no private company has simply scanned all the law books into a
> single database.

Ah, another misuse of the patent process. Why am I not surprised?

But, that is why I propose to eminent domain West's right out of business.

> Even the free US Gov't patent database works pretty good and is pretty
> comprehensive and very searchable at this point! Why is case law database
> practice so far behind?

Exactly my point. And, in the patent area, besides uspto.gov, there is the
IBM patent site (if you don't mind they are spying on you).

> intelligent, young law students will exhibit startling convergence
> when given similar research tasks independently of each other.

This supports my conjecture that it is quite probable that a human-
created object/ontology hierarchy could successfully formalize
case law. Please note, I said "formalize", not "automate".

> Your big costs in litigation relate more to the discovery process

I assume this is in reference to my argument that SWE formalization
would reduce legal costs. Again, thanks for the useful domain-specific
knowledge. My proposal is strictly limited to formalization. Perhaps
this fact would say that formalization will not solve the problem in
the legal/court system. But, again, my real goal is to clean up the
legislative process.

Once again, your input is most valuable.

arendt

[calm_blue_ocean]

20. Good chatting with you

I find these subjects very interesting, as you can probably tell.

The discovery discussion doesn't really relate to your plans.

My point in bringing up "discovery" is just to let you know that the legislation and legal research area that you are dealing with may not be as large of a component of legal expenses, or as much a part of a typical lawyer's work, as you might think.

This is probably why lawyers and their clients have put up with overpriced LEXIS and Westlaw (and their silly COPYRIGHT (not patent) claims) all of these years. Overpriced as this stuff is, it is all often dwarfed by the discovery-related line items on a legal bill.

As far as legislation, it ain't the cost of legislation -- its the cost of lobbying. Politicians can be even more expensive to buy than are lawyers.

[arendt]

21. Lobbying is the target of my new Constitution

> As far as legislation, it ain't the cost of legislation -- its the cost of lobbying.
> Politicians can be even more

By created literally hundreds of legislatures, and a total of over
20,000 elected federal officials, and by distributing said legislatures
in 50 or more cities, my goal is to make it simply unaffordable for
lobbying as currently practiced to continue.

Right now, there are roughly 80,000 lobbyists; and they target
the UNELECTED Congressional staff. By essentially making the
staff elected officials and bringing their dealings with lobbyists
into the daylight, I expect that lobbying will have to change.

First, I expect that lobbyists try to use their time efficiently by
covering many bases on their trip to the one Capitol. But, if
the committees are spread out in different cities, it makes it
much less efficient to keep lobbyists on payroll in all those
places.

Furthermore, right now there is an average of only four lobbyists
per staffer. In my plan, there will be roughly 100,000 voters per
elected official. Surely those 100,000 ought to be able to keep four
measly lobbyists under control. (If you don't understand how
I can have 20,000 legislators each having 100,000 votes,
you should read the thread on NK government at redeye's
VOY site.

Regards,

arendt

P.S. This is one of the most polite and non-confrontational
exchanges I've had on DU in quite some time. Thanks.

[redeye]

11. Your jury suggestion

The problem with every juror giving a confidence level is strategic voting - if I think C is 20% likely to be guilty, then I have the incentive to say he is certainly innocent, because this way my vote will have a greater power. You can use the confidence idea when you're dealing with impartial judges - judges, justices, even judges in athletic games - but not when you're dealing with juries, elections, and that kind of stuff.

[calm_blue_ocean]

12. Strategic voting

My perception is that under the current system of yes or no, unanimous voting, then voting is even MORE strategic.

What I mean is that if twoor three jurors are holdouts, then there is a lot of pressure routinely brought to bear on these people to change their votes. Pressure by the other jurors, who want to go home. Pressure by the judge, who wants to avoid a 2d trial. In this case the voting is strategic in the sense that it is based on pressures that come from outside the realm of actual evidence in the case.

It seems to me that confidence levels would greatly decrease strategic voting. This would be especially true if the jurors were not allowed to know how various confidence levels were actually translated, mathematically into a verdict. In other words the jury doesn't get to know what aggregate confidence levels amount to "guilty," "hung," and "not guilty." In fact, I would argue that jurors should be allowed to keep their votes secret to avoid the pressures we see now (can't really do that under the present system).

Also, mathematical curbs could be put on strategic voting. For example, a vote of zero confidence could be made into a vote 20 confidence (after all, the case hasn't been thrown out for frivolousness by the judge).

Ultimately, I no longer think that my confidence levels scheme should be implememnted. However, it is a nice example to scare people with. People usually get pretty uncomfortable when the messy humanity of the legal system is reduced to so much math.

[Ivory_Tower]

15. Forget law! I want to see SE applied to government!

Seriously, I'd love to see a class diagram of the executive branch (well, all the branches for that matter). I have a feeling we'd see a really inefficient design that could use some re-structuring.

(Note: I think "class diagram" is the correct term -- I'm mainly a flat C programmer, and don't have a lot of experience with Object-oriented design. Some, but not much...)

[arendt]

19. Well, that's really my final goal, if you study my post

The correct term might be "class hierarchy" if all laws were
subsets/supersets of each other. But, it is probably true
that laws are "compositions" of several independent component
classes. Some components are found in some laws, but
not in others.

So, yes, I want to apply SWE to the government.

Maybe the first thing to do is to try to turn the original US
Constitution into objects.

arendt

[GAspnes]

22. very interesting idea

I have no proof that this could be accomplished, but let me add a couple of data points to the discussion underweigh.

Being easily bored, last year I tried to write a 'law parser'. My intent was to feed the USC into the parser and build a lexical tree of the laws. My goal was to find conflicting laws.

I never succeeded in getting what would be considered a "good compile" of the United States Code. Terms mean different things in different sentences, which resulted in associated sentences which, when examined by a human (me) didn't really relate. I never made a decent taxonomy of even a portion of the USC.

My ultimate goal was to try to collect case law analyses and decisions to refine your 'fractal interface', so that a set of facts (written in standardized language) could be fed into the engine and return a set of relevant laws and decisions. My failure at step 1 put the kibosh on that.

I estimated what it might take to do a human/machine conversion of the USC to a set of standardized terms. Given the growth in the law, my estimate was of the close order of 80 years. Bah. This result led me off into the lexical analysis swamp for a month. This is the 'human language translation' problem -- a highly intractable problem that hasn't had any significant progress made against it by people far smarter than I.

So I branched off into the AI aspects for a while. Perhaps some kind of human knowledge base. I checked into the Cyc project, and they are doing some interesting work with 'common sense'.

For example, suppose a user asks the GeoAgent for "elected heads of government of countries north of the equator". This might be represented as:

(#$and
   (#$headOfGovernmentOf ?x ?y)
   (#$hasAttributes ?x #$Elected)
   (#$northOf ?y #$Equator))

The GeoAgent is able to find bindings for the third clause by using its own knowledge of the geography domain:

Britain is in Europe.
Europe is in the northern hemisphere.
The northern hemisphere is north of the equator.
If region A is part of region B, and region B is north of region C, then region A is north of region C.
Therefore, Britain is north of the equator.

This is a declarative rules-based language (like Prolog), and again we're back to the translation problem.

I think you're on the right track here (although I could go 15 rounds with you on whether or not there is such a thing as Software Engineering and, if there is, whether it does what you think it does), but I get the gist of your idea and I think it's a good one. It's just not going to be easy or simple.

Oh -- I ran into one other problem during my project: the Law of Unintended Consequences. Even very simple computer programs compile to binaries which are only human-readable if you're a very strange human. Large programs are essentially un-understandable once compiled. A concommittant of this complexity is the inability to ever confirm that a program is going to do exactly what you expect in every case.

In this regard, the law (code and case) is a computer program, executed on a machine of vast complexity (the legal system, encompassing clients, lawyers, judges, court clerks, police, government). The only thing that keeps it from being completely arbitrary is the fact that humans participate at every step of the process and the ridiculous, while a well-known occurrence in the law, is not a common one.

[calm_blue_ocean]

23. Let me know if . . .

you ever need a patent attorney.

I am out in the desert, so I am relatively cheap.

[arendt]

24. You must be computer-literate and legal-literate, you will go far.

PM> Hope you find my comments helpful.

Yes, indeed. You are either a most computer-literate lawyer or a
computer science/compiler guy interested in the law.

You have given me at least one new pointer (i.e., the USC);
but more importantly, you have already taken a shot at this.
Can you tell me a little bit about your parser? What language
(C++, Perl, ...) or tool (yacc,...) did you use? If it was a table
driven parser, how big was the table? Did it parse correctly,
but the result was like the old computer translation joke:

English: "Out of sight, out of mind."
Russian: Blind maniac

My personal opinion is that Douglas Lenat (the author of cyc)
is the last diehard of the old Hard AI school. The whole idea
that intelligence is done purely by formal logic is a long-disproved
canard. Still, that school keeps coming up with new wrappers
for the same old stuff. Their latest is "ontologies", where the
logic is hidden in the fact that you are still traversing graphs
or "relationships", which relationships contain the real world
"knowledge" that this "knowledge builder" is aware of.

Ontologies have their uses, however. The one I am most
familiar with is GO (genome ontology), which is a repository
for all the known metabolic and control pathways of the
human genome and proteome. Not only is this an ontology,
it is updated on a regular basis by an AUTOMATED RECOMPILE
of the entire ontology in order to restructure the pathway
graph on the basis of newly discovered info about the pathways.
In this case, ontology works because the domain it is mapping
actually is a graph, the graph of metabolism.

From what you say, I'm not so sure the law is a graph (or at
least not one simply connected in two or three dimensions).

> Large programs are essentially un-understandable once compiled.
> A concommittant of this complexity is the inability to ever confirm
> that a program is going to do exactly what you expect in every case.

This is merely an instance of the Turing stoppiing problem or Chaikin's
algorithmic complexity work, both of which play off of Godel's theorem
about complex logical systems. At some point, rationality reaches
its limits. I find that many scientists refuse to include the Godel/Turing/
Chaikin PROOFS in their worldview, but rather cling to 19th century
faith in the ability of logic to solve all problems.

I have no problem with the fact the program might do something
unpredicted. The theorems show that no one could do any better.
Rather, I look at these unpredicted results as opportunities to
discover inconsistencies in the existing program.

> the law (code and case) is a computer program, executed on a
> machine of vast complexity

Thanks for saying that. It really helps to know that other people have
similar thoughts. Sometimes, I think I'm so far out that I ought to just
check into the nearest funny farm.
Anyway, thanks for a most interesting response.

arendt

[GAspnes]

25. it was an interesting failure

I wrote the parser in C. I was learning about MySQL at the same time. The parser needed many exceptions, and there was a problem building the dictionary as well, especially with sentences like this one:

Sec. 1. - Trusts, etc., in restraint of trade illegal; penalty

Every contract, combination in the form of trust or otherwise, or conspiracy, in restraint of trade or commerce among the several States, or with foreign nations, is declared to be illegal.

I could find no simple way to determine what 'otherwise' should link to, and the parser would have problems with nested 'or's; there is also apparently a large difference between the legal definition of 'commercial conspiracy' vs. 'criminal conspiracy'. Determining which is which required real-world knowledge (outside of the USC), which is what led me to the Cyc project.

There's no doubt I'm a better computer jock than lawyer (which isn't saying much).

I've put this problem on the back burner while I work on other things, but I'm sure it will pop up again soon.

on edit: as much as I think using OO techniques (or something else -- formal methods, e.g.) to define laws so that they are clear and precise, I suspect that most normal humans prefer the current ambiguity. We're a race of arguers; just look at the debates over whether a pitched ball was a strike or not. A clear and accurate deliniation of laws is *not* what most people might want, because there is no wiggle room for circumstance, intent or a face-saving way to get out of a bad situation. People and their interactions are complex.

This doesn't mean that I don't think this is a good idea. It is. I just think there are wider implications than just the technical. That was why I went after analyzing the USC first. If one could show that it is inherently self-contradictory in places, or that similar offenses are not dealt with similiarly, that's a lever to help change perception.