Kesteven
 |
Logical Law
2008/05/20 18:45
|
Karma: 12
|
Admin |
|
I posted this in another forum, but thought I'd also leave it here just in case anyone had anything to say.
I was talking to a friend of mine about the problems of juries being vulnerable to emotional rhetoric, and he suggested I write a script in prolog to replace them. I told him it was a stupid idea because it is, but nonetheless, I found it worth considering.
Firstly, if you were creating a program to function like this, you'd want it to say, of an individual, if that individual was guilty of a specific crime. It would do this by determining if the individual had a set of properties sufficient to entail guilt of that crime.
Now, in terms of inputs, you'd really need two main kinds.
Firstly, you'd need to program in the conditions of guilt of a crime. This is the easy part, but still pretty much insanely difficult. In theory though it's possible, on the assumption that all laws can be phrased in logically unambiguous language which reduces to verifiable properties. The problem is that most laws are, at least in the vulgar, extremely rubbish. For instance, the OED defines murder as 'criminal homicide with malice aforethought (occas. more fully wilful murder); an instance of this.' and 'murderer' as 'A person who murders another'.
Breaking this down: For x to be guilty of the murder of y, x must have murdered y. For x to have murdered y at time t, the following conditions are both necessary and sufficient:
1) x must have been a human being (at t)
2) y must have been a human being (at t)
3) x must have killed y (at t)
4) x must have intended to do harm (to y) (before t)
5) The killing of y (at t) must have been a criminal act, qua being an act of killing.
Now some of these are pretty troublesome. Notably 5, which I'm assuming is an ellipsis for the ACTUAL legal definition of murder, because if not, god help us all. In theory though, this is all do-able with enough logicians and lawyers and sufficient time (even though the concentration of bloody-minded heartlessness would have probably created a sphere of negative energy killing every puppy in a 10 mile radius by the time anything was achieved).
The second part of the equation however is determining, based on the evidence, whether someone meets the conditions required for them to be guilty of the given crime. This is the part that would give pretty much any computer engineer a brain aneurysm. In theory, of course, it's possible, otherwise jurors, who are essentially just very advanced (or at least very complicated) computers, wouldn't be able to do it. The problem is the vast quantity of supposition, assumption and intuition that has to go into even relatively simple cases. The only way I can see to implement this in an even somewhat workable system is to have a simulation run through every possible scenario (given the constraints fixed by the evidence and suitable background information)(probably couldn't do this in prolog) and determine in how many of them the accused meets conditions for murder. That would then be cross-referenced with a system which weighted each scenario for likelihood, presumably by further background material, and then converted the resulting information into an estimation of the probability that the accused is, in fact, guilty of the crime of which they stand accused. 'Reasonable doubt' would, in this case, have to be reinterpreted as a certain figure, probably around 80%.
The scenario system would probably be the most troublesome part of all. Realistically, it would have to be able to render real world scenarios in terms of significant parts, since it would be impossible to simulate it in full without running through every conceivable atomic composition of the universe during the required period. Determining which parts were significant and creating a system that could place them in a series of scenarios would be a task for several generations of geniuses, and even then this kind of 'brute force' method would need phenomenal quantities of processing power.
to reduce the load to more plausible levels, the job of narrowing down scenarios could be delegated. The job of the solicitors is of course to provide the jury with the most plausible scenarios which favour their case, so if we're purely emulating the jury, we wouldn't need to do nearly as much. Just run each suggested scenario and correlate it with the evidence and some background assumptions. Of course, each scenario will have a large number of variables, and working through the implications of each of these would still be an incredible task. Creating and implementing an appropriate set of background assumptions, an even more incredible task, perhaps even insurmountable. However, concentrated intelligent scrutiny and thorough testing over an extended period might be enough to accrete a workable set. One might also prefer to have a collection of machines, each running on a slightly different set of assumptions, cumulatively representative of the common (or common intelligent) view. |
|
Log in to post messages and access additional forums.
|
Dan J. Brigham
|
Re:Logical Law
2008/11/11 17:50
|
Karma: 3
|
Junior Philosopher |
|
Condition (4) "x must have intended to do harm (to y) (before t)" is too wide scoping. It needs further qualification.
Here's a counter example:
I (Mr x) have an argument with my neighbour (Mr y). I throw a punch at Mr y, intending to really hurt his head and teach him the value of my 18ft Connifer. Unknown to me, Mr y has a particularily weak skull. My single punch, which was intended to hurt him BUT NOT KILL HIM - I'm no monster - has now just killed him.
Conditions (1) - (5) have been met, however this would be a case of manslaughter not killing.
Perhaps it could be re-worked as the following:
(4') x must have intended to kill y (before t)
Although even this might create problems. Wanting to kill someone in the past, then changing your mind, then accidentally killing them, for example.
I don't want to sound like I'm a pedant or anything but you are in the Logical Analysis section. |
|
Log in to post messages and access additional forums.
|
Peachy
|
Re:Logical Law
2008/11/12 06:59
|
Karma: 3
|
Experienced Philosopher |
|
I think the most pressing problem is one you skimmed at the end of your piece. What counts as 'the evidence'? What should be the background view? The programmer himself is responsible for inputting the background assumptions and plausible interpretations of character-types, which makes the computer, at best, him, sitting on a jury. even when we try to remove ourselves from a process, w come right back in through the tools that replace us. Now, there's not just the bias of the juror, there's the bias of the programmer, errors in hardware, overcompensating corrections and so forth. We shouldn't trust ourselves to not trust ourselves. |
|
Log in to post messages and access additional forums.
|
Kesteven
|
Re:Logical Law
2008/11/12 07:49
|
Karma: 12
|
Admin |
|
|
Oh, I certainly wouldn't want to implement this in real life, certainly not at this level of technology and constructor intelligence. But there are several reasons why the computer wouldn't just be the programmer sitting on a jury.
Firstly, the computer's logic, unlike the programmer's, is infallible, whatever else of it may be fallible, because its information processing is not emotionally-based or distributed. So is its memory; it will not forget or overlook any of the evidence provided to it. Of course, it will probably be called upon to discriminate evidence, in case, as is quite likely, faulty evidence has been input. But it will be able to draw attention to conflicting evidence and suggest the most plausible resolutions.
Secondly, and perhaps more importantly, there is no such thing as 'the programmer'. The system would be programmed, observed and administrated by teams of programmers, logicians, lawyers, statisticians, politicians and so on working in concert, and no doubt many of the distinctions would themselves be fixed by the democratic process and by judicial precedents. The system would also be constantly revised and improved by a range of different people (in a similar manner to Dan's demonstration) until it was in keeping with the intuitions of the vast majority (at which point it would become adequate for actual rather than merely experimental use).
And there isn't even any need for such a system to replace the traditional court proceedings, it would be a useful tool to use along with it. The system output could count as part of the evidence to be presented to the court. This would delay cases while the information was prepared of course, but I think it could prove indispensable for high-profile cases, and if the system was good enough, I imagine the human jury would become vestigial anyway.
This is the point where sabotage or political manipulation of the system would become particularly profitable.
Post edited by: Kesteven, at: 2008/11/12 07:53 |
|
Log in to post messages and access additional forums.
|
Peachy
|
Re:Logical Law
2008/11/29 09:25
|
Karma: 3
|
Experienced Philosopher |
|
|
That's very persuasive. Are you sure you wouldn't want this implemented? |
|
Log in to post messages and access additional forums.
|
|
Philosophy Forums 
