I'm confused by why JWZ says that the Mork database code has anything to do with killing people. It is not presented as a product to be used for the collection of forensic evidence. Every user agrees to terms of use that suggest it is not a mission-critical piece of software to be used for sending a manned mission to mars or saving the location of your buried ingots of gold.
Relying on its data to convict someone of a capital offense says more about the justice system then it does about the software, and if an alleged killer should go free because of its unreliability, that says more about the inability of the police to gather sound evidence than about the inability of its programmers to build software for the unstated requirement that it catch criminals.
>I'm confused by why JWZ says that the Mork database code has anything to do with killing people.
Here's the progression of thoughts.
1) Messed up database format is difficult to decipher.
2) Difficulty in deciphering the format, leads a forensic analyst to mistakenly conclude that the site was visited 84 times instead of just once.
3) Assume that Jury reaches conclusion that Casey Anthony killed her daughter using chloroform. Thus the choice is then whether it was simple manslaughter and she gets a few years in prison, or it was premeditated murder, and thus eligible for the death penalty.
4) Since the suspect visited the site 84 times, this helps lead to the jury concluding that the killing was premeditated murder, and thus Casey Anthony should be put to death.
5) Thus, as the result of a bad database format, someone was put to death instead of just serving a few years in jail.
I agree with your analysis of the argument being presented, however I disagree with the argument itself.
1) Street light burns out but hasn't been replaced yet, making street dim
2) Dim lighting leads bystander to mistakenly conclude that accused was present at scene of crime
...
5) Thus, for the want of a light bulb, someone is put to death.
My contention is that it is a failing of the process for the Jury to reach the conclusion based on a forensic analyst's mistake. I'm not an expert on courtroom proceedings, but I certainly hope that if I am ever asked to give expert testimony in court, I will not look the jury in the eye and claim in loud, ringing tones that I am 100% confident of my findings when spelunking through a database format I didn't write that was written by code I haven't examined line by line.
If I was foolish enough to make claims like this, I hope someone would cross-examine me properly. I hope that someone, somewhere would suggest that there might be any number of reasons why the word "chloroform" might appear on the hard disk in that spot and give alternate explanations. And so on, and so forth.
If people are being sent to Death Row based on what looks to me like incredibly scanty evidence, I don't think the scanty evidence is the problem, I think the process for evaluating evidence is the problem.
Unfortunately law seems to be one of the most conservative institutions (i.e. slow to change) in existence. It it set up to be an adversarial system between two parties, and most of the institutional structure depends upon that. Between the desire for an adversarial system and constitutional right to trial by jury, I'm not sure how you can really move away from my expert witness vs your expert witness when examining evidence for a jury. Any ideas? A "technical judge" that adjudicates evidence into more explainable forms to be explained to the jury?
Conservative (risk aversion) and slow to change are not the same thing. By being slow to change law and lawyers introduce excessive risk in their offerings which, ironically, makes them negligent at their stated purpose.
Your "I hope" sections are exactly how it happens. The amount of preparation that goes into a trial like this is ridiculous, and the cross-examiner will have a host of questions that the expert witness cannot answer with any certainty.
In less media-saturated trials which have the same stakes for the participants, expert witnesses can say whatever they're paid to say. They're rarely challenged seriously on their testimony, or even their credentials.
The way the adversarial judicial process works is that lawyers hire expert witnesses based on their ability and willingness to look the jury in the eye and claim in loud, ringing tones that they are 100% confident of their findings.
Reliability. It is sometimes thought of as correlating to how much weight a particular data point may have in making a decision. In this case, the data is not very reliable but a seems to be evidence crucial point (the intent demonstrated by earlier information gathering). It speaks to a particular case of a justice system presenting evidence without reliability meta-data.
Also, computer forensics is ALL ABOUT dealing with fucked up data in bizarre formats (if the design of the format doesn't make it bizarre, the baffling habits of its users and provenance of the data will).
True, but isn't it important as an expert witness to state the degree of confidence you have in the evidence you give? I suspect the problem here is that the US system of having high-priced professional expert witnesses gives the witnesses a massive incentive to overstate their confidence.
"Oh, yes, absolutely she did a Google search for chloroform 84 times, I'm certain of it, I'd stake my reputation on that."
Who says that in a murder trial and sleeps soundly at night?
Definitely. But in my experience (I used to work in this industry) the reality is much sketchier than most of the industry lets on.
There are long chains of subcontractors involved, and there were many points where I or any other low-paid employee (at the time, I was making $12 an hour, though my time was being billed out to the client at $150 an hour) could've easily tampered with or falsified data (and no, I never actually did this), despite a lot of security theater and data integrity theater.
Welcome to my job..... and this is only the tip of the iceberg. Mork is one of many painful "things" that no one seems to be able to produce a decent forensic tool for :S
Looking at that format, no wonder JWZ had trouble getting a list of recently visited URLs.
Quite a contrast with the present-day. A few months ago, I wanted to extract just that information, and despite not knowing a lick of SQL, I still managed to write a little script which would extract 'all URLs visited in the last month': http://www.gwern.net/Archiving%20URLs#local-caching
16 comments
[ 3.4 ms ] story [ 41.6 ms ] threadLooks like David McCusker is the original author and it was only supposed to be a temporary solution that lasted until 2009.
http://web.archive.org/web/20050315212725/http://erys.org/re... explains mork being an abstraction layer on top of MDB.
Also http://web.archive.org/web/20050324235032/http://erys.org/re... for some general info linked from the mozilla.org link.
Relying on its data to convict someone of a capital offense says more about the justice system then it does about the software, and if an alleged killer should go free because of its unreliability, that says more about the inability of the police to gather sound evidence than about the inability of its programmers to build software for the unstated requirement that it catch criminals.
Here's the progression of thoughts.
1) Messed up database format is difficult to decipher.
2) Difficulty in deciphering the format, leads a forensic analyst to mistakenly conclude that the site was visited 84 times instead of just once.
3) Assume that Jury reaches conclusion that Casey Anthony killed her daughter using chloroform. Thus the choice is then whether it was simple manslaughter and she gets a few years in prison, or it was premeditated murder, and thus eligible for the death penalty.
4) Since the suspect visited the site 84 times, this helps lead to the jury concluding that the killing was premeditated murder, and thus Casey Anthony should be put to death.
5) Thus, as the result of a bad database format, someone was put to death instead of just serving a few years in jail.
1) Street light burns out but hasn't been replaced yet, making street dim
2) Dim lighting leads bystander to mistakenly conclude that accused was present at scene of crime
...
5) Thus, for the want of a light bulb, someone is put to death.
My contention is that it is a failing of the process for the Jury to reach the conclusion based on a forensic analyst's mistake. I'm not an expert on courtroom proceedings, but I certainly hope that if I am ever asked to give expert testimony in court, I will not look the jury in the eye and claim in loud, ringing tones that I am 100% confident of my findings when spelunking through a database format I didn't write that was written by code I haven't examined line by line.
If I was foolish enough to make claims like this, I hope someone would cross-examine me properly. I hope that someone, somewhere would suggest that there might be any number of reasons why the word "chloroform" might appear on the hard disk in that spot and give alternate explanations. And so on, and so forth.
If people are being sent to Death Row based on what looks to me like incredibly scanty evidence, I don't think the scanty evidence is the problem, I think the process for evaluating evidence is the problem.
http://www.guardian.co.uk/uk/2008/mar/23/ukcrime.law
http://www.innocenceproject.org/Content/Cases_Where_DNA_Reve...
http://www.nytimes.com/2008/08/12/us/12experts.html?_r=1&...
I don't understand what you mean by this.
Isn't "data" just another form of evidence, whether it's your browser's history file or bank account records or cell phone history records?
Why is one type of data less worthy than others?
"Oh, yes, absolutely she did a Google search for chloroform 84 times, I'm certain of it, I'd stake my reputation on that."
Who says that in a murder trial and sleeps soundly at night?
There are long chains of subcontractors involved, and there were many points where I or any other low-paid employee (at the time, I was making $12 an hour, though my time was being billed out to the client at $150 an hour) could've easily tampered with or falsified data (and no, I never actually did this), despite a lot of security theater and data integrity theater.
Quite a contrast with the present-day. A few months ago, I wanted to extract just that information, and despite not knowing a lick of SQL, I still managed to write a little script which would extract 'all URLs visited in the last month': http://www.gwern.net/Archiving%20URLs#local-caching