91–100 of 520 posts
Earlier quoted context omitted.
Yes, just like voting machines. But how do we confirm that the software which was vetted is the software that was actually _used_?
Compile it, run it with the same inputs the prosecution claims was provided, and see if you get the same result. That's sorta the point of computers, right? Same inputs on the same program, same output?
Earlier quoted context omitted.
The hands-off reputation-based approach isn't enough to regulate food and drugs, or aviation. I don't see why for-profit forensics companies should be trusted implicitly. Also related: we know that funding bias [0] is a real problem in science, despite that scientists' reputations should (ideally) motivate them to resist such biases. [0] https://en.wikipedia.org/wiki/Funding_bias
No, the question is what is the incentive for lying and claiming guilty? Do prosecutors choose different labs?
I would argue that if prosecutors and/or cops get to pick the tool, false positives would result in more sales.
Just as false negatives would result in more sales if the defendant gets to make the pick.
First, the defence doesn't necessarily have to evaluate all 170,000 lines. They just need to find one buggy line which could potentially overturn the result.
Second, even if it did take a full 8 years, is that a good reason to deny the defendant due process?
Earlier quoted context omitted.
> Forensic labs and companies are expert witnesses with black box processes and the incentive to protect the authority of their profession. I imagine they're also incentivised to favour a result indicating guilt.
Why? The risk to their reputation & livelihood for a false positive should temper that.
Their goals are always to get plea bargain because that is automatic guilt or if they take it to court, they'd better win. That's a lot of state resources just to lose.
Overtime, you'll get people willing to bend the rules, turn a blind eye to 'weak evidence' or whatever.
You're only as good as your measurement/incentivisation.
Earlier quoted context omitted.
My exact thoughts. This sounds like a classic example of launching a prototype created by domain specialists (biostatisticians and bioinformaticians) as production software and skipping on the expensive stuff, like sound development practices
There is so little emphasis in production software development in bioinformatics and biostatistics. Despite a lot of groups open sourcing their code it's is nearly unusable and not reproducible due to hard coding, ignoring edge cases, and dumping the majority of the code on a single giant R or python function.
Earlier quoted context omitted.
I'm not so sure we should be allowing challenges to expert witness, with no evidence. That sounds like a rich source of abuse of the system, like a big company filing thousands of briefs to delay a case. It's important for experts to use validated, tested equipment. That's clear. And by the sound of it that is happening (stories of how code was tested and found wanting confirm that; stories of double-blind testing).…
In most criminal complaints the defendant has far fewer resources than then the prosecution. The prosecution has a conflict of interest to get a conviction. Absolutely everything about their process needs to be above reproach, including their experts. A challenge to that process, approved by a judge, seems like a small consolation in a biased system.
On one side were 6 impeccably dressed men in what I presumed were multi-thousand dollar suits. On the other side was single overwhelmed, overworked, and under-dressed defender who was steam-rolled at every sentence she spoke.
I ended up in traffic court last year for a parking ticket that I can only think must have been stolen from my car during a festival. After spending just a few hours waiting my turn, it was pretty clear who exactly the police target for traffic violations. But this is a tangent.
Earlier quoted context omitted.
Yes, just like voting machines. But how do we confirm that the software which was vetted is the software that was actually _used_?
At some point you just have to trust people to not intentionally lie in a trial. We have very stiff penalties for anyone caught lying in court, but there isn't much more we can realistically do.
This is similar to a case a few years back where someone asked to see the source code of a breathalyser that had found them guilty. AFAIR, the breathalyser was incorrectly averaging the readings, giving disproportional weight to the first reading. I don't know if it was enough to rule in their favour, but I'm sure it called the data into question Edit: Looks like it was a Draeger breathalyser https://www.schneier.com…
Earlier quoted context omitted.
> Forensic labs and companies are expert witnesses with black box processes and the incentive to protect the authority of their profession. I imagine they're also incentivised to favour a result indicating guilt.
Why? The risk to their reputation & livelihood for a false positive should temper that.
How would a team of independent reviewers even go about examining source code of this scale? How can you possibly find bugs that aren't super obvious?
Something tells me that 170k of Matlab code is going to contain plenty of obvious bugs.