But it is clear that this legislature is pushing for crime prevention based on aggregated data from computer systems(where from, exactly?)
The arguments are that a manual extraction and collection would be too slow and that this should be limited to only a handful of severe potential crimes. As a prevention rather than mitigation technique.
This is quite bad.
Now the objection is not purely because the data is coming from computers, but because it will reveal data on people tangentially connected to the targets.
This is so wrong in so many ways and the people who are pushing from this have learned nothing from the stasi and nazi regimes, or perhaps they are nostalgic.
It is very clear that these things are pushed by rather smart minds, matter experts and this makes it clear they do not care about the innocent peoples privacy, as they could not possibly be so ignorant to overlook this matter.
Justice even the conventianal way spawns injustices, now the lazy government wants to automate and scale their procedures. W.t.h.
It isn't just about their privacy but also their very freedom. Being in some police database might very well land you in the crosshairs of some investigation and related niceness like searching your home, taking you in for questioning and keeping you there because you didn't say something...
I was working that this behaviour by FAANG will normalise it, and will lead to our governments doing the same.
Same 'smart people' work for government and fang, they are gonna behave the same way.
Skipping ahead to the fact that automated data analysis is not prohibited per se, just needs to be reasonably targeted, the two most interesting paragraphs are the last ones for me.
Basically the court mandates that any future laws passed for legalizing automated data analysis must explicitly
1. For any given crime to be automatically analysed, define the data model and define appropriate analysis techniques or methods. Those laws should be interesting reads.
2. For any one analysis happening, document and later publish to the public domain the reason for initiating an analysis, the methods chosen, the criteria for labeling someone as "guilty/innocent" - basically the entire process so that the analysis is auditable end to end.
Again, this is not a suggestion in the ruling, but a requirement.
This is really good news for personal liberties in my book.
Stark contrast to US/ China data mining on citizens.
(scroll down to 3.1.(g) Public sector)
https://www.mondaq.com/france/technology/1059760/artificial-...
> whenever an individual decision is taken on the basis (even partially) of an algorithm, the administration must [...] explain [using easy to understand language]†, at the request of the individual, how the relevant algorithm works [...], by providing the following information:
the degree and mode of contribution of the algorithm to the decision making;
the data processed and its sources;
the processing settings and their weighting applied to the situation of the data subject; and
the operations carried out by the processing.
† a very important detail that this translation for some reason omits ??Now, something that has been bugging me since then, especially in the «AI» context that was also being discussed more or less at the same time by the legislators, is that so far this is impossible to do for neural network based programs, since the final program is not based on any algorithm (as commonly understood), or at least not one that even the programmers themselves could understand, much less distill into a commonly intelligible form !
Whereas you say: >Skipping ahead to the fact that automated data analysis is not prohibited per se, just needs to be reasonably targeted, the two most interesting paragraphs are the last ones for me.
I think this reads like the court curtailed reasonably broad data grabbing behaviors such as those Snowdon has revealed happened illegally and are still happening in the US. It still permits effective, focused/targeted data mining.
So at least the court learned from the Nazi/Stasi times but also from the times of left-wing terrorism ("Rasterfahndung" during the Bader-Meinhoff terrorist gang) - i.e., that these techniques are effective yet must not be applied to broadly. If this reading is correct (I'm not a legal expert), then the main danger is that lawmakers are incapable (or unwilling) to implement requirements (1.) and (2).
"Automated data analysis or interpretation requires justification under constitutional law."
I'm sure something is lost in translation. But even assuming a lot of context, I find it hilarious. Germany's highest court seems to have stumbled into avoiding a dystopia not because they don't want some biased ML model bullying minorities, but because they plain and simple don't like all this computer nonsense. (Grumble grumble.)
You either need a lot more budget (= oversight) to hire all the people to do the work, or do less of it.
Sure computers also cost money but you can keep using them as long as you pay the (already budgeted) power bill and (also already budgeted) maintenance/upkeep.
The issue that the constitutional court is pointing out, is that there is no appropriate threshold for using such analysis software. In other words they that that you can't use is for anything, just when it is really necessary and proportional.
Police in the United States are close to this already. Chicago's police department used to brag about its data-crunching ability to predict where crime would happen. At the time, it was just by area, but it was stated that with enough data, it could be narrowed down to the block. It's not that big a leap to extrapolate that to the individual.
My guess, though, is that it didn't work, and the city was just parroting the promises made by the salespeople of whatever system they bought for this task. If it worked, then police would preemptively flood certain areas where they know crime is about to happen. But, as is seen on television most nights, and in the complaints of the aldermen, the police mostly seem to react to crime, rather than prevent it. The use of social media by crime mobs has only made it worse.
No constitutional law expert myself, trying to contextualize in the hope it helps:
Core pillars of the German constitutional system are informational self-determination (i.e. personal data has a very high degree of protection under the constitution, see Germany's history) and proportionality (in using its powers, the state must choose methods appropriate to specific situations, obv. to hinder misuse of power - see Germany's history again).
So the ruling in essence, as I understand it, says among other things:
1. Automated data analysis needs to respect informational self-determination of anyone (i.e. no data mining per se on people who are not suspects for a given type of crime being investigated)
2. Automated data analysis needs to be proportional, meaning that the purpose of running any analysis "for the benefit of the public" must be reasonably well established and the analysis must have a strict goal.
I.e. if John stole something, you have to analyse the data such that you can credibly prove that identification for theft was the purpose.
Starting with looking for thieves but ending up identifying X movie pirates, Y weed consumers etc. is not proportional and crosses the right to informational self-determination.
So in the above example, the movie piraters and weed smokers might have not had an official investigation into their offenses on record, so by law the presumption of innocence must hold - you cannot do blanket investigations without reasonable cause under German law.
In the final paragraphs, the text actually states under which conditions automated data analysis would fullfil constitutional requirements:
1. The legislative needs to codify which data points exactly can be used for a given offense where offenders are to be identified with data analysis. Laws also need to explicitly codify limits of usable data.
2. If automated data analysis is performed by authorities, they must document the reason for initiating an analysis, the methods chosen, the criteria for labeling the outputs as "guilty/innocent", document the entire procedure and be able to release all of it to the public domain so that public data protection officers - or anyone else - can audit the results.
All of this sounds super reasonable to me.
To be clear, automatic data analysis for the purpose of extreme threats to national or public security are exempt if I understand correctly. These would be cases where it is proportional to perform such analyses for the greater good.
But you cannot do mass number crunching on ordinary citizens.
Yay in my book.
This seems like a judgment that focuses on process and not outcome.
I have probably butchered what my friend said.
If you introduce automation doing the same thing at a scale corresponding to 10,000× the human capacity, you will have 10,000× as many errors. This will violate Blackstone's ratio and erode trust in the system.
Just imagine, everyday 6AM an officer come at your door to check your apartment "to prevent" your crimes, suddenly every random guy in the street looks suspicious and spying on you, then some officers just randomly walks by your workplace, police stop you everytime you travel... This would eventually lead to an insult to the mafia guy and then boom you have already commited contumacy and the model was right! You are a criminal and you should be punished!
It's not a broad decision that automated data analysis for crime-prevention is unconstitutional.
This roughly translates into "there has to be a sufficiently concrete danger to warrant such a ambiguously formulated infringement."
I truly appreciate this sentiment. Any collection or analysis of data "just because there might be something in there" has to be weighed against the right of everyone that is being data-mined.
This freaked me out about Minority Report when I first watched it: those precogs are essentially hacking into the future memories of everyone involved in _and by happenstance surrounding_ the lives of future criminals and their crimes.
Mass-scale datamining is effectively the real-world equivalent of telepathy. And I want neither precogs nor the government in my mind.
The truth is that Law Enforcement investigations are currently taking information from dozens of silos/databases, manually copy&pasting them into Excel and Word and then finally making a pdf or powerpoint to share the results. This is obviously a really inefficient way to do things and wastes a lot of taxpayer money.
A technical solution that would allow LE to crosslink data they already have is a no-brainer.
P.S: Very biased view, as I'm building/selling exactly this tool :)
A technical solution that is Open Source, that falls under federally mandated criteria for how it can be legally used, that automatically documents any and all access and makes the entire audit log part of any output result so that all happenings are easily auditable by 3rd parties... might be a no brainer.
Anything else I'm highly skeptical.
If going after syndicated and or international crime, fine.
First feature I'd like to see is linking Finanzamt with LE and go after money laundering - Germany being an absolute paradise to the detriment of everyone else.
If it takes manual copy paste to protect ordinary citizens from overreaching police, I welcome that technical limitation.
Very biased, as I am an ordinary citizen.
For eg: denying service from a German retailer depending on the ML prediction result of the transaction being classified as fraud probable(Germany has pay by invoice), would also be constitutional or not?
Or credit score/risk rating using ML, trained on a feature extracted from parameters like Zip code, ethnicity could also classified as automated data analysis for prevention of criminal acts, right?
Or this is only applicable to police?
The reason they do this isn't to fight crime, but to save the business from losing money. So I don't see how a ruling about preventing crime would apply.
I think it's more like borderline presumption of guilt.
And I think that is ok...
In the end the government cancelled the program because of free will. A person always has the choice to change his actions because his actions aren't predetermined.
just like once upon a time it was constitutional to own slaves and now it isn't
just think about the monetary cost-savings! (nevermind the human misery)
Germany leans heavily to more and more privacy because they have the experience in recent memory of what a government can do with vast information collection and centralized authority.
America leans lightly towards more and more centralization of data because they have the experience in recent memory of what individuals who "slip through the cracks" can do. So in Germany we see decisions like this and in the US we see police using privately-curated DNA databases to defrost the cold case of a serial killer, and they're lauded as heroes.
(It's worth noting, of course, that this is all relative. Relative to most of Europe, the US is starting from a place of massive decentralization and minimal individual-tracking. Some people still find social security numbers controversial).
https://www.stasi-unterlagen-archiv.de/mfs-lexikon/detail/ze...
Link is german: Shows how denunciation aka controlled leaking of private life like affairs was used, to destroy the lifes of resisting citizens.
PS: FB and Google could in theory use similar instruments against citizens that put pressure on them. Nobody could ever catch them at it.
If the ccp ever gets swept away and replaced by a new better china, the resulting parties will outdo one another to promise freedom and privacy.
For instance, in France specifically since the 1974 SAFARI scandal that resulted in the creation of the French data protection authority and assorted laws in 1978, and happened when the government in secret attempted to (and had to stop due to the outrage) link various (then paper) administrative files into a centralized computer database through a single social security number.
I theorize that in a hundred years or so, North Americans will have learned this lesson the hard way and will have atrocities in recent history, too.
I'm sad I won't see the day when Bluffdale is repurposed as a memorial to state surveillance, much like the Stasi headquarters are now in Berlin.
It's the working class who pay the price at the moment. We get burgled, mugged and killed, meanwhile the wealthy lawyers and politicians can pontificate over constitutionality from their gated communities and private vehicles.
It's insane that the EU is trying to ban automatic facial recognition with the terrible AI act - https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=celex%3A... - nevermind the sheer amount of bureaucracy it introduces for startups (a government register of "possibly harmful" AI - wtf?).