Regulate This! Time to subject algorithms to our laws Algorithms are almost as pervasive in our lives as cars and the internet. And just as these modes and mediums are considered vital to our economy and society, and are therefore regulated, we must ask whether it's time to also regulate algorithms. Let's accept that the rule of law is meant to provide solid ground upon which our …
Another opportunity for the government to do all the wrong things for (possibly?) the right reasons !
Perhaps the first step in this process is legislating that the lawmakers are required to have a more than "just adequate" understanding of the technology they are regulating, or we shall see repeats of wholly unacceptable situations like the head of a government agency not knowing what the thing they were legislating against actually was ..... anyone remember cookies ?
The mortgage one is hilarious. When I was buying my first house, the bank told me at first that the computer didn't think I could afford the repayments month to month.
The fact that it was a clear £100/mo less than what I was paying in rent, that I was already paying and managing to save up for a deposit while paying, should have been self-evident that I could afford a smaller amount.
MK_E, you were lucky. In my youth I was denied a mortgage, despite the repayments being lower than the rent on a ****hole room without luxuries like hot water in the communal bathroom.
I fled the country to escape that, and so missed my generation's chance to buy at a reasonable price in the 1990s. But at least today's rental market is much-improved.
That's quite a good case in point - precisely because it is so simple. For a start, as soon as the bank people said anything like, "The computer doesn't think you can afford the payments", they revealed their utter ignorance of what was really happening.
Computers do not think. One day they possibly might, but as of today they don't. What they should have said was, "We have done some predetermined sums on our computer, and we don't think you can afford the payments".
The decisions were all made by bank staff - probably managers - and then programmed into the software. If any mistakes (or legal offenses) resulted, they were the fault of those people.
We will have precedent and case law in a few years.
A few trips up to the highest courts and lawyers will be quoting cases for and against blaming the vehicle (and thus the SW and HW therein) for various kinds of incidents.
The Tesla/truck crash which killed the driver must already have started the ball rolling - the system can't just allow Tesla to state 'Hmm, didn't see that truck - will tweak the algo' despite requiring the driver to essentially never use the actual 'self-driving' tech.
Once they start blaming the car tech, the flood gates will open and other tech will start getting blamed for stuff - 'I didn't WANT two tons of creamed corn, it was Alexa's fault"!
one of the oldest rules in computing
Garbage in , garbage out.
You can have the most wonderful algorithms ever, and completely proven to work doing what they are suppossed to be doing. sadly the data entry gets a minus sign in the wrong place, and next thing you know is that mariner 1 does a 180 degree turn at mach 2
But the idea of regulation does raise a concerns. are they qualified to do the job? how long will they take to do the job? how much will it add to the cost? and finally what happens when they f*** up as well?
I think this article would have benefitted greatly if at the start you had defined what you considered an "algorithm" to be. Without this, especially given your apparent definition differs from what most programmers would use (I think, I can't be sure because you didn't give a definition), it isn't really possible to give any value to the subsequent points made about them.
By the end of the article I got the feeling you could have just replaced the word "algorithm" with "scary thing".
"Algorithm - a process or set of rules to be followed in calculations or other problem-solving operations".
The eponymous origin being the approach to solving mathematical problems by Mohammed ibn-Musa al-Khwarizmi (780-850). Applications in electronic computers are only a recent use - algorithms have been a recognised method of processing information for over a thousand years.
"Algorithm - a process or set of rules to be followed in calculations or other problem-solving operations".
I once had some university professor ask me what algorithms I was using on my current project, which was a simple data transfer from one system to another.
No calculations or problem-solving as far as I could see, it was simply shifting data from A to B.
As such, I didn't understand the question.
I suppose it did demonstrate that I wasn't used to obfuscating stuff by the use of academic language...
"The answer must be a right of appeal to legal precedent [...]"
Unfortunately in the UK you need deep pockets to take the case all the way through the appeals process. Only an Appeal Court ruling forms future legal precedence - no matter how many times juries refuse to convict people for the same offence.
Too often these days the law makers rush bills through with insufficient scrutiny. Leaving some poor sod to be prosecuted by the police and CPS who will push the limits of a law's vague wording.
Clearly use algorithms that are stupid
When working out your monthly payments they assume that your energey use will be the same in December as it is in June. Clearly this is not the case for 99.99999% of the population but still they use it.
It would not be rocket science to introduce an ALGORITHM that worked on the basis of actual average data obtained from your meter readings but will they? Like heck they will. Using a direct debit allows them to adjust your payments over the year depending upon your use. The only changes they will make would be to increase your payments. Like petrol prices, up in a flash and down sometime never.
That money of yours they have when you are in credit is free cash and can even earn them interest. Do they give that back to us?
Like heck they do.
I may be wrong, but I think the article is confused, in the way that all or almost all people are when it comes to law and ethics.
Law is neither here nor there, and quite often it is a problem in and of itself.
The key principle is that everything two entities - companies, people, Governemnts, whathaveyou - do together must be voluntary (they must both agree to it) and well-informed (they must know what they're agreeing to). So you can't coerce, and you can't deceive. The sole exception of self-defence and then all bets are off.
Where law is compatible with this principle, it's good and right; where it's not, then law is being used by one group to enforce its will on another group, without their consent and not in sefl-defence.
When we talk about algorithms and all this - it's kinda neither here nor there. What matters is *disclosure*. If you sign up for something, and you've been well-informed about whatever it is, and you agree to it, *then by definition it is ethical*.
One case where there are problems is when there is no choice - so let's say there's only one electricity provider in an area, and they want to use smart meters which are known to be incorrect in their readings. You then have Hobson's choice - electricty and bad meters, or no electricty. What I usually find in these situations is that *prior* to this point, someone or something else was imposing non-voluntary contracts - such as regulating an industry and reducing competition - which in turn *led* to this problem, further downstream.
In such situations then you are in fact being coerced, at one remove; all bets are off. You are free to act in whatever ways are necessary to ensure you are not coerced or deceived. Hack the meter.
"The key principle is that everything two entities - companies, people, Governemnts, whathaveyou - do together must be voluntary (they must both agree to it) and well-informed (they must know what they're agreeing to). So you can't coerce, and you can't deceive. The sole exception of self-defence and then all bets are off."
Which is usually never the case. Each side is usually trying to hide things from the other: either to outplay the other or as a defense against the other trying to backstab them. Confidence is something that usually only comes with trust, and trust is the exception, not the rule. Unless there's a crisis, one man usually doesn't trust the other and will tend to act in competition. That's why theoretical things like pure capitalism don't work in reality (asymmetry of knowledge) and why you have thought experiments like the Prisoner's Dilemma.
"In such situations then you are in fact being coerced, at one remove; all bets are off. You are free to act in whatever ways are necessary to ensure you are not coerced or deceived. Hack the meter."
Nope, because the electricity companies tend to have government mandate on their side. IOW, if all bets are off, what do you do when it's the OTHER side that has all the guns? Oh, and a willingness to use scorched earth tactics?
We should not be using algorithms to determine sentences for criminals, the judiciary is there for that to determine the facts of the case and the intent of the criminal. I understand it is still an opinion but its better than something that does not know or see what is in front of it and is basing it's opinion based on information fed in to it.
Mortgages are based on ability to pay so I have no problem with them, money in - money out = amount you can afford with an adjustment to factor in interest rate changes.
Googles algorithms should be open to scrutiny by a legally backed watchdog because they can make or break a company/product/person/opinion.
At the end of the day the box (Pandora's) is now open and it's not going to get shut.
"Googles algorithms should be open to scrutiny by a legally backed watchdog because they can make or break a company/product/person/opinion."
That might be easier said than done. We keep getting told that even the authors of the learning programs don't know how they work after they've been subject to training.
In any event I'm not sure even the results would pass scrutiny. Google seems to specialise in presenting hits which are irrelevant to what I actually want too much of the time. The likes of Amazon & eBay all too often present prompts along the lines that "You might also be interested in ${What I just bought or something similar and won't need to buy again for a long time}" or "People who looked at $Thing also looked at ${Amazing variety of other things which just goes to show that looking at $Thing doesn't correlate well with looking at anything else}". I'd have thought that anyone who's experienced this would think very carefully about devolving any serious decision to such processes. Clearly, however, too many in business can't understand what lies in front of them.
"We should not be using algorithms to determine sentences for criminals, the judiciary is there for that to determine the facts of the case and the intent of the criminal. I understand it is still an opinion but its better than something that does not know or see what is in front of it and is basing it's opinion based on information fed in to it."
But if you depend on humans, what happens with a charismatic suspect?
"Mortgages are based on ability to pay so I have no problem with them, money in - money out = amount you can afford with an adjustment to factor in interest rate changes."
But since mortgages tend to be long-term things, they also have to take vulnerability into account. How likely is the borrower to suffer a significant event that severely alters his/her ability to fulfil his/her end of the deal (say, the industry he/she is in is prone to collapse leaving him/her not just unemployed but unemployABLE).
"Googles algorithms should be open to scrutiny by a legally backed watchdog because they can make or break a company/product/person/opinion."
But Google is multinational. They can probably play foreign sovereignty against you. What will you do then? Block Google and get complaints up the wazoo?
""We should not be using algorithms to determine sentences for criminals, the judiciary is there for that to determine the facts of the case and the intent of the criminal. "
The politicians in England have imposed rules for determining minimum sentences on conviction. If someone pleads guilty immediately there is a fixed discount on the sentence. If they insist on a jury trial and are found guilty - then there is no discount. If an innocent person changes their plea to guilty because of the fear of a subjective jury finding them guilty - then apparently the judge is not allowed to impose a lesser sentence than the minimum that the rules prescribe.
Minimum sentences rules were introduced because tabloid-fearing politicians didn't want judges exercising any intelligent discretion based on the evidence presented.
Fear of the unknown on a societal level can be a very ugly thing.
Out of curiosity, has anybody put these politicians on the spot and asked them point-blank to define exactly what they mean by "algorithm"? The answer (or lack thereof) might be amusing ...
Tom, if you have enough energy to boil your blood & yell at DearOldTelly, you'd probably be better off spending your copious free time educating the voters in your neighborhood. If we all did, something might actually get done.
(Can you imagine fifty people a day? I said FIFTY people a day walkin' in, singin' a bar ... Oh, wait, that was a different protest entirely. But it just might work ... )
Some laws stop us taking each other's stuff (property, liberty, lives)
Whoa, slow down there fellow - laws do NOT stop anyone from doing anything.
Murder and theft have been officially illegal for at least 4,100 years that we know of[1] and yet stuff still gets stolen and people are still murdered every single day in every single country on the planet. Laws merely establish a fixed and uniform punishment for specific acts so that everyone knows ahead of time what the consequences are should they be caught committing one of said acts.
While some people *might* weigh the punishment against the crime and choose not to murder their coworker or steal that shiny-shiny from the jewelry store, there are plenty of other people who don't perform such calculus, or reach a different conclusion, and thus steal and murder as they please.
It would be more appropriate to say that laws discourage us from taking each other's stuff.
[1] The "Code of Ur-Nammu" dates to ~2100-2050 BCE and specifies punishment for (among other crimes) murder, robbery, adultery and rape. Spoiler alert - the penalty for all of them is death.
"[1] The "Code of Ur-Nammu" dates to ~2100-2050 BCE and specifies punishment for (among other crimes) murder, robbery, adultery and rape. Spoiler alert - the penalty for all of them is death."
I suspect given the conditions of that society (no such things as jails, for example, and no practical destination for an exile, etc.), death was basically the only option that would stick.
PS. In the end, laws are just ink on a page. What matters law to a charismatic sociopath able to raise a army big enough to overrun you?
AFAICS this lies at the heart of data protection legislation which includes control of processing. A good deal of the intent of such legislation is to prevent inappropriate processing. As I said in a comment on another topic maybe tightening up here might lead banks to re-evaluate branch closures.
Sadly the people who will ultimately make the decision on this are virtually guaranteed to have obtained their positions via a way around the rules that apply to everyone else.
They are not going to make things fair, they have to maintain the status quo else they find themselves unable to compete.
This is very much what they fear along with being found out, so no it will never be allowed to happen and if ever it did then things would return to how they are now within one generation and even more biased within two.
It doesn't matter where or how you live, only in deperation are people not biased against the stranger no matter what they claim.
"And yet, if the last decades of open-source software have taught us anything, it is that simple availability does not incentivise investigation."
You don't have to look at open source for that, just look at much more substantial examples. How about projects which are basically build upon pseudo science and which can be proven to be bollocks by merely applying some simple mathematics on them.
For example: a project which will allegedly solve the worlds water scarcity by extracting it out of the air. People made a project, a nice presentation with featured unrealistic claims ("it'll easily extract 40liters of water per day") and as a result they managed to gain a lot of funding, including government funding. Even though it can be proven that the whole concept is flawed and won't work.
Project even made the news and hardly any reported bothered to also look at this from a scientific perspective or to get someone to do that for him.
This is about something in plain sight, fully out in the open, yet people still manage to allow themselves to be conned by it.
So then someone things that algorithms need to be more transparent? Uhm, right...
Let's accept that the rule of law is meant to provide solid ground upon which our society can function. Some laws stop us taking each other's stuff (property, liberty, lives) while others help us swap our stuff in a way that's fair to the parties involved (property, liberty, time)."
I'm sorry; you must be new here.
The law is simply a generally-accepted monopoly on the initiation of violence. This may, in theory be used to help society function as described above. In practice, it throws as few crumbs in that direction as is necessary to maintain the violence monopoly. The overwhelming majority of its energy is directed towards further enriching and empowering those who control the violence monopoly. This attracts people who prefer to improve their position through the unilateral application or threat of violence, as opposed to other groups who prefer (or are at least willing) to improve their position through other means, ranging from fair trade to outright fraud. In the US, people complain about having to choose between the openly sociopathic Donald Trump, the openly sociopathic Hillary Clinton - but the system only admits people like this. There is an enormous apparatus of local, state, and federal government that acts as a training and grooming program to weed out the weak and idealistic. Some people do get as high as the US Congress and keep their values mostly intact, but these people are so rare that they don't really mean anything. On the (R) side you have someone like Ron Paul or Justin Amash, and on the (D) side you have someone like Dennis Kucinich. These people are denigrated as loons, and the system tries very hard to reject them. I'm not saying that their positions are right or wrong, merely that they are or were fairly immune to corrupting influences against their values. Amash's own party spent an insane amount of money trying to force him out during the last election. People who cannot be bought and controlled are not wanted in government (note that I specifically exclude Bernie Sanders, who is just as corrupt as the rest when people aren't paying attention - which is most of the time - from the good guy list).
So then we have laws. Laws are designed to further enrich or empower the people writing them. This is sometimes done directly in a pay-for-play fashion, and other times indirectly by buying votes or otherwise paying enough attention to the outrage of the day to make the unwashed masses shut up and go away. But the kicker is that most of the time it's both - legislation designed to help "protect" the Morlocks from the big bad business people who are.... actually the ones who write the laws. Big business then complain loudly, bitterly, and publicly about how unfair the legislation is, the politicians claim Victory for the People, and the media (almost entirely comprised of pathetic, self-important rubes) gushes in admiration. The end result is that Big Business has fewer competitive worries and few (if any) meaningful costs or restraints placed on them. Every once in awhile a Sacrifice Must Be Made For the Greater Good (a company gets whacked hard), but again, this is relatively rare and almost always done in favor of some competing oligarchy so that money and power are not lost overall by the political class.
A lot of people say that this description of laws is grossly unfair, unrealistic, etc. Nope. In my younger days I was rather involved with it. I've literally been in the room where businesses helped write regulations to "improve public (whatever - safety, health, etc.)" in ways that covered what they were already doing and specifically hurt or eliminate their competitors that couldn't afford to keep up with them. The word "monopoly" was originally coined as a privilege granted or sold by governments to give an organization exclusive rights to some sort of business practice. These days, we call it legislating for the public good.
There are a lot of metaphors for politics; the most useful one I have found is that it's just a specialized form of gang warfare controlling the most rarified form of turf.
Now we bring in AI, which some people fear will be even smarter and more evil than Big Business. And we should have bigger and badder laws to deal with it! But... if AI is that big and smart, then why won't it game the system even better than Big Business has? Especially when you consider the fact that all of the big, bad AIs will be controlled (at least initially) by Big Business? Does anybody have a plan to make this work that doesn't involve unicorns and fairies?
We can't even control dumb megacorporations with nation-states. We're pretty thoroughly screwed trying to apply the same techniques to these hypothetical ultra-smart ones. Nation-states themselves are the biggest polluters on the planet, and the worst of the pollution that they don't create directly is the result of them selling off the rights to pollute. They kill more people through violence - by many orders of magnitude - than even the worst corporations. DuPont has killed a lot of people, but they're pikers compared to even a mid-tier national government. Democide (the slaughter of people by their own governments, excluding war) has killed around a quarter of a billion people in the last century alone. It's hard to figured out whether the number of people locked in cages for political or victimless crimes is in the 8-figure range or the 9-figure range. In exchange we get schools (that graduate illiterates by the tens of millions), roads (that cost exorbitant amounts of money to build and maintain, if and when they're actually maintained), national health care (that is escaped by anyone who can afford to), national defense (that creates more enemies and terrorists than it eliminates), and a justice system (whose results are more dictated by the wealth of the defendant than the guilt of the defendant), etc. I realize that a lot of people are quite concerned about losing these "essential benefits," but personally I think that we're already living pretty close to the worst-case scenario there.
Time to start cracking on the next stage of organizing humanity....
But then comes the armor-piercing question.
"Can you think of any better without changing the human race as a whole?"
In other words, what you describe sounds like the absolute pits...until we start looking at the alternatives.
Otherwise, we may be better off just waiting for the Taelons or whatever to come and become "beneficent guardians" for our own protection.
But then comes the armor-piercing question.
"Can you think of any better without changing the human race as a whole?"
Oh good grief, how can you not? Nothing is provable in a comment (that requires tons of a priori reasoning followed by experimentation), but there is an absolute mountain of theory to throw at the subject:
1) A system of ethics (and government is just a practical application of this) should function unilaterally: that is to say, my ability to coexist and interact peacefully with others should not require that the others share my values. Tons of stuff has been written on this.
2) Ethics are simple. The problem is that people want things or want to do things that can't be accomplished ethically, so we introduce rationalization. Rationalization is complicated. If ethics appear to be complicated, then start eliminating rationalization.
3) Representative government does not scale well. I don't pretend to know where the line is where it starts to fall apart (and that would vary by culture), but my wild guess would be 50,000 people for an overall entity.
4) I'm discarding direct democracy, as that's just an ISO-Certified methodology for mob rule.
5) Forcibly grouping people into blocs based on which map coordinates of the planet they happened to be at when they exited their mother's womb is perhaps not the worst way to organize societies, but it has to be close. Lumping people together based on geography stopped making sense at least 50 years ago, and gets worse every day. Government should be something you join, like a church or a Rotary Club or something, because it actually reflects your values. And you should be reasonably free to change governments fairly easily. If you, in your heart of hearts, want to be a hardcore communist you should be able to go be a hardcore communist. If you want to be the opposite (anarcho-capitalist?), then go do that. If you want something in between, you should be able to find something that suits you. Some people love leisure, and some people love work. Some people love to work with others, some prefer to work alone or in small groups. None of these approaches is inherently wrong, but mixing these types in the same government is problematic at best - people then are incentivized to force others to be Their Type.
6) Any organizational methodology that involves being constantly at war is almost certainly defective. The rationalization for this is that the world is full of assholes that must be dealt with through military might, but if this never ends then chances are you're the asshole. One thing I've learned through world travel is that most people just want to be left alone in peace. That doesn't mean that they don't get angry or have grievances, but that almost never drives them to violence on its own because most stuff just isn't worth getting you or your family and friends killed over. It's governments (and occasionally religion) that rile them up, under pretenses that almost invariably turn out to be dishonest.
7) Somebody will always get screwed in the end. If you try to eliminate this entirely, you only guarantee screwing everyone. I would personally value systems that are more quick and flexible with self-correction over systems that are slow.
And I could probably go on for another 20 pages, but this should be enough to get somebody going.
Except every single method you propose has fatal flaws.
1-2) People will CHEAT and then hide the fact they're cheating. Humans will be bastards if it'll give them a leg up on their neighbors. It's damned near instinct.
3-4) NOTHING scales well for 7 billion (and direct democracy pretty much doesn't scale past tribal size), and unless you have a system that can encompass EVERY human, one side or the other's going to feel slighted and want revenge.
5) Eventually, two such blocs will end up at odds. Usually over resources like arable land or women. If it can happen to two people, it can certainly happen to two blocs of people. Geography WILL matter at some point because the Earth is finite. EVERYTHING is finite at some scale.
6) Except when there are too many people. You eventually end up with a "Baker's Dozen in an Egg Carton" situation: too many people for whatever geography can accommodate. At that point, war isn't just desirable, it's inevitable. Either SOME die or ALL die from sheer exhaustion of resources.
7) And many humans hold grudges. If you screw someone, you risk an act of revenge, and some people see Mutual Assured Destruction as an acceptable scenario in that context.
From the text of the article, the laws and sentences are being subjected to algorithms. The problem is that the humans who should be giving the results a second thought and using them as a guideline are instead rubber stamping the results.
In that they are embedded in products and services. If you sell a product or service with an algorithm that makes that product unsafe, then your company is liable. If you create an algorithm that misrepresents data, as in a financial scam or your average VW diesel, you are liable. If your algorithm unjustly impacts the economic opportunities of one or more racial/ethnic/gender groups, then you are liable.
Yes, you have to be careful when writing or incorporating algorithms, but the same can be said for code in general, or even many physical components of a tangible product. I worked at a medical device company where our materials people changed out the type of plastic used in a device, (stupidly) without testing the toxicity of the new plastic--oops. The Food and Drug Administration slapped us into next week over that one.
In terms of public services/government, you can in many cases also sue if an algorithm impacts your benefits/employability/economic or property rights. You can also vote to "kick the bums out" if a specific elected official or group of officials is involved.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
