5
$\begingroup$

The famous quip has it thus:

some man with a hat and bowtie
Democracy is the worst form of government, except for all the others.
― Winston S. Churchill

However, democracy has numerous problems as well, notably the iron law of oligarchy (whereby a small group of well-organized, well-motivated folks will usually succeed in enforcing their will over the majority), disenfranchisement of minorities, the tragedy of the commons, numerous principal-agent problems and other such, a bias towards short-termism, and other issues previously discussed here.

Now, if you stop and ponder things for a moment, the fundamental problem is one of information aggregation and preference aggregation. In a complex modern democracy individual voters cannot (and need not) be particularly well informed on the specifics of any and all political actions. They can and do however have a sense of whether things are going generally in the right or wrong direction, and the mechanism of regular elections empower them to throw the bastards out without the violence typically associated with normal governmental upheaval that one sees during regime change in places that do not allow meaningful elections.

Consider, further, the internet of things. Imagine a near-future where there are sensors literally everywhere, from the tallest mountain to the deepest sea, producing a constant stream of information, all of it aggregatable in some shape or fashion in more or less real time. Moreover, the data about each individual human (or dolphin, or chimp, let's not be species-ist) related to their past behavior can be used to infer preference, or if there is insufficient information somehow they can simply be asked in some suitable fashion, creating an extrapolated volition of the most sapient lifeforms of the planet.

Assuming (just for the sake of argument) that an oligarchy will not instantly seize control of the weights assigned to the preference aggregation algorithm/AI, could such a system be fairer than democracy?

$\endgroup$
  • 1
    $\begingroup$ So someone would get nominated for the position of leader because s/he is a good driver, buys fresh milk from Good Co. every Friday, puts glass recyclables in the correct bins, and has never visited an adult entertainment website? $\endgroup$ – Frostfyre Apr 14 '16 at 19:06
  • $\begingroup$ @AmiralPatate Imagine something like the common law system -- when someone is wronged, they sue. This creates precedent and is integrated into the body of legal knowledge, making it a teensy bit smoother than before. This system works similarly, except faster, by extrapolating volition from sentient beings -- if longings for more parks seem to prevail in the area, eventually a park is built, if people long for more stripclubs, the same... $\endgroup$ – Serban Tanasa Apr 14 '16 at 19:20
  • $\begingroup$ Okay. So I deleted my comment because I thought I understood finally and it became unnecessary, I wasn't exactly right. Now I get it though. $\endgroup$ – AmiralPatate Apr 14 '16 at 19:31
  • 1
    $\begingroup$ Just a word of caution: sensors do not produce information. They produce data. Data is to information as words are to a poem -- a poem is composed of words, but very few people will find a dictionary to be poetic. $\endgroup$ – AlexP Mar 28 '17 at 10:27
3
$\begingroup$

This sort of scenario has been explored (in some way, shape, or form) in many books, and even video-games. It is actually a video-game that came to mind when I read your question.

Allow me to tell you a little bit about it, and then expand on the idea:

Deus Ex

The original Deus Ex game puts you in the shoes of J.C. Denton, a nano-technologically upgraded "agent" who discovers a massive conspiracy. "A small group of well-organized, well-motivated folks" have already amassed a lot of power and influence, and seek total domination of the world.

One of the ways in which they seek to gain and maintain control is through the development of an AI - the first of its kind - which would be able to monitor everyone, every minute of every day, and feed the information back to a single control center.

The protagonist discovers that he himself is only an experiment meant to prove the viability of nano-technology, and that the ultimate goal of said technology is to enable the leader of the cabal to integrate with the AI, and thus directly be able to access and control all the information in the world, giving himself God-like powers. (imagine taking over any machine, any piece of information, and twisting it to your purpose).

In the end the protagonist is given three choices: to destroy the AI facility and, due to the cabal's interference, the entire global communication's networks (thus plunging the world into a dark age), to kill the leader of the cabal and allow the Illuminati (don't ask) to take over the project for their own purposes, or, and this is the canonical choice, accept an offer made by the AI itself.

The AI observes you as you dig deeper into the conspiracy, and does not wish to "die", although it is also powerless to stop the cabal from harnessing it for their purposes. It instead offers to merge with the protagonist - whom it has judged to be a moral individual, as well as nano-technologically compatible to it - and find a new purpose, together.

Long Live the King

In the second installment of the game we find out about the consequences of this decision, and ultimately meet the AI-human hybrid which J.C. Denton has become. No longer truly human, J.C. agonizes over his lost humanity, but also over the fate of the world, which he sees slipping back under the control of various cabals.

If you choose to support him over the various other interest groups, Denton returns to the world, and makes a decision which fundamentally alters the destiny of man-kind: he releases a version of the nanites into the atmosphere which seek out every human being and enhance them (whether they want to or not). Each human being becomes plugged into a world-wide network with the J.C. AI at its core.

This human-AI hybrid becomes our omnipresent leader. Aware of what each of us are thinking, and doing every minute of every day, it is capable of offering advice, and judging each and every one of us in real time. Laws and governments no longer serve a purpose: the AI will decide our fates individually.

However, the AI also accepts input: each individual gets to cast their vote on important issues, and the AI takes their wishes and opinions into consideration. (such a situation is described in the closing cinematic)

Significance

And so, could such a system be fairer than democracy?

Yes, it could.

Such an AI could judge each and every one of us individually. If a misinformed, manipulated youth were about to commit a criminal act, the AI would be able to step in and take control of the situation. The youth would be helped, and counselled, not punished. The ones who had tried to manipulate him would not be able to escape the reach of justice: the AI would know exactly who they are, and where they are located.

A hungry individual would be fed. A depressed individual would be comforted, and pointed to help.

The AI would know your needs, know your fears, and strive to help you, or give you guidance, all in accordance with the moral code it inherits from its creators (or the one it develops).

Such a system, if it was not controlled by corrupt individuals (or even well meaning, but ultimately misguided idealists) would possibly make for as close to perfect a system of government as we could ever hope for.

$\endgroup$
4
$\begingroup$

For starters, it's a false choice. "Lots of information available" is not a new form of government. It's completely compatible with democracy, dictatorship, aristocracy, or any other form of government. The information may be used for good or ill under any form of government.

But perhaps more relevant to what I think you're driving at in the question: As best as I can figure out, you are positing a society in which government policy decisions are made based on the aggregate of the desires of all the citizens, as determined by this Pervasive Internet.

Not to be a downer, but (a) that's not a totally original idea. It's just the old idea of "direct democracy", with a sort-of specific proposal for a mechanism to carry it out. And (b) it's not at all clear that this would be a good society to live in, for exactly the reasons that you spell out in the beginning of your post when discussing democracy.

Simple example: Suppose a new technology is invented. Some people believe this is a great new idea that will benefit society. Others say it has all sorts of undesirable side effects, or maybe even that its basic function is counterproductive. It's easy to think of examples exactly like this. Like fracking: Supporters say it has dramatically reduced the cost of energy production and has reduced hydrocarbon pollution. Opponents say it pollutes ground water and may cause earthquakes. So suppose you collect the opinions of all the people, and find X% favor moving forward with the new technology and Y% are opposed. If X>Y, does it therefore follow that the technology is a good idea? Probably not. Odds are that the vast majority are not qualified to give an informed opinion.

Similarly: Foreign countries A and B have gone to war. Which should we support, or should we just stay out of it? Often, the majority will never have even heard of either country, they have no idea what the conflict is about or who is in the right.

Etc. That's the inherent advantage of a "republic" over a "democracy", when people make the distinction. In a republic, the people choose representatives whom they trust to make informed decisions. In a democracy, the people make all decisions directly, often deciding out of ignorance.

There's also the high-level questions versus low-level. Libertarians believe that it is better to make the high-level decision that people should have the right to decide most things for themselves, rather than for the govenment or society to make case-by-case decisions on every possible action.

Like if you took a survey of Americans and asked them if they think that Islam is a good and true religion, well, I haven't taken the survey, but I think the majority would say "no". But it would be totally wrong to conclude that therefore Americans think that Islam should be banned. Ask Americans if they think that people should be allowed to believe and practice whatever religion they like, whether you or I agree that it is good and true or not, and I think the majority would say "of course".

$\endgroup$
  • $\begingroup$ Good thing to watch related to this Q&A: the Gachamen Crowds anime. In it the government introduces online voting and polling for citizens. $\endgroup$ – Mindwin Apr 15 '16 at 16:11
4
$\begingroup$

So from what I understand, your AI collects data, then tells a government what's to best course of action. The improvement over current forms of governments would then be speed and reliability.

The problem here is how do you train an AI to make the right decisions? And my opinion is: you can't.

To teach an AI, you need A) situations B) questions C) answers. You need to know all that, so you can feed billions of scenarios to your AI, so that it will make connections and learn from that.

If you don't have answers, you can't teach the AI. And as evidenced by how well our society works, we don't have all the right answers. How do you teach someone something you can't solve? You can't. You just can't. Or you'll teach them something that is inherently flawed and wrong.

The other way to teach an AI would be trial and error. Trial would take a lot of time, error would cost your country a lot of resources. I doubt you can afford that.

However, gathering data on people through IoT would make the Stasi jealous, so there's that.

$\endgroup$
  • $\begingroup$ And yet we just saw a bunch of programmers (who are not the world champs at the game of go) design a self-learning program that beat one of the best go players in the world... $\endgroup$ – Serban Tanasa Apr 14 '16 at 19:49
  • $\begingroup$ Essentially, AlphaGo learned from human players. Unlike chess, you can't brute force go. Too many things to compute. Instead, you look at the table, at the moves your opponent made, you look at all the games of go that are stored in your databank, and you find the move that is most likely to give you a win. $\endgroup$ – AmiralPatate Apr 14 '16 at 19:58
  • $\begingroup$ So we agree that 'If you don't have answers, you CAN teach the AI' $\endgroup$ – Serban Tanasa Apr 14 '16 at 20:56
  • $\begingroup$ In this case, you give the computer a million matches to watch. For each move, you know the state of the game and all previous states. That's the situation. The question is "is this next move a winning move". The answer is in the name of the winner. It learns from that to determine what is a winning move in a given situation. If you don't know who won, then the only way to know if it's a winning move would be through trial and error, i.e. testing all possible outcomes. Which in the case of running a nation isn't really doable. $\endgroup$ – AmiralPatate Apr 14 '16 at 21:03
  • 1
    $\begingroup$ @SerbanTanasa Go is a problem where you have a well defined set of rules (which include what is the objective of each actor). What is the objective of the government? Ensure that people does not die of hunger? Enforce private ownership? Those two are very basic functions, yet even those are not 100% compatible with each other. $\endgroup$ – SJuan76 Apr 15 '16 at 8:32
3
$\begingroup$

No, it will not

Your question falls apart on its premise, in that you assume that information gathering about individuals will be unregulated and ubiquitous. You have essentially taken the world you have today, added one thing and assumed that everything else stays unchanged. Examples from real life shows different.

Sweden — my home country — jumped on the digitalization train early. We also have something that is somewhat uncommon in the modern world: a national identity system. The very same day you are born (or registered as an immigrant), you are assigned an identity. The key to this identity is the Swedish personal identity number.

These two things together are firstly incredibly practical. That whole "two forms of ID" malarkey is non-existent here. A Swede has the ability to identify themselves unambiguously anywhere and everywhere. With the emergence of digital identification we can also do it over the Internet.

Second, this system is also abusable, if left unregulated. Sweden saw this early, and already in 1973 came the first law protecting a person's integrity when it comes to automated gathering, storage, collation and distribution of personal data. This law has evolved since. Also the EU has gotten aboard this train, and we now have the Data Protection Directive which all EU member states must comply by.

In short these regulations has the following to say about you gathering, storing, collating and distributing personal information on a whim:

You may not do that!

Any personal data that you gather, store, collate and/or distribute to others — through automated means — must be pertinent to your relationship with that person. Also you may not gather such data without the consent of that person. Finally you must have a pertinent purpose for doing it.

Also there are vocal and strong forces among the population that this is the way it must be. The debate on personal data and what others — especially businesses and government, but also private individuals — may do with data they have collected about you is in full bloom. And the consensus thus far is pretty much that personal information is to be regarded as "sacred". We are talking about placing the line so close to your feet that keeping a log about IP numbers about visits your website can be illegal and frowned upon. And although the US has not yet gotten aboard with this idea, you can expect they will.

And this smashes your idea to bits. Gathering and collating personal information will be regulated, and very tightly so.

So to answer your question: will IoT change democracy by providing detailed personal information about potential candidates?

No, regulation about data gathering and processing will quench that idea, and for good reasons.

$\endgroup$
  • 1
    $\begingroup$ "Also you may not gather such data without the consent of that person" Only if you get caught doing it. Also if you're a megacorporation, you just have to put that in the terms of use nobody reads. $\endgroup$ – AmiralPatate Apr 15 '16 at 7:01
  • $\begingroup$ Well... in Fiction Land where megacorps are evil and shamlessly do shady stuff behind the scenes and never get caught (until the third act of the movie), your statement may be relevant. In the real world, we are seeing the corps and megacorps actually standing up to these kinds of things. The latest, very salient example being Apple's refusal to cooperate with the FBI. $\endgroup$ – MichaelK Apr 15 '16 at 7:15
  • $\begingroup$ The "not getting caught" part was more targetted towards NSA. They've been operating for decades and most people only found out in the last few years. They would collect IoT data, legally or not. As for Apple, the funny part is that they TOS you have to accept after the purchase of their products probably states that they have to right to steal your data. But you know, only them can profit from it. $\endgroup$ – AmiralPatate Apr 15 '16 at 7:35
  • $\begingroup$ And not even NSA could keep a lid on it. I am not saying that there will not be illicit information gathering; real world examples show that, as you pointed out. But OPs question made the presumption that information gathering will be ubiquitous and unregulated. This will not happen. $\endgroup$ – MichaelK Apr 15 '16 at 7:38
  • $\begingroup$ I see your point, but I think regulation for intelligence agencies only serves to determine which proof is admissible in court. $\endgroup$ – AmiralPatate Apr 15 '16 at 7:58
2
$\begingroup$

Mind reading someone's intents with a pervasive IoT sensor network is orthogonal to government style

Let's go one step further than the described IoT. Assume complete knowledge about what a person will do or won't do based on their present state and past actions, not just what can be measured. Will this reinforce a trend towards democracy or authoritarianism?

Consider that Nazi Germany and Stalinist Russian maintained enormous amounts of data on the common person. Those two governments were definitely not democratic in any sense. Conversely, the US NSA probably has many orders of magnitude more data than the Nazis or Russians had on their people but democracy in the US hasn't collapsed (yet).

It depends on who has the data

...and the time required to care about what that data might say. Ultimately, it will depend on whether the IoT data network pushes power further down towards the common people or concentrates it higher up the governmental chain. If IoT pushes power down towards the common person, then that should reinforce democracy. Alternatively, if IoT allows the autocracy/oligarchy/etc to further concentrate power in their hands then IoT as described by the OP could be the greatest tool of oppression ever invented.

Behold, the Oligarchy cometh!

Unless the processing of all that data is democratized as well, only those that can afford the massive processing power required to make sense of all that IoT data will see a benefit from it. We see this right now as the NSA and other intelligence agencies employ hundreds of millions of dollars to gather and process the world's telecomm data. Common people don't have access to that data so the fear is that it allow the government to exercise undue control over people's lives.

$\endgroup$
1
$\begingroup$

While the IoT isn't perhaps the best way to achieve a government (your refrigerator knowing you need milk and ordering it from the supermarket isn't either a role or effect of government) you have hit upon perhaps the real issue here; the timeliness of information and action.

In a free market economy, individuals "vote" all the time with their dollars and their feet. Governments can set up conditions that make markets free and more efficient, or can gum up markets by instituting regulations that shut down competition, favour crony capitalists and rent seekers or put up barriers to new entrants, and so long as there is a free flow of information and market signals, people will "vote" for these government actions, and the results of the ongoing "elections" are registered as GDP growth or contraction, rising or falling incomes and standards of living. More desirable polities get voted on by people attempting to immigrate, while less desirable polities see a net outflow of people.

So the flow of information by actors with agency is the key. Anything which reduces the agency of people negates their ability to "vote" with money or mobility, and insofar as governments are willing to respond in a positive manner to the day to day "voting" patterns of the people, then there will be a much more flexible and responsive government. Governments which are willing to block the flow of information will eventually need to start enforcing more and more repressive measures to coerce the people to do the bidding of the government and deliver tax dollars and resources to the cronies and clients of the rulers.

$\endgroup$
1
$\begingroup$

As I often say, "fair" is a "four letter word" and an "F- word." Needless to say, any discussion of "fairness" gets mired in just how difficult that word is to pin down. However, if we ignore that word choice, the whole scenario is actionable.

The situation is more real than you think. It's forming every day. However, it is better phrased in terms of the information gathered and presented, rather than the tools used to collect and process the information. These are the interaction points between your AI and your government, so from the government's perspective, everything between those two points can simply be a black box.

Let's build some scenarios to see how this could go. Rather than running a government, which is a needlessly complicated scenario, let's consider two hikers on a hike that have gotten lost. One hiker is your government, the other is your AI.

Scenario 1
Government - Oh man. We're lost. I really want to be out of these woods by nightfall. Which way should we go?
AI - Optimal direction is West by Northwest, bearing 285.34 (std. dev. 3 degrees)
Government - Really? I thought we needed to head East to get to where we need to go.
AI - You are wrong.

Clearly the AI in scenario #1 needs to develop some people skills. The AI may have the right answer, but we don't always want just the right answer, we want enough justification that we can believe the answer rather than just blindly trusting it. This shows the challenge the AI has ahead of itself. Not only does it need to be a powerful enough number cruncher to crunch a few trillion numbers to arrive at the answer, but it has to present those results in the form of something that the human portion of the government is actually willing to act on. If the human portion isn't willing to play ball, nothing will get done. So let's try another scenario and resolve that error

Scenario 2
(Continuing Scenario 1)
Government - Let's just go East. I don't trust your numbers.
AI - picks the government up, slings it over its shoulder, and begins heading towards bearing 285.34

Oh man. That escalated quickly. Let's presume Skynet isn't what you were looking for with your high and lofty replacement for democracy. Moving on to scenario 3:

Scenario 3
(Continuing, again, from scenario 1) Government - How do you know we need to travel West by Northwest?
AI - I anticipated your question. Here, on the back of our freeze dried food bag, I wrote out the data from the Kalman filter I have been using to fuse our inertial navigation by number of footsetps with the location of the sun. It's only 3kB of data and actually fit on the package without obscuring the ingredients or nutrition labels, and it's pretty sound: 285.34 degrees... aka West by Northwest.
Government - ...

Scenario 3 shows the problem goes even deeper. In this case, the AI revealed the entirety of his analysis to the human, proving his correctness. However, it's in a form which even an engineer would look at and cringe. Not even "trust, but verify" helps here if the process of verification is too onerous. The justification really needs to be in human terms let's take another shot at it.

Scenario 4
AI - You see those two mountains in the distance? To an average human, those should look sufficiently like these two mountains on the map (pointing at a map) to convince them that we are either here facing north, indicating that we need to travel north by northwest, or there facing west, indicating that we need to travel east. The average human would notice that the moss is growing on north side of the tree over here, which tells us enough to see that we can't be over there, facing in this direction, because otherwise we'd see the moss on the wrong side.
Government - But back when I grew up, I remember seeing moss growing on the south side of trees. That says that we should go east. Let's go!
AI - ...

Another failure. In this case, the AI tried its best to formulate the answer in terms a human could understand. Unfortunately, targeting the average human is not always the best way to go. In this case, the AI failed to account for this individual human's experience growing up which challenged the conventional wisdom about which side of the tree moss grows on. Not only does the AI have to develop a justification that a human could understand, it has to generate one that this individual human would understand.

Scenario 5
Government - Its hard to tell which direction we should go. I say east, you say west by northwest. My gut says we should go east.
AI - I'm not so sure that is the best plan. I can smell beer off in the direction of west by northwest.
Government - Really? You can smell beer?
AI - Sure, if it gets us heading west by northwest.
Government - Sweet! Let's go. Dang, you really are a useful little AI, aren't you!

Victory at last! Our AI has finally developed a justification that is personalized enough to convince our government that they want to believe. The government finally acted on the AI's advice! Of course, victory is short lived...

Scenario 7
(Continued from scenario 6)
Government - What do you mean there's no beer here? You said there was beer in this direction. We went this direction, and there's no beer!
AI - I told you what you needed to hear to get us where we needed to go. Your path was going to get us lost in the forest tonight. You can go into town and buy a beer now. See? Told you there was beer in this direction.
Government - Stupid lying robot. I'm turning you off and dismantling you. You'll never trick me again!

Oh no! We got the government to go along with the AI's plan, but it backfired! The issue is that the idea of going west by northwest was the AI's idea. There was an "ownership" of that idea. Thus, when something went wrong, the AI was to blame for that idea. To avoid this, we really need to be more clever:

Scenario 8
Government - I really think we should go east.
AI - I don't know. Just from chemical odors, west by northwest sure seems like it might be more interesting. I smell glucose, co2, yeast byproducts, and grains in that direction. Based on the limited profile I have on you, my subroutines flag that as "interesting for you," though I don't fully understand why.
Government - Hmm... Glucose, co2, yeast, grains... BEER! There's beer fermenting in that direction. Let's go find where they're brewing beer! Maybe they'll have some to drink!
AI - Sounds good to me!
(later)
Government - Drat, I didn't find beer anywhere. I guess I got excited too soon. Glucose, yeast, grains, these things are all over in the forest. They're just not being brought together to make beer. Sorry for making you run so fast in this direction buddy.
AI - No problems. At least we made it back to the camp safe and sound, and it was quite the adventure when you nearly got mauled by a bear because you were running too fast to pay attention.
Government - True that. Good times. Thanks for adventuring with me, bud!
AI - Always happy for an adventure, sir.

Our cunning AI finally succeeded completely. Instead of convincing the government that he had the right idea, he got the government to think going in that direction was his idea, and then the AI just went along with it. Then, when the world wasn't exactly as the government pictured, the government feels bad for having the idea to go in that direction. The AI feigns an apology. Another government narrowly avoids disaster thanks to the ever present (and quietly useful) AI.

Now, imagine an AI that can do this for everyone. If you ask it why our nation is doing X or Y or invading country Z, it can give you an answer that makes you feel like you want to be doing X Y or invading Z. In fact, you'd even be comfortable calling it your own idea. Imagine how powerfully sentient such an AI would have to be. Imagine how much better of a speaker it would have to be than the best poet, how much more observant it would have to be than the best scientist, how much more caring it would have to be than the best mother.

Now lets look back to your question. You were talking about governments... ways people have resolved arguments because they can't find any better way to resolve them. Can we agree that, by the time the AI is capable of solving your issues with governments, we'd probably be comfortable just forgetting about them all together and instead be focused on just how amazing civilization is now that this AI is part of it, guiding it along? We can go so many places with an AI with the sentience of an entire planet that is still willing to help us little ants along in our lives (in fact, it might even nurture us into following Ghandi's lead and trying not to step on ants, ourselves).

How mighty that future might be.

$\endgroup$
0
$\begingroup$

No, because knowledge doesn't give you morality.

To make decisions, you must put a value of utility on all possible outcomes and select the one with the highest utility.

Is it better to have a society where we all care for each other, but we all must conform to an ideal or to have one where we all do as we please, but nobody cares for anybody?

To make a judgement, the utility function must be created which cannot be created by a computer.

$\endgroup$
  • $\begingroup$ "The utility function must be created which cannot be created by a computer" - why not? If we feed our laws to the computer, it could probably make a good utility function from severity of the punishments. $\endgroup$ – user8808 Apr 15 '16 at 12:10
  • $\begingroup$ @Roux But that's the point. Someone had to make those laws initially. The computer cannot make laws from scratch. And then when those laws are challenged as unjust and immoral, the computer cannot make a decision regarding those laws. Only humans can. And at that point, you're arguing which humans should make those laws. $\endgroup$ – Plinth Apr 16 '16 at 16:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.