Why widespread electronic surveillance is pointless

he thesis here is that while widespread surveillance is possible, it is fundamentally incapable of penetrating defenses that even computer novices can deploy will minimal effort. It therefore follows that if there are individuals that are intent on harming the US or its interests, and they have the requisite knowledge and resources to be harmful, widespread electronic surveillance will be ineffective in stopping them.

The following bullets detail barriers that prevent widespread surveillance from being effective:

  1. The basics of what is knowable: If one assumes that everything that is going across the internet is visible to US government, in order for that to be useful to preventing an attack of some kind, the information must be linked to a real person. Assuming that none of the information that is being transmitted contains the identity of the parties for certain[1], the first step in finding the parties is the location information. This is mostly going to be impossible, as while it would be possible for the government (or an ISP for that matter) to determine who the intermediaries are, they have no way of knowing if those intermediaries are merely acting as proxies, or are the actual parties to the communication.
  2. The above is rather permissive in assuming that it is possible to read everything that is going across the internet (assuming that they have some type of on-site access). Encryption is a way to prevent anyone other than the intended recipient from being able to view the contents of a message (including the sender). The mathematics of such a system have been carefully examined for decades, and the fundamental system is secure. Even if the government was years ahead of the rest of the world, and we assume that they have access to the combined power of the world’s top 500 supercomputers (around 223.7 petaflops)[2] it would still require about 101198 times the age of the universe to break a 4096 bit key (easily possible to create on even a small old laptop today). Quantum computers will make cracking some types of encryption easier, but there are forms of encryption that are resistant to that as well (and quantum systems have not been demonstrated at any appreciable scale).
  3. Another issue is that it is hard to find the data that is actually useful when so much is being collected. The proof for this can be found in numerous articles that claim that the NSA stores all the data, and then when that have a lead on something, they then go back to their massive data store and retrieve all of the relevant information. While this is a nice capability to have, the first problem listed does not go away. How do you know what user that you have logged is the one that you are looking for? Again, making the assumption that some protection has been employed (a relatively trivial task)[3], there is still no way to map all network information that has been collected to a person (or vice versa).

While each of the above three points excludes a significant amount of technical detail, and dumb users can make everything easier to track, anyone with a modicum of interest in keeping their online activities truly confidential will be able to. As an aside, all of this technology was developed far in advance of the NSA data collection leaks. That one is able to circumvent this type of surveillance is not news. It has always been possible.

[1] Even if information that is being sent contains the name of the person, this is not guaranteed to be true. Anything could have been typed in there. Additionally, credit card numbers could either be fake, or be anonymous gift cards that serve the same purpose.

[3] Tor browser, any one of a number of VPN or proxy services, etc

Fairer Accommodations for all

The general view of educational accommodations is that they level the playing field, and therefore, to qualify, one has to pass some educational tests that show that a “disability” exists. The problem with calling it a disability is that being disabled does not necessarily have anything to do with outcomes (unlike say, physical disabilities). It is entirely possible for someone that is “disabled” to do better on an exam unaided than someone that is not. In this sense, many of these “disabilities” are unlike blindness, hearing loss, and other physical handicaps, and more akin to a general inability to do well in school.

Given that stupidity and ADHD both have large effects on educational outcomes, it would seem that they should receive the same type of treatment, but for someone reason, society only considers the impact of ADHD worth dealing with, and not stupidity. This is not to say that low IQ people cannot be educated, but rather than it simply takes more effort. It would therefore seem reasonable to have an accommodation that essentially says: “For every 10 points of IQ that you have above 100, you get <arbitrary number> of minutes less in class, and for every 10 points of IQ that you are below 100, you get <arbitrary number> of minutes more in class.” Given that IQ is centered around 100, this is guaranteed (in the aggregate) to have the same aggregate instruction time as giving everyone the same amount of class time.

While tracking does have some disadvantages, on the whole, they mostly seem to be implementation details. Making the tracks more fluid where possible would solve many of the problems that develop over time, and randomly assigning teachers to teach different classes would solve the problem of less skilled teachers teaching the lower tracks. Additionally, even though there would be a fair number of tracks (~6 would cover 95% of the population) given the sizes of high schools (average grade size for all levels is 100+, even excluding people below 85) so it can be done or the grouping sizes can be reduced (the US army has an standard of about 85 as below that level one is not capable of soldiering, so the range can be reduced to 85 to 130, with people below that going elsewhere).

Aside from the absence of tracking as an accommodation, there is a larger problem with accommodations as a whole, namely that they aide people that don’t have a problem in addition to those that do. Examples of what I would call “fair” accommodations are things like textbooks in braille. This is something that helps blinds students, but for sighted students confers no advantage. Other examples include getting directions read aloud, large type on tests,  and auditory amplification devices.

On the other hand, there are many accommodations that will aid just about everyone who receives it, disabilities or not. The simplest example of this is extra time. While there may be students with a “disability” whose problem is solved with extra time, many non-disabled students would function better with extra time as well. Other examples include medication (those for ADHD and ADD come to mind, but there are probably others), and use of a calculator.

Additionally, the number of people that are getting these types of accommodations is growing dramatically. According to an article in the NY Sun, 25% of students are getting some measure of accommodations on exams, and by high school 20% of males have been diagnosed with ADHD (not including anything else, like ADD) . While it is true that not all of these are of second type of accommodations, a large enough number are, to the point where “That’s a percentage which is large enough basically to invalidate the test.”

As a solution, I would propose getting rid of either all type two accommodations, or granting them to everyone that asks for it (and making a note of that in the transcripts, SAT score reports, etc). The simple reason for this is that it would create a true level playing field. While there are those that say that the field would only be level with the accommodations (extra time can’t really be called extra if anyone can get it), the reality is that people are inherently unequal. There will be some people that are better than others in some or all subjects no matter what the respective effort is. This fact will not change. It will not go away when accommodations are granted. It simply grants a boost to some selected population (and one could argue that it is selected poorly).

Instead of trying to help solely those with proven “disabilities” we should try to help everyone. Given the educational attainment rate in this country, we clearly need the help, and it is the fairest thing for all students.

Second Point

The other issue is that of disclosure. It is not just that students are not on the same playing field, but that everyone else (other than the schools) are unaware of how the field is arranged. It is not as if two people are performing the same, one can imply that they are of equal ability. There is currently no way to make that determination. As such, no matter what happens with point one, there should be nothing that stops colleges or employers from seeing what accommodations were used to obtain the transcript/SAT scores that they see (note that this is not the same as sharing the disability itself, but rather its mitigation).

Gun control and reality

With the advent of guns that can be made of stock parts and a 3D printer the question should be asked: “What can be done to prevent anyone that wants one from getting a gun?”

The answer is simply put, nothing. This does not mean that I am a supporter of gun rights as they currently exist, but at the same time, I recognize a reality in which guns are plentiful and easy to acquire  currently, and with the ability to print key parts of guns, this will only get easier.

There is no way to conceivably restrict either 3D printing technology (it is too useful, easy to recreate, and multi-purposed) nor the files that can be given to the printer to create the pieces given the raw materials. The MPAA has been fighting a similar battle for years, and their problem is comparably well defined. They know exactly what content they are looking for. With the state of technology today, there are many many distinct ways to create designs for these parts, they can be hosted in many formats, in ways that would be difficult to determine that they were all the same blueprint algorithmically (if the language that is used to create the blueprints for the printer is sufficiently complex, it could easily be impossible (this would be the case if the language was c++ for example)).

Since the technology side cannot be conceivably regulated, another thought would be to regulate the raw materials that would go into the gun pieces. Unfortunately, this too is impossible, as all it takes is plastic, and some simple metal pieces available at many hardware stores.

The question then becomes, “How can we prevent murder from being commonplace?” I believe that the answer lies in a fundamental misunderstanding of how guns kill people. According to the CDC, 31,672 people were killed by firearms in 2010. While this number may seem high, for a sense of perspective, more people were killed in car accidents. Additionally, more than half of the people that are killed by firearms pulled the trigger themselves. While this is very sad, this is not going to influence the incidence of murder by firearm. So we are really more concerned about the other 50% of people that were the victim of a shooting, accident, and the like.

Considering that in 2010 2,468,435 people died of all causes, one has to wonder whether our priorities are in order. Deaths by firearm excluding suicide accounted for approximately .64% of all deaths, and while these deaths are sad, in the scheme of things, this is not a large impact on society.

As terrible as it may sound, the same thing can be said of mass shootings. While they are tragic, the fact that a few tens of people were killed by a lone individual on the same day is not significant in terms of the death rates. It may make for many consecutive news broadcasts, but they don’t kill many people as a percentage of pretty much any other cause of death.

The government is attempting to regulate something that in the near future (or even today) will be impossible. While I appreciate the effort to solve the problem of mass shootings (and regular homicides by gun), the country has far far bigger problems. Perhaps we should stop worrying so much about gun control and focus on those.

While it might not be possible to prevent people from getting guns, it has been proposed by the NRA and others that a possible solution would be to have guards everywhere, this is both impractical and useless. The vast majority of homicides take place between people that know each other well. Unless someone is proposing that guards should be placed everywhere (20% of the homicides are between family members) there is nothing that can be done. Additionally, having guards brings with it its own problems (can a guard shoot in time, or will it only be reactive, in which case in most scenarios it is too late anyway, or considering the accuracy of the NYPD what if the guards kill more people than the murderer ever would).

The elderly as teachers

There are at least two large problems in this country regarding education:

  1. Lack of good teachers
  2. Lack of jobs for the elderly

Given that people are working longer than ever, and retirement savings are very low, it would make sense to have jobs that have short working days (the average public school day is less than 7 hours long), low amounts of physical labor required (a given for almost all teachers), and breaks. Furthermore, if the demands that re made on a regular teacher are too demanding, it is relatively simple to lower the workload of an individual by having them teach less. As far as the elderly making better teachers, this is probably will not be true in high school as the material there is much harder to teach and remember for long periods of time, than the material for say, 5th grade english. The question is not whether they know the material (at least as well as the younger teachers), but whether they can teach it. While research on this point is lacking, something that bears mention is that teachers in the US apparently work far longer hours than their peers in other countries, with rather dismal results.

While it is possible that the elderly could be far worse teachers than the current teachers, I would tend to doubt it given that older people are stereotypically (research is lacking unfortunately) more patient than their younger peers, more caring, and have less conflicting priorities (generally speaking, their lives are more settled and less hectic). It is also worth noting that “and non-education majors performed just as well in the classroom as education majors” and so the lack of specific training is likely not to be severe.

Additionally, since their average workload will be lower (or even much lower) than that of regular teachers, Illness and death, and similar conditions are a legitimate concern, the odds of a teacher dying (assuming that they are of average health, and they will probably be of better than average health when they start) are relatively low (less than a 3% chance a year). Considering that teacher attrition is currently about 8.4% (not counting teacher turnover due to mobility), this risk is insignificant.

Since older people do not have the same career aspirations as their younger peers, they need less job security, less union protections, and the like. Because of this, they can be used as a more flexible workforce than regular teachers, and can therefore be more fully utilized.

Lest people accuse me of trying to take jobs away from career teachers, I would simply say that the only reason to allow people to be career teachers is if they get better with time. If they don’t, then there is no logical reason to give them raises, tenure, seniority, or any of the other perks that come with being a teacher for a long time. According to some research detailed here and here while teachers get better during their first 3 years, it is not that large (.04 standard deviations). While it might be statistically significant, its not significant in its impact on kids. Most tellingly, any sources that I could find that seemed to put a positive spin on the effect of experience failed to quantify the effect.

In light of this information, and given how much of teacher’s salary is based on experience, I am proposing that there is no reason to defend career teaching as a good  in and of itself. In light of the data, teachers get an average 3% raise year over year (mean: 3.08%, median: 2.92%) and this is for at best, a .04 standard deviation increase in performance. There is no good economic reason to pay for this kind of performance. If a drastically cheaper, similarly efficient labor supply can be found, as a society, we have a duty to use it.

To those who say that children deserve the best education at any cost, I have just two things to say:

  1. Throwing Money at Education Isn’t Working is a paper that distills the facts about the rise of education spending in this country, and how it largely has had no impact, so saving that money so that it can be used for better things (or not raised at all) would make sense
  2. At any cost is a funny thing. That means that in principle, if there was a way to raise student achievement by a few tenths (or hundredths) of a standard deviation at the cost of hundreds of millions if not billions of dollars, it should be done, no matter what the tradeoff is (trillions for a 1% gain? sure!). This kind of thinking is simply not rational. There is clearly a tradeoff in terms of what is acceptable, and a better metric to determining the value is perhaps the increase in salary later in life. While research has shown significant an ability of teachers to make a dent in student earnings (to the tune of $9285 over a lifetime in present day dollars, when comparing a teacher in the bottom 5% to an average one) this is not a majorly significant number (remember that these are lifetime totals, not yearly losses). Additionally, while a bad teacher has this kind of impact, it is not symmetric (one cannot achieve the same gains again by replacing an average teacher with one in the top 5%).

For all these reasons, encouraging the elderly to enter the teaching profession is a good idea for society, though perhaps not for current teachers.

Why programming is so great and rote education is terrible

As a programmer, I am slightly biased on this issue, but the more I try to learn about different subjects (history, politics, the sciences, etc) the more it becomes clear to me that  programming is a very useful skill to have.

The first thing that it does is force me to think in a logical way, and express myself clearly and unambiguously. This is a quality that easily transfers to other domains, and makes working in those domains simpler.

Secondly, it is useful for doing some cursory research. Most of the time when I find a document or documents that contain some statistics that I am looking for, they are in some kind of annoying format, with lots of extraneous information and much of the information that I am looking for needs to be calculated. The standard way that most people solve these problems is lots of manual copy and pasting. This works, but it quickly becomes tiresome.

This is a classic example of when a little programming can come in handy. There is no need (in most cases) to go to a fully fledged language like C++, as macros in Word, regexes in Notepad++, and Excel formulae go a long way (Google Refine is also very cool, and is actually designed for this, but its not as common a product). Also a useful thing is the ability to take many files, and turn them into one file, and IDM or DownThemAll (firefox add-on) to download many files all from the same page.

Combining all of these tools makes it simple to compile large amounts of data and do useful things with them (summing, averaging, getting statistical variations, and the like), and allows me to come up with bespoke data to research claims.

The problem is that it is hard to teach this kind of thinking with rote education, as it is all about what data sources you happen to come across on a particular day. Additionally, while rote learning may make some people great a task in isolation, it often makes it harder to abstract away parts of that task and combine those ideas with another task that was learned in a rote fashion.

Additionally, while the rote learning makes sense in some cases (vocabulary words, anatomy, etc) in those cases, the internet is greatly supplanting the need for memory. If you don’t remember the definition of a word, all you have to do is google it (or use any one of many free services online). While there is some small amount of vocabulary that is needed to get by, past a certain threshold, it no longer makes sense to make an active effort to memorize complex words.

As technology becomes more pervasive, this threshold will only get lower. Consider that in the past, one needed to find a dictionary (takes minutes). This was made easier by being at a computer (electronic dictionary), and even easier with mobile phones with web access (many people carry it around all the time). If augmented reality ever catches on, it will be instantly able to do translation (Google Glass can already do this) and I imaging that word lookups are even easier. There will be little need to memorize words then. The question becomes, what makes you better than wikipedia+google+ half a dozen other websites that are littered with knowledge.

I think that the answer is that people still do a lot better an analysis than machines. We can determine what is important, connections between data, causation, and many other tasks easily where the same problems have been giving natural language researchers headaches (like sentiment analysis). To turn people into things that machines can easily replace is not something should be desirable. We should strive to excel where they cannot, and leave the things that computers are good at to computers.

A more meaningful way to vote

Strange things have been shown to correlate with the odds getting elected, that ideally should have nothing to do with being a good leader/politician (unless being tall makes someone a better leader/politician).

Given that the US is a democracy and that is a good way to pick a government (that may be the topic of a later post), it would be better if these irrational biases are eliminated.

Here are a few ways to eliminate the biases:

  1. Don’t put any names on the ballot. Instead, create lists of positions on the issues for all the candidates, and then display all of the candidates on some of the issues (requires a computer screen). Have it display a random selection of the issues to every person that votes, with the option to display more. This would prevent people from voting for a candidate that they know nothing about. As a downside, this will make voting take longer, but considering the impact of voting, it would be worth it to extend the voting period to a few days or even a week (this would also raise voter turnout)
  2. Allow people to split their vote into number of candidates+1 parts. This would allow people to better express the way that they feel in more elections, and to allow third party candidates more of a chance to get a vote.
  3. Prevent candidates (or people promoting them) from lying, or being deliberately misleading, or assess some kind of penalty. The problem with this idea is that it is most probably unconstitutional, and if the penalty is meaningful enough, it may be worth it for the opposition to promote lies about a candidate.
  4. Throw bogus candidates onto the ballot (mixed with number 1) to make it a roster of lets say 8 (the number is arbitrary) total candidates. This will a number of things:
    1. Encourage more non-mainstream candidates. This will happen as if slots are not taken, then bogus candidates will appear, and so getting votes for minority parties will be easier.
    2. If people vote for a candidate that does not exist (i.e. randomly picking a person) then they are not really making a choice anyway, and discarding their vote is the right thing to do.
    3. Encourage people to learn more about the candidates. If they only know one point of difference between the candidates, there is now a risk that they will pick the wrong thing on the ballot. In order to make sure that their vote will go towards the person that they want, which will lead to a better educated electorate
  5. Use something like STV, though in general that works better when there are more candidates. This will also encourage more centrist candidates, as there is no downside to picking as a first choice a candidate that can’t win.

These ideas are not mutually exclusive, and some of these ideas are implemented in other places, but I think all of them would make the voting process more meaningful.

The end of full employment

In the past, as technology has come around, society’s response has been to make people more educated, get them into the workforce later, and then get them a higher skilled job.

This works when the level of education that is required for a productive life has grown to the point where is it hard for many people to reach that level. It is clear that the unemployment rate for high school dropouts and people with just a high school diploma is much higher than those with a college degree (2 to 4 times higher in fact). What goes unstated is that the educational attainment of the average high school graduate is not at a 12th grade level, or even close. To give a sense of scale, on the NAEP (National Assessment of Educational Progress) an advanced 4th grader knows more than a basic 12th grader. Additionally, the average score on the NAEP is less than proficiency by a hefty margin for 12th graders.

It seems that, at least under the current educational system, there is a limit on the maximum that people can really learn, for whatever reason. It seems unreasonable to expect that people will be able to actually read, write and do math at a 12th grade level. That means that there is a limit to the jobs that many people can get. As automation moves forward, the number of jobs that people who cannot read, write and do math at a 12th grade level can do will shrink. At a certain point, the number of jobs that these people can do is will be significantly exceeded by the number of people.

As far as room for growth, I don’t think that this will become a real problem in the short term, but I don’t think that this is a problem that can be solved in the long term. While the educational situation may get somewhat better, I don’t believe that people can be educated  to successively higher levels. At some point, people will not be able to be educated enough, and will be unable to get jobs.

The day is coming, the question is, what happens then?

The right to make bad choices is not absolute

The primary problem that most people have with Mayor Bloomberg’s ban on large sodas (and the ACA for that matter) is that it turns the state into a “nanny.” The problem with that argument is primarily that the state already does that.

Common examples include things like seat belts for the driver. Most people agree that this is a good idea, but at its core, the goal is simply protect a person from themselves. There is no “public” interest that has to be protected, just the people from themselves.

The problem is that people have a terrible idea of what is good for them, and more specifically they have consistent problems thinking for the long term. When this manifests itself in the savings rate, credit card debt, and seat belts for that matter, the state steps in (as much as legislators are comfortable with anyway) and they protect people from themselves.

The case of health insurance and the soda ban is even more clear than the above examples. In these cases, the government the government is not only protecting people from themselves (people in the aggregate, not necessarily every individual) but people from other people. Assuming that soda really does make people obese (which seems to be well documented (cited by 1670)) it has real costs to society. Since many people that are obese are concentrated in the lower economic strata, Medicaid, and Social Security disproportionately have to pay for the additional treatment that is required.

The money does not come out of the ether. It comes from people’s pockets that are not obese, as well as those that are (and disproportionately those that are not). To claim that the state should both pay for treatment for a large number of self-inflicted ailments (again, in the aggregate) at the same time as tying its hands for fixing the root cause is absurd (a legitimate point of view is to say that the government should do neither of these things, but that view gets much less popular support).

To the claim that the government could pass a law that says that people have to eat an apple a day, I have one response: Don’t elect such people

Unethical human experimentation

In order for researchers to do human experimentation, they (in almost all cases) the research proposal must pass an ERB (ethical review board). This seems like a good idea, to prevent abuses.

More important than the ERB review is the consent form that all participants of the study get. In theory, it explains all of the risks, rewards, side effects, etc so that the person knows exactly what they are getting into when they agree to be part of the study. Problems can arise when people are willing to consent to something, but the ERB feels it is unethical. The first case of informed consent was with Yellow Fever. Healthy people were willing to get themselves infected with a disease was lethal about 30% of the time so that the disease could be better understood and fought. Such a thing would most probably not be possible today because of an ERB. Similar experiments that would be useful are things that would determine is possible carcinogens are actually carcinogenic, faster turnaround for drug development, and any disease about which not that much is known (or usually comes with complications).

If people are willing to participate in these kinds of studies, why should we stop them?

Why many arguments are not worth having

Many arguments fall into one of three categories:

  1. Objective and known (it is daytime now)
  2. Subjective (vanilla is better than chocolate)
  3. We don’t have all the data, so we must reason with what data we do have

The only one worth arguing is number three. The result of the first one is simply to check (that could be a study or studies, looking outside, or anything similar). There is a right answer, go find it.

For the second one, you will never be able to convince a chocolate lover who is also a vanilla hater than vanilla is better than chocolate. There is nothing to start from, as its all subjective. Changing the subject to vanilla is healthier moves it into category one.

The third one is actually reasonable to discuss, but before attacking someone’s position, make sure that it really does fall into category three (and then keep in mind that just because it is a topic that can be argued does not mean that facts go out the window).