7 ms·

And that means what? Data values are nothing without the exact context in which they were created and the exact context in which they are used. That's like lev

by gloflo 2y ago

And that means what? Data values are nothing without the exact context in which they were created and the exact context in which they are used. That's like level 1 data analysis. Publishing such data without context is deceitful.

mike_hearn 2y ago
How is it published without context? We know that this is the age field from the social security system. And that the query omits records that are recorded as dead. Therefore, the social security system in America has records that claimed to be for people who are alive for whom the date of birth field is incompatible with that status. That seems like quite a lot of context, actually. I don't know why some people are finding it so hard to accept that there is likely to be fraud in this system. Look into the determined origins of the so called blue zones to see that every country has problems with this, albeit some more than others. It's the government giving out free money, so naturally it attracts very sophisticated fraud schemes and civil servants are rarely motivated to track it down and investigate properly.
- troupo 2y ago
  How many of those are test data, simple clerical errors (and how many of those are already in the process of being rectified), and how many of those are in actual use (e.g. how many actually use those SSNs in the wild)? This is the important missing context. Musk can and does claim a lot. He rarely, if ever, provides any evidence or context. And none of his or his team's actions can be verified or monitored.
  - mike_hearn 2y ago
    He does. That's what started this thread: he provided evidence and context to back up his claims of widespread government waste and fraud. What that got him was a bunch of smears from people spreading misinformation about COBOL. Anyone else in the world would just stop sharing stuff publicly given this kind of public abuse. There's no requirement to do so. But Musk just shared evidence and context again, by showing the output of the equivalent of SELECT GROUP(age), COUNT(*) FROM SSNS WHERE DEAD = FALSE: https://x.com/elonmusk/status/1891350795452654076 https://x.com/elonmusk/status/1891350795452654076 It shows approx 1.3 million SSNs in the 150-159 age bucket, along with many millions more that are actually over that age, and over 1000 SSNs allocated to people over 200 years old, including one that is for someone marked as 360-369 years old. There are over 20M people listed as 100+ years old. Total sum is around 395M people. Therefore there is no 1875 epoch and that claim was simply misinformation. So here's what we know given his statements: 1. There are a lot of data entry errors in the SSN database. 2. These aren't test data, misinterpretations of an epoch, etc. 3. Such errors are extremely common and not being rectified. 4. Many SSNs aren't unique. 5. There are far more records in the database than people in America. If we combine those statements with prior knowledge of other related topics, we can infer: • The data quality issues are much more extensive than just age and number of records. • Whatever processes are meant to ensure data quality don't work. • This was not previously known to the public. • Civil servants know all this but are often unable to do anything. We also know that this situation is expected. It would be much more crazy if Musk announced his team couldn't find fraud in SS. Just look at the graph for payouts from the American disability benefits system - it tracks general economic performance. Other countries don't see such a thing in their payouts, where improving economic conditions magically make long term disabilities disappear, but they also tend to be more aggressive at cracking down on benefits fraud. The UK did a big purge some years ago where every single person claiming disability benefits was re-assessed. Anytime someone comes in and does basic checks of government finance systems they always find lots of very basic stuff. At one point it was discovered, again in Britain, that a Labour council was regularly paying invoices multiple times and nobody had noticed for years. Paying money out to dead or non-existent people is a common problem in all such systems.
    - camjw 2y ago
      Given the pretty large scale fraud here, do you expect people to go to prison for this? In the case that no-one is prosecuted and no-one goes to prison, how would you update your opinion on this story and on Elon in general?
      - Amezarak 2y ago
        Not speaking for GP, but even if this story is 100% true, why would you think anyone would go to prison? Obviously, nobody wants to clog up the court system with cases like this, and the people are probably not wealthy enough to fine or even claw the money back. I would just expect it to be terminated. Fraud like this is almost never prosecuted. The only exception I'd see is if there are people who have somehow gamed the system to collect multiple payments at once.
      - troupo 2y ago
        It's not a "large scale fraud". This is: https://en.wikipedia.org/wiki/Eric_C._Conn?wprov=sfti1 https://en.wikipedia.org/wiki/Eric_C._Conn?wprov=sfti1
        colejohnson66 2y ago
        His last name being "Conn" is very apt.
      - joshuahedlund 2y ago
        I think it’s important to maintain curiosity rather than assume “pretty large scale fraud” here. It seems very likely that the vast majority of these old records are not actually receiving active benefits, and even the few that are may have valid reasons (ex living younger spouses). https://xcancel.com/ThatsMauvelous/status/1891356192502399023?mx=2 https://xcancel.com/ThatsMauvelous/status/189135619250239902...
    - diffeomorphism 2y ago
      > 5. There are far more records in the database than people in America. Isn't that entirely expected? Someone coming to the US on a work visa gets assigned an SSN, works for some time and then leaves.
- contravariant 2y ago
  It all boils down to what you consider more likely. Someone not quite understanding the data they're looking at or that nobody noticed several tens of millions of dead people receiving social security. Whichever it is I would be very careful before making any grand public statements about it. And as far as fraud goes this doesn't sound anywhere near sophisticated.
  - mike_hearn 2y ago
    Those aren't the only two possibilities. I guess it depends on how much prior knowledge someone has of such problems. The correct explanation isn't "nobody noticed", it's "people noticed but could not/would not do anything about it". But how much you know about these problems will depend where you get your news from. If you never read right wing sources, you aren't exposed to stories about benefits/grants fraud as places like the Guardian or CNN just don't cover it, so all this will appear fantastical and absurd. If you're used to such stories then it all seems plausible and normal. These types of things usually have some aspects in common. One is that front line workers, when asked in a safe environment, will give enormous estimates for how much of their payout is fraudulent. The people who are in denial about it are usually the people at the top, not the bottom. They don't ask, nobody tells, they don't want to know because they'll be held accountable for it and they fundamentally, ideologically, do not care about waste or inefficiency, viewing it as preferable to even one innocent person not getting their money. I can't find it now but Musk has alleged similar things; he got private estimates from civil servants that 50% of the payouts from their department were probably fraud of some kind (it might not have been SS, I don't remember), and that the staff in charge of blocking payments at the Treasury were told to never do so. This kind of thing is what you'd expect. Often the front line civil servants are frustrated by this situation, as it's their taxes being wasted too. It's the people at the top who cover it up.
    - contravariant 2y ago
      And in what scenario would it be a good idea to give all defrauders a heads-up instead of going after potentially millions of them?
- notachatbot123 2y ago
  Context would be for example knowing how the code around the data handles those values. Some numbers in a database do not in any way imply real world effects.
- hyperpape 2y ago
  It does seem to me that logically, you have two choices: 1) Take the numbers at face value. In that case, you are predicting millions of accusations of fraud and an enormous number of prosecutions in the next year or two. 2) the situation is somehow more complicated, and most of those millions of records with 140+ ages do not represent fraudulent activity. P.S. Mentioning the blue zones is incredibly silly. Those regions have modest numbers of individuals being reported in the 100-120 age range, which probably are fraudulent. None of those areas have millions being reported to be 140+. For instance, Sardinia had 13 reported centenarians per 100,000 population, which would be equivalent to ~39,000 centenarians in the US. So it's orders of magnitude less than this database shows.
  - mike_hearn 2y ago
    Or the third choice: (3) The data quality is so bad that there is no way to reliably know if a case is fraudulent or not without an investigation so expensive the ROI is negative, so everything remains unclear and unresolved forever. That's usually how it goes with cases like these. Sometimes there are gangs who organize large scale fraud against the system and those might attract the attention of prosecutors, but people not reporting a death or double payment or similar isn't worth it. There might never be a clear answer to how the database got into this state. But the basic point stands that data quality for a critical dataset is really low in obvious ways, so what about all the non-obvious ways? I mentioned blue zones because there are a lot of people on this thread who are really having a hard time believing there can be problems as obvious as people who have died but continue receiving payments in social security schemes. Blue zones is just an easily searchable keyword to learn more about other times when it's happened at scale, e.g. In 2010, the Japanese government announced that 82 percent of its citizens reported to be over 100 had already died. In 2012, Greece announced that it had discovered that 72 percent of its centenarians claiming pensions – some 9,000 people – were already dead. Puerto Rico’s government said in 2010 that it would replace all existing birth certificates due to concerns about widespread fraud and identity theft. https://www.aljazeera.com/news/2024/9/26/the-secret-of-blue-zones-where-people-reach-100-fake-data-says-academic https://www.aljazeera.com/news/2024/9/26/the-secret-of-blue-... Obviously nobody is claiming the database reflects reality. Governments often have multiple data sources that are badly out of alignment. A census can give more accurate data, but that doesn't mean SS is synced to it. For instance, in the UK during COVID, more people came forward in some age ranges for a COVID vaccine than theoretically existed in the country at all. The UK's population data is so badly screwed up that people started using the quantity of NHS numbers issued instead to try and estimate it.
mmusson 2y ago
Another thing is that COBOL records commonly have complicated unions (to save space) where a separate code affects how you interpret the fields. You need to understand all the business logic to make sure you are reading the data correctly.

by gloflo 2y ago

mike_hearn 2y ago

How is it published without context? We know that this is the age field from the social security system. And that the query omits records that are recorded as dead. Therefore, the social security system in America has records that claimed to be for people who are alive for whom the date of birth field is incompatible with that status. That seems like quite a lot of context, actually. I don't know why some people are finding it so hard to accept that there is likely to be fraud in this system. Look into the determined origins of the so called blue zones to see that every country has problems with this, albeit some more than others. It's the government giving out free money, so naturally it attracts very sophisticated fraud schemes and civil servants are rarely motivated to track it down and investigate properly.

troupo 2y ago
How many of those are test data, simple clerical errors (and how many of those are already in the process of being rectified), and how many of those are in actual use (e.g. how many actually use those SSNs in the wild)? This is the important missing context. Musk can and does claim a lot. He rarely, if ever, provides any evidence or context. And none of his or his team's actions can be verified or monitored.
- mike_hearn 2y ago
  He does. That's what started this thread: he provided evidence and context to back up his claims of widespread government waste and fraud. What that got him was a bunch of smears from people spreading misinformation about COBOL. Anyone else in the world would just stop sharing stuff publicly given this kind of public abuse. There's no requirement to do so. But Musk just shared evidence and context again, by showing the output of the equivalent of SELECT GROUP(age), COUNT(*) FROM SSNS WHERE DEAD = FALSE: https://x.com/elonmusk/status/1891350795452654076 https://x.com/elonmusk/status/1891350795452654076 It shows approx 1.3 million SSNs in the 150-159 age bucket, along with many millions more that are actually over that age, and over 1000 SSNs allocated to people over 200 years old, including one that is for someone marked as 360-369 years old. There are over 20M people listed as 100+ years old. Total sum is around 395M people. Therefore there is no 1875 epoch and that claim was simply misinformation. So here's what we know given his statements: 1. There are a lot of data entry errors in the SSN database. 2. These aren't test data, misinterpretations of an epoch, etc. 3. Such errors are extremely common and not being rectified. 4. Many SSNs aren't unique. 5. There are far more records in the database than people in America. If we combine those statements with prior knowledge of other related topics, we can infer: • The data quality issues are much more extensive than just age and number of records. • Whatever processes are meant to ensure data quality don't work. • This was not previously known to the public. • Civil servants know all this but are often unable to do anything. We also know that this situation is expected. It would be much more crazy if Musk announced his team couldn't find fraud in SS. Just look at the graph for payouts from the American disability benefits system - it tracks general economic performance. Other countries don't see such a thing in their payouts, where improving economic conditions magically make long term disabilities disappear, but they also tend to be more aggressive at cracking down on benefits fraud. The UK did a big purge some years ago where every single person claiming disability benefits was re-assessed. Anytime someone comes in and does basic checks of government finance systems they always find lots of very basic stuff. At one point it was discovered, again in Britain, that a Labour council was regularly paying invoices multiple times and nobody had noticed for years. Paying money out to dead or non-existent people is a common problem in all such systems.
  - camjw 2y ago
    Given the pretty large scale fraud here, do you expect people to go to prison for this? In the case that no-one is prosecuted and no-one goes to prison, how would you update your opinion on this story and on Elon in general?
    - Amezarak 2y ago
      Not speaking for GP, but even if this story is 100% true, why would you think anyone would go to prison? Obviously, nobody wants to clog up the court system with cases like this, and the people are probably not wealthy enough to fine or even claw the money back. I would just expect it to be terminated. Fraud like this is almost never prosecuted. The only exception I'd see is if there are people who have somehow gamed the system to collect multiple payments at once.
    - troupo 2y ago
      It's not a "large scale fraud". This is: https://en.wikipedia.org/wiki/Eric_C._Conn?wprov=sfti1 https://en.wikipedia.org/wiki/Eric_C._Conn?wprov=sfti1
      - colejohnson66 2y ago
        His last name being "Conn" is very apt.
    - joshuahedlund 2y ago
      I think it’s important to maintain curiosity rather than assume “pretty large scale fraud” here. It seems very likely that the vast majority of these old records are not actually receiving active benefits, and even the few that are may have valid reasons (ex living younger spouses). https://xcancel.com/ThatsMauvelous/status/1891356192502399023?mx=2 https://xcancel.com/ThatsMauvelous/status/189135619250239902...
  - diffeomorphism 2y ago
    > 5. There are far more records in the database than people in America. Isn't that entirely expected? Someone coming to the US on a work visa gets assigned an SSN, works for some time and then leaves.
contravariant 2y ago
It all boils down to what you consider more likely. Someone not quite understanding the data they're looking at or that nobody noticed several tens of millions of dead people receiving social security. Whichever it is I would be very careful before making any grand public statements about it. And as far as fraud goes this doesn't sound anywhere near sophisticated.
- mike_hearn 2y ago
  Those aren't the only two possibilities. I guess it depends on how much prior knowledge someone has of such problems. The correct explanation isn't "nobody noticed", it's "people noticed but could not/would not do anything about it". But how much you know about these problems will depend where you get your news from. If you never read right wing sources, you aren't exposed to stories about benefits/grants fraud as places like the Guardian or CNN just don't cover it, so all this will appear fantastical and absurd. If you're used to such stories then it all seems plausible and normal. These types of things usually have some aspects in common. One is that front line workers, when asked in a safe environment, will give enormous estimates for how much of their payout is fraudulent. The people who are in denial about it are usually the people at the top, not the bottom. They don't ask, nobody tells, they don't want to know because they'll be held accountable for it and they fundamentally, ideologically, do not care about waste or inefficiency, viewing it as preferable to even one innocent person not getting their money. I can't find it now but Musk has alleged similar things; he got private estimates from civil servants that 50% of the payouts from their department were probably fraud of some kind (it might not have been SS, I don't remember), and that the staff in charge of blocking payments at the Treasury were told to never do so. This kind of thing is what you'd expect. Often the front line civil servants are frustrated by this situation, as it's their taxes being wasted too. It's the people at the top who cover it up.
  - contravariant 2y ago
    And in what scenario would it be a good idea to give all defrauders a heads-up instead of going after potentially millions of them?
notachatbot123 2y ago
Context would be for example knowing how the code around the data handles those values. Some numbers in a database do not in any way imply real world effects.
hyperpape 2y ago
It does seem to me that logically, you have two choices: 1) Take the numbers at face value. In that case, you are predicting millions of accusations of fraud and an enormous number of prosecutions in the next year or two. 2) the situation is somehow more complicated, and most of those millions of records with 140+ ages do not represent fraudulent activity. P.S. Mentioning the blue zones is incredibly silly. Those regions have modest numbers of individuals being reported in the 100-120 age range, which probably are fraudulent. None of those areas have millions being reported to be 140+. For instance, Sardinia had 13 reported centenarians per 100,000 population, which would be equivalent to ~39,000 centenarians in the US. So it's orders of magnitude less than this database shows.
- mike_hearn 2y ago
  Or the third choice: (3) The data quality is so bad that there is no way to reliably know if a case is fraudulent or not without an investigation so expensive the ROI is negative, so everything remains unclear and unresolved forever. That's usually how it goes with cases like these. Sometimes there are gangs who organize large scale fraud against the system and those might attract the attention of prosecutors, but people not reporting a death or double payment or similar isn't worth it. There might never be a clear answer to how the database got into this state. But the basic point stands that data quality for a critical dataset is really low in obvious ways, so what about all the non-obvious ways? I mentioned blue zones because there are a lot of people on this thread who are really having a hard time believing there can be problems as obvious as people who have died but continue receiving payments in social security schemes. Blue zones is just an easily searchable keyword to learn more about other times when it's happened at scale, e.g. In 2010, the Japanese government announced that 82 percent of its citizens reported to be over 100 had already died. In 2012, Greece announced that it had discovered that 72 percent of its centenarians claiming pensions – some 9,000 people – were already dead. Puerto Rico’s government said in 2010 that it would replace all existing birth certificates due to concerns about widespread fraud and identity theft. https://www.aljazeera.com/news/2024/9/26/the-secret-of-blue-zones-where-people-reach-100-fake-data-says-academic https://www.aljazeera.com/news/2024/9/26/the-secret-of-blue-... Obviously nobody is claiming the database reflects reality. Governments often have multiple data sources that are badly out of alignment. A census can give more accurate data, but that doesn't mean SS is synced to it. For instance, in the UK during COVID, more people came forward in some age ranges for a COVID vaccine than theoretically existed in the country at all. The UK's population data is so badly screwed up that people started using the quantity of NHS numbers issued instead to try and estimate it.

mmusson 2y ago

Another thing is that COBOL records commonly have complicated unions (to save space) where a separate code affects how you interpret the fields. You need to understand all the business logic to make sure you are reading the data correctly.