To me, experimenting with drugs seems like playing Russian roulette. It is safe and fun… until suddenly it’s not.
It seems simple. Just do your research, and only try the stuff that your favorite part of the internet says is safe. What could possibly go wrong? Did someone else try the same algorithm and hurt themselves? Oh, am I sure that unlike me, that person was stupid and did something wrong. That could never happen to me.
(Also, that other guy was probably at risk, and I am not. How do we know whether someone it at risk? We will know they were at risk, if it later happens to them, duh.)
“I have it perfectly under control. I could stop any time I want. I just don’t want to.” say people unfalsifiably.
*
I think there is a difference between “taking a specific drug X that some research showed to be safe” and “the general attitude of willingness to experiment with all drugs that some research shows to be safe”. I expect the latter to be orders of magnitude more dangerous, because it encourages you to push your boundaries until something bad happens.
Consider the epistemic trap: You try a drug X. Either it hurts you or not. If it did not hurt you, your trust in “trying drugs that seem safe” increases. So you try a drug Y. Again, it either hurts you or not. If it did not hurt you, your trust in “trying drugs that seem safe” increases again. Do you see the trap? You either get hurt, or you get increasingly convinced that you are invincible. At first, you needed the research to be replicated seven times in respectable journals with N at least 100, in order to believe that something is safe. After a while, two replications with N = 20 will seem okay. Then, maybe a convincingly sounding blog post, or a story from a friend.
The alternative is to also learn from other people’s errors. Like, stop doing something when someone else gets hurt doing that thing. Tell me honestly, are you going to use this strategy? Because if the answer is yes, you might stop right now. There are plenty examples of people who tried to experiment with seemingly safe drugs, and hurt themselves. Oh, they do not count because they were not rationalists? I could probably find a few examples in the rationalist community too. Oh, they also do not count because… whatever. OK, realistically, is there someone you would consider to be in your reference group? -- Note that merely having friends who experimented with drugs and turned out to be okay is literally the survivor bias.
*
...I am not really expecting to convince anyone; I just would feel bad about myself if I didn’t say the obvious, in a situation where I feel that the obvious is being ignored.
Doesn’t that prove too much? You can apply the same logic to, for example, using transport (each time you try to ride a train and it doesn’t hurt you, your trust in “use transport is safe” increases, so you try to ride a motorcicle). Or to sport (each time you play ping-pong and it doesn’t hurt you, your trust in “sport is safe” increases, so you try BASE jumping).
Just… when you decide what to try you should not use a mental category that includes both LSD and heroin. I don’t think it’s that hard.
Well, yes, people feel safe doing things that didn’t hurt them yet, and that is sometimes how they get hurt. It even makes sense from certain perspective: if the thing you randomly chose is better than average (evidence: you didn’t get hurt yet), it is on average better to keep doing that, rather than to randomly choose something else. But that is a very low bar.
You can get much better estimates if you look at your reference group. How many ping-pong players get hurt per year? Suppose you had 10 friends at school that liked to play ping-pong. Now that you try to contact them a few years later, you find out that 4 of them are dead, because the ping-pong ball crushed their brain, or something like that. It would probably scare me so much that I would stop… even if nothing bad happened to me yet. Even if I could make an excuse like “they probably held the racket wrong, that could never happen to me”. Similarly, if I knew in person someone who died in a plane crash, especially multiple people at different crashed planes, I would probably stop using planes, even if one never crashed with me (which is the evidence I cannot get without also getting hurt).
What is the right reference group for drug use?
When people say “I read a research, and it showed...”, they kinda imply that the participants in the research are their reference group. Which is nonsense in my opinion. Participants in a research are probably getting their drugs from a reliable source, carefully measured, probably have their health checked, and are not subsequently encouraged by the organizers to go try also something else. This is not what will happen to you, if you start experimenting with whatever seems safe after reading some research online.
(By the way, are we going to “trust the Science” on safety of drugs? What happened to the usual skepticism and comments about replication crisis?)
So who is your reference group? Rationalists who use drugs?
Here, I find it important to openly push back against the pro-drug messages, because if no one does, then it will seem like all rationalists use drugs, when in reality… okay, I am just guessing here, but I think it could be even 10% or less.
Why is that important? Because the risk is, to put it bluntly, the number of people who died divided by the number of people who tried. If we make it seem like everyone is using drugs, that makes the risk seem smaller, by making the denominator large. If there are, let’s say, 1000 active rationalists, and we know about 10 who overdosed on drugs, it makes a difference whether it is 10 out of 1000, i.e. 1%; or 10 out of the 100 who experimented with drugs (because the remaining 900 did not), i.e. 10%.
you should not use a mental category that includes both LSD and heroin
I agree. It would be interesting to know, how many people who started experimenting with LSD stayed with LSD, and how many moved on. I opened the Wikipedia page about Timothy Leary (my first mental association with LSD) and… surprise, surprise, “Leary continued to take a wide array of drugs (ranging from serotonergic psychedelics to the nascent empathogen MDMA and alcohol and heroin)”. So, apparently even the person who popularized LSD failed to keep the mental buckets sufficiently separate.
I’m interested in the specific subquestion on how you decide who is in your reference group. And even more specifically, can we make this decision such that we would have detected earlier the wave of medication abuse in north America. Would you mind to expand your thoughts on this point?
If your concern is whether people who tried some medicine X will keep taking it forever even if they do not need it, then the reference group is “people who took medicine X at least once”, and you try to find out how many among them still need it, and how many still take it.
If your concern is rather whether people who do need medicine X will buy it regardless when it is freely available at shops, the reference group is “everyone”, and you try to find out how many have bought it, and how many needed it. (Or if it is freely available in a city, then the reference group is “everyone in the city”.)
If there are different risk profiles e.g. for men and women, or educated people and uneducated people, you can reduce to that. But if you go too far, you will have not enough data (in the extreme case you end up in the group alone). There is a methodological risk of… choosing a subset when its outcomes are better, but keeping the large set if they are not. Or just having a very low N and getting lucky.
My concern is how I can construct a category withput knowing which construction method to select. Say my category is rationalist who use drug, does that mean anyone who did try something once upon a time? Does alcool count? Painkillers and meds? If I’m too strict, everyone and her grand-mother is testing addiction. Not useful. If I’m too relax, that’s a free pass for shamans, old rockers, wallstreet, and Ontario prime ministers who have been on drug for decades. So, how do you define rationalist who use drug so that’s a priori useful for either(you, the median occasional LW reader, the specific question of how we could have detected faster the present wave of complications from med abuse)?
As I say in the post, don’t be a dumbass. Do an expected value calculation. If it doesn’t pan out, don’t do it.
Some drugs are risky and might kill you. Some drugs are safe at even very high doses. Some can have long-term negative side effects. Some don’t.
The point of this post is to point at the ugh field around doing drugs and say “hey, that’s an ugh field, you should look into it, because some people do drugs and think it’s a good idea even after we discount folks who are addicts, and maybe it turns out you were wrong to dismiss certain drugs out of hand because you weren’t bothering to check the math yourself”.
Sounds from your comments like you did the math and in your estimation they still aren’t worth it. Great! But different people value different things and not everyone will get the same result.
1) People systematically underestimate the effect drugs have on them. This is anecdotal evidence, but I guess many people have an experience with a drunk person yelling at them “I am not drunk”, or know a junkie who insists that he has it completely under control, can stop any time he wants, and it has no impact on his life, while his family is like “since he started taking drugs, he abandoned all his former hobbies and friends, dropped out of school, and all he talks about are the drugs”.
You could interpret this as a revealed preference, like maybe from his perspective the life with drugs is indeed better than the life without drugs, his new friends better than his old friends, etc., even if his parents disapprove. But to me this seems like dramatically modifying your utility function from U1 to U2, where according to U2 your new situation is great, but according to U1 it possibly is not. I currently happen to be at U1, so I am judging things from the perspective of U1.
2) There seems to be big variance in reactions to (some) drugs. Like, one person takes 10 doses without developing addiction; another person tries once, gets addicted. Even the same person can have a different reaction to the same dose on a different day. This is further complicated with illegal substances, because of varying purity and additives. The anecdotal evidence about how safe it is to use substances may be the survivor bias, with evidence to contrary dismissed as “those people were at risk” or “they did something wrong” (which seems to imply that you are not at risk, and your probability of doing something wrong—even at the moment you are high—is negligible).
Here we could list people from the rationalist community who experimented with drugs and died. I am not in Bay Area, where the prevalence seems to be greatest, but I could still add a friend of a friend to that list.
3) There is a well-known tendency of people who start experimenting with one illegal substance to also try another. This is probably a side effect of illegality—if you drink alcohol or smoke tobacco, your alcohol or tobacco producer does not have an incentive to convince you to also try something else; your drug dealer however has a strong incentive to convince you to try something with a greater profit margin. If you do not interact with dealers directly, you should still expect peer pressure towards the same. So a full calculation of danger of X should also include a term for people who moved from X to Y. If you tell me that obviously X is totally unlike Y, I don’t disagree; I am just saying that this is what “you that didn’t try X” believes. There is a reason to assume that “you after using X” will believe differently. Similarly, “you after using X” will probably be more comfortable to try higher dosage, and to take the drug more often. A calculation that does not take this into account is unrealistic.
In summary, you say “do the math”, I say that the math is typically done incorrectly, based on filtered data, often based on reports of people who are notorious for being untrustworthy; plus there are a few people who were rationalists, did the math, and died regardless. If you still believe that you can do the math correctly, okay, maybe you are right, or maybe you are just overconfident, I don’t know.
To me, experimenting with drugs seems like playing Russian roulette. It is safe and fun… until suddenly it’s not.
It seems simple. Just do your research, and only try the stuff that your favorite part of the internet says is safe. What could possibly go wrong? Did someone else try the same algorithm and hurt themselves? Oh, am I sure that unlike me, that person was stupid and did something wrong. That could never happen to me.
(Also, that other guy was probably at risk, and I am not. How do we know whether someone it at risk? We will know they were at risk, if it later happens to them, duh.)
“I have it perfectly under control. I could stop any time I want. I just don’t want to.” say people unfalsifiably.
*
I think there is a difference between “taking a specific drug X that some research showed to be safe” and “the general attitude of willingness to experiment with all drugs that some research shows to be safe”. I expect the latter to be orders of magnitude more dangerous, because it encourages you to push your boundaries until something bad happens.
Consider the epistemic trap: You try a drug X. Either it hurts you or not. If it did not hurt you, your trust in “trying drugs that seem safe” increases. So you try a drug Y. Again, it either hurts you or not. If it did not hurt you, your trust in “trying drugs that seem safe” increases again. Do you see the trap? You either get hurt, or you get increasingly convinced that you are invincible. At first, you needed the research to be replicated seven times in respectable journals with N at least 100, in order to believe that something is safe. After a while, two replications with N = 20 will seem okay. Then, maybe a convincingly sounding blog post, or a story from a friend.
The alternative is to also learn from other people’s errors. Like, stop doing something when someone else gets hurt doing that thing. Tell me honestly, are you going to use this strategy? Because if the answer is yes, you might stop right now. There are plenty examples of people who tried to experiment with seemingly safe drugs, and hurt themselves. Oh, they do not count because they were not rationalists? I could probably find a few examples in the rationalist community too. Oh, they also do not count because… whatever. OK, realistically, is there someone you would consider to be in your reference group? -- Note that merely having friends who experimented with drugs and turned out to be okay is literally the survivor bias.
*
...I am not really expecting to convince anyone; I just would feel bad about myself if I didn’t say the obvious, in a situation where I feel that the obvious is being ignored.
Doesn’t that prove too much? You can apply the same logic to, for example, using transport (each time you try to ride a train and it doesn’t hurt you, your trust in “use transport is safe” increases, so you try to ride a motorcicle). Or to sport (each time you play ping-pong and it doesn’t hurt you, your trust in “sport is safe” increases, so you try BASE jumping).
Just… when you decide what to try you should not use a mental category that includes both LSD and heroin. I don’t think it’s that hard.
Well, yes, people feel safe doing things that didn’t hurt them yet, and that is sometimes how they get hurt. It even makes sense from certain perspective: if the thing you randomly chose is better than average (evidence: you didn’t get hurt yet), it is on average better to keep doing that, rather than to randomly choose something else. But that is a very low bar.
You can get much better estimates if you look at your reference group. How many ping-pong players get hurt per year? Suppose you had 10 friends at school that liked to play ping-pong. Now that you try to contact them a few years later, you find out that 4 of them are dead, because the ping-pong ball crushed their brain, or something like that. It would probably scare me so much that I would stop… even if nothing bad happened to me yet. Even if I could make an excuse like “they probably held the racket wrong, that could never happen to me”. Similarly, if I knew in person someone who died in a plane crash, especially multiple people at different crashed planes, I would probably stop using planes, even if one never crashed with me (which is the evidence I cannot get without also getting hurt).
What is the right reference group for drug use?
When people say “I read a research, and it showed...”, they kinda imply that the participants in the research are their reference group. Which is nonsense in my opinion. Participants in a research are probably getting their drugs from a reliable source, carefully measured, probably have their health checked, and are not subsequently encouraged by the organizers to go try also something else. This is not what will happen to you, if you start experimenting with whatever seems safe after reading some research online.
(By the way, are we going to “trust the Science” on safety of drugs? What happened to the usual skepticism and comments about replication crisis?)
So who is your reference group? Rationalists who use drugs?
Here, I find it important to openly push back against the pro-drug messages, because if no one does, then it will seem like all rationalists use drugs, when in reality… okay, I am just guessing here, but I think it could be even 10% or less.
Why is that important? Because the risk is, to put it bluntly, the number of people who died divided by the number of people who tried. If we make it seem like everyone is using drugs, that makes the risk seem smaller, by making the denominator large. If there are, let’s say, 1000 active rationalists, and we know about 10 who overdosed on drugs, it makes a difference whether it is 10 out of 1000, i.e. 1%; or 10 out of the 100 who experimented with drugs (because the remaining 900 did not), i.e. 10%.
I agree. It would be interesting to know, how many people who started experimenting with LSD stayed with LSD, and how many moved on. I opened the Wikipedia page about Timothy Leary (my first mental association with LSD) and… surprise, surprise, “Leary continued to take a wide array of drugs (ranging from serotonergic psychedelics to the nascent empathogen MDMA and alcohol and heroin)”. So, apparently even the person who popularized LSD failed to keep the mental buckets sufficiently separate.
I’m interested in the specific subquestion on how you decide who is in your reference group. And even more specifically, can we make this decision such that we would have detected earlier the wave of medication abuse in north America. Would you mind to expand your thoughts on this point?
If your concern is whether people who tried some medicine X will keep taking it forever even if they do not need it, then the reference group is “people who took medicine X at least once”, and you try to find out how many among them still need it, and how many still take it.
If your concern is rather whether people who do need medicine X will buy it regardless when it is freely available at shops, the reference group is “everyone”, and you try to find out how many have bought it, and how many needed it. (Or if it is freely available in a city, then the reference group is “everyone in the city”.)
If there are different risk profiles e.g. for men and women, or educated people and uneducated people, you can reduce to that. But if you go too far, you will have not enough data (in the extreme case you end up in the group alone). There is a methodological risk of… choosing a subset when its outcomes are better, but keeping the large set if they are not. Or just having a very low N and getting lucky.
My concern is how I can construct a category withput knowing which construction method to select. Say my category is rationalist who use drug, does that mean anyone who did try something once upon a time? Does alcool count? Painkillers and meds? If I’m too strict, everyone and her grand-mother is testing addiction. Not useful. If I’m too relax, that’s a free pass for shamans, old rockers, wallstreet, and Ontario prime ministers who have been on drug for decades. So, how do you define rationalist who use drug so that’s a priori useful for either(you, the median occasional LW reader, the specific question of how we could have detected faster the present wave of complications from med abuse)?
As I say in the post, don’t be a dumbass. Do an expected value calculation. If it doesn’t pan out, don’t do it.
Some drugs are risky and might kill you. Some drugs are safe at even very high doses. Some can have long-term negative side effects. Some don’t.
The point of this post is to point at the ugh field around doing drugs and say “hey, that’s an ugh field, you should look into it, because some people do drugs and think it’s a good idea even after we discount folks who are addicts, and maybe it turns out you were wrong to dismiss certain drugs out of hand because you weren’t bothering to check the math yourself”.
Sounds from your comments like you did the math and in your estimation they still aren’t worth it. Great! But different people value different things and not everyone will get the same result.
Here is a summary of the problems I see:
1) People systematically underestimate the effect drugs have on them. This is anecdotal evidence, but I guess many people have an experience with a drunk person yelling at them “I am not drunk”, or know a junkie who insists that he has it completely under control, can stop any time he wants, and it has no impact on his life, while his family is like “since he started taking drugs, he abandoned all his former hobbies and friends, dropped out of school, and all he talks about are the drugs”.
You could interpret this as a revealed preference, like maybe from his perspective the life with drugs is indeed better than the life without drugs, his new friends better than his old friends, etc., even if his parents disapprove. But to me this seems like dramatically modifying your utility function from U1 to U2, where according to U2 your new situation is great, but according to U1 it possibly is not. I currently happen to be at U1, so I am judging things from the perspective of U1.
2) There seems to be big variance in reactions to (some) drugs. Like, one person takes 10 doses without developing addiction; another person tries once, gets addicted. Even the same person can have a different reaction to the same dose on a different day. This is further complicated with illegal substances, because of varying purity and additives. The anecdotal evidence about how safe it is to use substances may be the survivor bias, with evidence to contrary dismissed as “those people were at risk” or “they did something wrong” (which seems to imply that you are not at risk, and your probability of doing something wrong—even at the moment you are high—is negligible).
Here we could list people from the rationalist community who experimented with drugs and died. I am not in Bay Area, where the prevalence seems to be greatest, but I could still add a friend of a friend to that list.
3) There is a well-known tendency of people who start experimenting with one illegal substance to also try another. This is probably a side effect of illegality—if you drink alcohol or smoke tobacco, your alcohol or tobacco producer does not have an incentive to convince you to also try something else; your drug dealer however has a strong incentive to convince you to try something with a greater profit margin. If you do not interact with dealers directly, you should still expect peer pressure towards the same. So a full calculation of danger of X should also include a term for people who moved from X to Y. If you tell me that obviously X is totally unlike Y, I don’t disagree; I am just saying that this is what “you that didn’t try X” believes. There is a reason to assume that “you after using X” will believe differently. Similarly, “you after using X” will probably be more comfortable to try higher dosage, and to take the drug more often. A calculation that does not take this into account is unrealistic.
In summary, you say “do the math”, I say that the math is typically done incorrectly, based on filtered data, often based on reports of people who are notorious for being untrustworthy; plus there are a few people who were rationalists, did the math, and died regardless. If you still believe that you can do the math correctly, okay, maybe you are right, or maybe you are just overconfident, I don’t know.