3 ms·

It sounds to me like you are justifying a continuance of frauds that masquerade as scientific. The danger here is is not just direct effects but also what they

It sounds to me like you are justifying a continuance of frauds that masquerade as scientific. The danger here is is not just direct effects but also what they displace. One can not actually do psychological studies beyond the test of chemical remedies anymore. The entire field is overrun with charlatans and 101 classes to make more of them because they will all be treated as valid enough to fund and they are all much more exciting than the scientific process.

godelski 2y ago
Oh dear god no. You can check my comment history to see I rant about that frequently. I with in ML and dear god is there so much fraud. What I arguing for is honesty. It's perfectly okay to do work where there's a high level of uncertainty *UNDER THE CONDITION THAT YOU ARE UPFRONT WITH THAT UNCERTAINTY.* Short of that is at best deceptive. I'll encourage you to look at my recent comment history because I've said a lot about that. But what I'm trying to discuss in the previous comment is the systematic and structural components that encourage this deceptive and fraudulent behavior. Because if we don't recognize why it happens then we can't really solve it. Sure, we can hit a reset button and it'll go away, but it's we don't recognize the pressures that push the fields in these directions then it is just going to happen again. The reason I'm arguing that we have to take risks is because many of these pressures are created by naive metrics that attempt to measure things that are not realistically measurable (so of course the metrics get hacked. It's literally Goodhart's law in action). But there are so many issues at play. The lack of incentivization of replication means few people earnestly read the works of others. It means science doesn't actually happen because the whole fucking point is replication. People aren't even reading the papers of their peers in their own departments! Literally the person in the office next to them! The whole academic structure is screwy. You train someone for years to become an expert researcher (grad student) and then immediately put them in a manager role (professor)? The fuck were we doing that training for? I get managers, but with so many admins in universities why are professors doing less and less research? Advising grad students isn't research as every grad student will tell you their professor doesn't get details. I could go on. But my point is the system is fucked up and overrun by bureaucrats who are risk adverse. They'd rather rely on metrics they are unwilling to understand than acknowledge that they live in a world that is fuzzy and chaotic. That kind of blindness is for religion and conspiracies, not science.

godelski 2y ago

Oh dear god no. You can check my comment history to see I rant about that frequently. I with in ML and dear god is there so much fraud. What I arguing for is honesty. It's perfectly okay to do work where there's a high level of uncertainty *UNDER THE CONDITION THAT YOU ARE UPFRONT WITH THAT UNCERTAINTY.* Short of that is at best deceptive. I'll encourage you to look at my recent comment history because I've said a lot about that. But what I'm trying to discuss in the previous comment is the systematic and structural components that encourage this deceptive and fraudulent behavior. Because if we don't recognize why it happens then we can't really solve it. Sure, we can hit a reset button and it'll go away, but it's we don't recognize the pressures that push the fields in these directions then it is just going to happen again. The reason I'm arguing that we have to take risks is because many of these pressures are created by naive metrics that attempt to measure things that are not realistically measurable (so of course the metrics get hacked. It's literally Goodhart's law in action). But there are so many issues at play. The lack of incentivization of replication means few people earnestly read the works of others. It means science doesn't actually happen because the whole fucking point is replication. People aren't even reading the papers of their peers in their own departments! Literally the person in the office next to them! The whole academic structure is screwy. You train someone for years to become an expert researcher (grad student) and then immediately put them in a manager role (professor)? The fuck were we doing that training for? I get managers, but with so many admins in universities why are professors doing less and less research? Advising grad students isn't research as every grad student will tell you their professor doesn't get details. I could go on. But my point is the system is fucked up and overrun by bureaucrats who are risk adverse. They'd rather rely on metrics they are unwilling to understand than acknowledge that they live in a world that is fuzzy and chaotic. That kind of blindness is for religion and conspiracies, not science.