I think this over-states the benefit, and misses out on some of the costs/risks. Shit-tests are a classic proxy, and subject to all the caveats of any measurement which is not perfectly correlated with your actual desires. Goodhart is one of them—vaguely self-aware people will recognize and game the test. Another problem is that if the test is different from your normal behavior, you’re likely to see a different response than you would to your normal behavior. The differences will be correlated with just how different the test is from your baseline activities and signals.
Importantly, the testing is itself a signal. You mention this, but don’t mention that it’s anti-correlated with self-respect and competence, things which presumably you value in a partner. Early in a courtship, when you’re tempted to use this kind of test, is exactly when this test will drive away your best prospects. Later, when you know each other better, the test is less harmful, but also less valuable.
There probably are cases where a shit-test is justified—the time savings of fast-failures is worth the false-positives and additional friction that the artificial filter will create. But for many many cases (of romantic and other relationship-based exploration), you’re best off looking for natural experiments than intentionally creating stressful situations.
Thanks for bringing up Goodhart’s law. It’s a real problem. Rwanda girl probably did figure out intuitively that I wanted low NFC and respond to it. Fortunately for me, those attributes are hard to fake. Honestly I wouldn’t trust myself to shittest for fakable attributes like affection, loyalty, interest and social status.
Is testing anti-correlated with self respect and competence? That seems likely. If I had greater social intelligence I would shittest less because I would be more confident to assess attributes naturally. And I should assume my partners have better social intelligence than me.
There probably are cases where a shit-test is justified—the time savings of fast-failures is worth the false-positives and additional friction that the artificial filter will create. But for many many cases (of romantic and other relationship-based exploration), you’re best off looking for natural experiments than intentionally creating stressful situations.
I agree with this. I plan two shit-test per relationship. And in a different mating market I would do none.
Another problem is that if the test is different from your normal behavior, you’re likely to see a different response than you would to your normal behavior. The differences will be correlated with just how different the test is from your baseline activities and signals.
Good point! I don’t think it’s actually a problem when selecting long-term partners. You want to see the full spectrum of behavior, not just how they respond to you. Hence the “watch how they treat the waiter” advice.
Interestingly, people actually just give tons of unqualified reinforcement to partners during courtship (unconsciously of course). DF argues that unqualified reinforcement helps show you the full range of the persons behavior. This is a valuable adaptive behavior, because you want as much info about the person as possible. For example, unqualified reinforcement might reveal that someone is very selfish or prideful. I’ve just realized this probably works on me.
I think this over-states the benefit, and misses out on some of the costs/risks. Shit-tests are a classic proxy, and subject to all the caveats of any measurement which is not perfectly correlated with your actual desires. Goodhart is one of them—vaguely self-aware people will recognize and game the test. Another problem is that if the test is different from your normal behavior, you’re likely to see a different response than you would to your normal behavior. The differences will be correlated with just how different the test is from your baseline activities and signals.
Importantly, the testing is itself a signal. You mention this, but don’t mention that it’s anti-correlated with self-respect and competence, things which presumably you value in a partner. Early in a courtship, when you’re tempted to use this kind of test, is exactly when this test will drive away your best prospects. Later, when you know each other better, the test is less harmful, but also less valuable.
There probably are cases where a shit-test is justified—the time savings of fast-failures is worth the false-positives and additional friction that the artificial filter will create. But for many many cases (of romantic and other relationship-based exploration), you’re best off looking for natural experiments than intentionally creating stressful situations.
Thanks for bringing up Goodhart’s law. It’s a real problem. Rwanda girl probably did figure out intuitively that I wanted low NFC and respond to it. Fortunately for me, those attributes are hard to fake. Honestly I wouldn’t trust myself to shittest for fakable attributes like affection, loyalty, interest and social status.
Is testing anti-correlated with self respect and competence? That seems likely. If I had greater social intelligence I would shittest less because I would be more confident to assess attributes naturally. And I should assume my partners have better social intelligence than me.
I agree with this. I plan two shit-test per relationship. And in a different mating market I would do none.
Good point! I don’t think it’s actually a problem when selecting long-term partners. You want to see the full spectrum of behavior, not just how they respond to you. Hence the “watch how they treat the waiter” advice.
Interestingly, people actually just give tons of unqualified reinforcement to partners during courtship (unconsciously of course). DF argues that unqualified reinforcement helps show you the full range of the persons behavior. This is a valuable adaptive behavior, because you want as much info about the person as possible. For example, unqualified reinforcement might reveal that someone is very selfish or prideful. I’ve just realized this probably works on me.