4 ms·

> Mentally healthy people will never be less lonely as long as they know they are talking to a model. This isn't an empirical claim; it’s a definitional loop.

> Mentally healthy people will never be less lonely as long as they know they are talking to a model. This isn't an empirical claim; it’s a definitional loop. You've defined "mentally healthy" in a way that makes your conclusion true by default. It's like saying, "Only mentally ill people commit suicide, because a sane person wouldn't do that." You've smuggled your conclusion into your premise. It doesn't prove anything; it just circularly reinforces your bias. If someone who passes every clinical benchmark for mental health reports feeling less lonely after talking to a model, your definition simply reclassifies them as "not mentally healthy" to preserve the thesis. That's unfalsifiable - Karl Popper would call it a pseudo-theory. If you want to know whether talking to a model can reduce loneliness in mentally healthy people, you have to measure loneliness directly - not redefine "healthy" so your preferred answer is guaranteed.

kylecazar 1y ago
"If someone who passes every clinical benchmark for mental health reports feeling less lonely after talking to a model, your definition simply reclassifies them as "not mentally healthy" to preserve the thesis." I would more readily accuse them of misusing the word "loneliness" than diagnose them as mentally ill. I would suggest it cured their boredom. If they developed a relationship with a model that they deem on par with that of a human, the mental health floodgates would indeed open, because that is objectively a delusion. The sentence you picked to zero-in on had context. I first asserted that loneliness is inextricably linked to humanity, and solved by human connection. That's a fine claim to take issue with (people do), but my statement about mentally healthy people flows logically from that premise. BTW, the mentally healthy qualifier was included to begin with because it's been shown that AI can be particularly dangerous for people with SMI. The vulnerable are more likely to fall for the illusion.

kylecazar 1y ago

"If someone who passes every clinical benchmark for mental health reports feeling less lonely after talking to a model, your definition simply reclassifies them as "not mentally healthy" to preserve the thesis." I would more readily accuse them of misusing the word "loneliness" than diagnose them as mentally ill. I would suggest it cured their boredom. If they developed a relationship with a model that they deem on par with that of a human, the mental health floodgates would indeed open, because that is objectively a delusion. The sentence you picked to zero-in on had context. I first asserted that loneliness is inextricably linked to humanity, and solved by human connection. That's a fine claim to take issue with (people do), but my statement about mentally healthy people flows logically from that premise. BTW, the mentally healthy qualifier was included to begin with because it's been shown that AI can be particularly dangerous for people with SMI. The vulnerable are more likely to fall for the illusion.