This is a special post for quick takes by dinosaue. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
As someone who is AuDHD and is relatively new to LessWrong, I've found that I apply Bayesian learning when navigating social interactions. I do not intuitively understand body language and social cues so I form hypotheses on observable behaviors to infer what someone is communicating.
I also think of myself as a reinforcement learning agent. My inward reward function is pursuing special interests and minimizing sensory sensitivities. I strictly adhere to rules as my inward policy. My inward action space is sometimes empty when I struggle with executive function.
I might also display a weak form of narrow intelligence. I can tell you a lot about my special interests and in the past when the domain was sufficiently narrow and I had lots of free time, I had an encyclopedic amount of knowledge on them. I also have a spiky skillset. I am very good at certain things, and very bad at other things -- just like how machines are (in relation to humans).
I've found that I might trigger a form of uncanny valley effect if you talk to me irl. I think this comes from having flat affect and alexithymia. This is similar to how AI writing is so easily spottable. They just have a form and cadence that doesn't match the human intuition.
I struggle with cognitive empathy, but I do feel lots of affective empathy. Basically, I do genuinely care about others. I mentioned before how my inward reward function was pursuing special interests and minimizing sensory sensitivities. My outward reward function is mostly altruistic. In practice though, I do not think my consequences[1] are always positive. As such, I try to avoid situations where it feels like my mistakes are catastrophic. Real life is often irreversible, but I'd argue a healthy mental state must treat its environment as episodic.[2]