There's literally less of your brain under your volition.

Well, yeah. The primary worry among tulpa creators is that it might get pissed at you and follow you around the house making faces.

This tulpa stuff resembles mental illness.

And what, pray tell, is the salient feature of mental illness that causes us to avoid it? Because I don't think it's the fact that we refer to them with the collection of syllables "men-tal-il-nes".

Now, you wanna show off your "rationality" according to local rules of showing off your rationality, by rejecting the simple looking argument that it should be avoided like mental illness is. "Of course" it's pattern matching, "non central fallacy" and other labels that you were taught here to give to equally Bayesian reasoning when it arrives at conclusions you don't like. Here's the thing: Yeah, it is in some technical sense not mental illness. It most closely resembles one. And it is as likely to be worse as it is likely to be better*, and it's expected badness is equal to that of mental illness, and the standard line of reasoning is going to approximate utility maximization much better than this highly biased reasoning where if it is not like mental illness it must be better than mental illness, or worse, depending to which arguments pop into your head easier. In good ol caveman days, people with this reasoning fallacy, they would eat a mushroom, get awfully sick, and then eat another mushroom that looks quite similar to the first, but is a different mushroom of course, in the sense that it's not the exact same physical mushroom body, and get awfully sick, and then do it again, and die.

Wow.

EDIT: OK, I should probably respond to that properly. Analogies are only useful when we don't have better information about something's effects. Bam, responded.

Let's suppose it was self inflicted involuntary convulsion fits, just to make an example where you'd not feel so much like demonstrating some sort of open mindness. Now the closest thing would have been real convulsion fits, and absent other reliable evidence either way expected badness of self inflicted convulsion fits would clearly be equal.

"Convulsion fits" are, I understand, painful and dangerous. Something like alien hand syndrome seems more analogous, but unfortunately I can't really think of any benefits it might have, so naturally the expected utility comes out negative.

Also, by the way, what ever mental state you arrive at by creating a tulpa, is unlikely to be a mental state not achievable by one or the other illness.

Could well be. Illnesses are capable of having beneficial side-effects, just by chance, although obviously it's easier to break things than improve them with random interference.

if its self inflicted, for example standard treatments might not work.

If you had looked into the topic, you would know the process is reversible.