So it's theory time. I'm going to post most of something I posted elsewhere a while ago, for your reading (dis)pleasure. There is a summary at the end. The below is all my opinion and so on, but I can't prefix everything with "I think" without rendering it completely unreadable. And on readability, I've tried to break it up, and underline different theories for clarity.
All of these ideas assume a more or less conventional theory of mind, and none have been tested empirically as far as I know. Also, I know there are far more viewpoints out there that are't covered here but I think a lot of the big ones are.
Taking a black box approach, functionally, a tulpa is capable of conscious thought, or at least a very good imitation of it. In normal conversation they will generate responses to your communications in a way that is unconscious to you; not only are these responses convincingly sophisticated but also more or less consistent with a personality separate to your own. Moreover, this personality is not necessarily as expected by the host, as countless experiences of deviation and such will tell. You could argue that instinctive responses are not conscious and that this is all a tulpa can do, but I think experience does discount this too.
A lot of people say that [conversations with] tulpas are either (1) self-generated responses that we label as not us, or (2) entities (and responses) created entirely by expectation. I don't think these ideas make a whole lot of sense when considered with the previous paragraph in mind.
For example, if the second idea were true then a tulpa would not be able to differ from the host's expectations. More formally, for no conscious thought to occur on behalf of the tulpa their actions must be pre-calculated (somehow) and thus expected by the host, since otherwise these actions will have been generated unconsciously. However, tulpas can and do respond in ways that are not expected by the host, which disproves the idea. Given that, I think it's reasonable to say that tulpas do possess some sort of conscious capability.
For number one, you have the issue that if a tulpa is unconscious to you but generates conscious thought it is a stretch to say that it is still 'you' in the first place. If you hold onto the view then it becomes a semantic point of where 'you' end in your mind - and not, in my opinion, a sensible one - rather than a question of psychology.
This conscious capability does not have to be consciousness as in 'true', philosophical sentience, nor does it even have to resemble conscious thought inside the black box. Therefore, without opening the box I think there are two separate approaches to take:
The first is that tulpas do possess conscious thought. It goes rather like the commonly-stated idea that you "split off a part of your own consciousness", although not phrased that badly. The idea is that you make part of your neural machinery for conscious thought (this is hypothetical, sure) unconscious to you. This then becomes governed separately to you, i.e., while it still generates responses, it is no longer influenced by your personality and instead receives its own.
There are a few issues with this view, I think. Firstly, it supposes a distinct 'conscious thought' structure in the brain which is not only not wholly tied to your own awareness, but can be modified, non-trivially but without too much trouble, to form two separate structures that (at least one-way) do not interfere. Secondly, it supposes that this structure is also distinct from personality, such that it can be uncoupled from it or reassembled to exclude it in favour of another.
For these together I think there are two solutions. One is that you 'create' a new personality and thought structureyou merely personify existing aspects of your psyche. I said "psyche" so I think you know where this is going. Carl Jung's ideas about archetypes are quite useful here. Jung said that unconscious factors affecting our personality could be identified as essentially distinct sets. The main one of interest here is the anima; given that the majority of hosts are male I can be sexist like he was and neglect the others. Put simply, the anima is the feminine side of a man, but more precisely it is the influences that take effect when dealing with women. In personifying it and giving it voice, you appear to create a female personality which can react fundamentally as it did when wholly unconscious, but to you and in a more human-like manner.
Again, this idea is not without its problems; mainly, it does not account at all for hosts with tulpas of the same gender. An extension of the archetypes might account for this, but would not quite be in keeping with Jung's original theory: the masculine side of a masculine self was, of course, conscious. Nevertheless I prefer this model to the first, not least because it is grounded in accepted psychology but also because the actual creation - the personification - has a fairly strong precedent in said psychology.
The second is that tulpas do not possess conscious thought, but rather, an imitation or simulation of it. This approach in general is vaguer and therefore more justifiable. Plus, you can pick and choose unconscious processes to be a tulpa, so it's more fun to consider. My personal pet theory is that tulpas are products of empathetic machinery: you (need to) have a way to predict the actions of others and empathise with them, which requires some sort of estimation of their mental processes, done unconsciously. Expectation of a response from an as-yet unformed tulpa pushes your empathetic framework to estimate behaviour, as would happen in real-life interaction. Repeating this strengthens the structure, and so on, eventually being able to generate responses clearly and easily.
As always, there are problems, principally that people do not generally empathise with their tulpas as strongly as would be expected. Despite this the model fits with what works for creation as well as others do, and fits in the seemingly overwhelmingly powerful factor of expectation neatly. Moreover, the neurology of empathy is of much interest to the scientific community, and concepts such as 'mirror neurons' are of much help fitting the model in with the brain itself.
Another thing I want to mention is the oft-quoted "tulpas are like you but in your head"
Lastly, all of the above, as I said, supposes a fairly standard theory of mind. If we go beyond, we are given even more freedom at the cost of any substantiation. For example, we'll assume that rather than consciousness being active, it is merely a passive window onto 'conscious' thought. Now all that is required to create a tulpa is to shift this 'window' in such a way that we become aware of other, latent fields of conscious thought that we mark as existent and personify. It fits the facts, for sure, but that may be because it has essentially no grounding in anything approaching accepted theory and is therefore vague or malleable enough to fit anything. Nevertheless, such things are worth bearing in mind. It's not like accepted theory is all that strong in the first place.
To summarise, I prefer to keep options open and support a range of different theories, as none of them are without flaw and none of them have been tested empirically. I think highly of the idea of a 'personified unconscious', or 'thought simulation' models, and I dislike the conventional "another consciousness, end of story" idea because fast/multiple tulpas breaks it for me. I also don't think it's reasonable to say that "tulpas are just illusions", although the specifics of such propositions may render them more sensible in my eyes.
And Sands if you say "a miserable little pile of secrets" you'll be banned from this thread forever.
