Hi, I have been a lurker for a while here. I have never gotten well with people. Most of the time I overstep our boundaries or the person I am getting along with just ends up taking advantage of me. This has been a problem running in my family for atleast 3 generations and now, quite frankly, I am tired of it.
So, what I am thinking of doing is making an AI model to interpret other’s behaviour. See whether they are being bored in a conversation with me, what they are feeling and other things, whether our relationship is unhealthy, etc. I have explored affective computing for the same and I am making progress on it. The issue is that most of them seem to focus on identifying emotions and not analyzing relationships.
I was hoping that someone can guide me to more resouces for the same. Sorry in advance if this is the wrong community for this question or if it seems rude.
(I have edited the post because I realised I asked for resources on something else)
deleted by creator
words x 0 + facial expression + action x 3
Wouldn’t “words x 0” result in 0 every single time? That would make the formula “facial expressions + action x 3”.
Also, can you elaborate on how to use this formula and provide an example? It seems really interesting to me. I actually have a protocol that I use to assess other people’s intentions, disposition, and engagement the relationship with me.
deleted by creator
I’ve heard some good things about GPT4ALL. It’s basically chatgpt hosted locally. I would never trust openai with my private conversations with people but I think it would be fine to feed it to a local program.
I have heard about it, but I was looking for something more along the lines of books, someone elses work, datasets and other such stuff.
Given that you would need an exceedingly extensive, labeled dataset to start working with actions/behaviours/body language directly, the only realistic way to start would be with a language model. (And while these are sometimes decent, they’re absolutely terrible at other times (and there’s generally no way to tell which time is which).)
You can’t apply computers to humans and expect good results. Period. Human made text is not easily parametrizable; much less someone’s manner of speaking or bodily expressions.
And human relationships as a whole, where at least two beings fundamentally incomprehensible to the machine interact, within a larger societal context that is equally incomprehensible? Acting like you can fit it into a neat model is peak computer science hubris.
And, in the end, it would only reinforce the preexisting notions of what is “healthy” or how others are feeling. I’m not sure that it applies to autistics as well as it does to neurotypicals. And I’m not certain if it applies to them either.
