I think this framing makes it seem like you want to track the raw emotions of users and make decisions based purely on their emotions.

I strongly suspect it isn't the case. What you're really after is detecting user frustration or user fulfillment: That is a more complex emergence where emotions can be a symptom.

Reading those emergent signals is well within expectations. Users probably do hope you'll make their lives easier or better, that's why they're using your product :)

Another consideration is that, frankly: Reading emotions to make a business decision implies that your product is the user's entire world. Even for mentally engaging products involving something like therapy, it's not: You can have an angry user that loves your feature, and a happy user that hates it.

Emotions are messy, fluid, imprecise, and circumstancial. And in many people's opinion, simply nobody else's business -- It's a good question to wonder why, but as a baseline, that's just the way it is for many people.