What is an Emotion?
Here's what I'm working on: neural models of emotions!
So, this is what I've come up with for a definition of an emotion that I can work with:
Evaluative `functions`(OPTIONALLY ENFORCED PARAMETERS)->Emotion<name>(GAIN|DOMAIN|EVENT|CLARITY|ACCEPTANCE)
An emotion is a function that evaluates statements (even recursively), returning an "Emotion" object with universal parameters for machine identification and classification, which we reference by name (sadly, not a universal parameter, but that's okay, because it's all relative).
The "Orthographic Model of Emotions" now has an operational design, implementing what you can read over on Medium. Although the model was designed as a minimal set of operational parameters for the classification and interpretation of emotions, its primitive and simplistic design doesn't dictate any operations. The emotions simply exist, as far as the model is concerned.
But emotions have a purpose: motivation! And because of that purpose, I was able to devise an emotional kernel, like the governing principal of an operating system. Of course, the kernel of an emotional system is trust. In addition to returning an emotion, the kernel also returns itself upon evaluation. Gotta watch out for those feedback looops!
Trust is just one of seven evaluative functions, and provides a core that evaluates itself as well as the inputs, on as many threads as there are inputs. Each function is an operational layer (when organized within a neural network, each will have its own multiple layers of nodes), and each has their own distinctive parameters and purpose. Here's the proposed "function signature" for Trust:
trust ~(POSITIVE GAIN|OUTER_DOMAIN_TO_INNER_DOMAIN_VECTOR|FUTURE EVENT)->self()
The parameters are enforced and evaluated, according to the signature. If the trust function were to evaluate the statement, "You will give me a gift," the positive gain may be negated if the gift turns out to be a live gator, for instance. The trust function would reject the toothy present and the negative implications if it were known!
Much more to come, including the building and training of the network (TBD, but compatibility with iOS's CoreML would be perfect)!