Something along the lines of auto research is what I have in mind for this psychology agent. It is currently working on training a model, with handholding right now.