you know most human emotion is constructed, inferred, there is no root object, you can kind of create the emotion you want? well, i was looking at human emotion experiments and thinking of adapting them to llms. i was thinking of this one because we've already found narrative priming to be super-effective on llms: https://pmc.ncbi.nlm.nih.gov/articles/PMC2758776/
[link] [comments]