I mean it failed because > Once people learned the messages were co-created by a machine, it didn’t work. Simulated empathy feels weird, empty. The distinctio

I mean it failed because > Once people learned the messages were co-created by a machine, it didn’t work. Simulated empathy feels weird, empty. The distinction between human-authored vs human-edited is only salient so long as it aligns with the "people leaning the messages where co-created by a machine" dimension. A followup experiment would be disentangling the content from the frame. The control would be human-authored messages framed as such, and the three treatments would be human-edited messages framed as such, human-edited messages framed as human-authored, and human-authored messages framed as human-edited. Whether or not the conclusions are actionable is another question entirely. What would we do if human-edited messages framed as human-authored work just as well or even better? Are there ethical lies?