For more details, contact me or Jenn Wortman Vaughan (jenn@microsoft.com) from Microsoft Research. The study also involves our collaborators Mina Lee (UChicago) and @ndiakopoulos.bsky.social#ScienceJournalism#GenerativeAI#ScienceWriting#Journalism#HCI

Your insights will help shape the future of AI and journalism, and you'll receive a $75 gift card as a token of our appreciation. Interested? Please fill this form: forms.office.com/r/dNG5zMuvWb (2/3)

Here's the weird Taylorism again. The system produces work at the level of an early trainee requiring substantive supervision. This is not good ROI for producing papers. The primary output of time invested in trainee research is the development of independent scientists—not the research papers.

same thought, I also think the structure of only connect makes it both harder to find connections and easier to reject spurious connections. sometimes the nyt's decisions on what items so not connect is very "no that's not what we want, we want a different one"

definitely due to how many times it played on tv when i was growing up

I read that and I agree -- it also may force users to actually think about the nature of the information they care about. but I also wonder if there are cases when iterative interaction actually aids understanding? (I'd also just personally spitball w/ a person at that point tho so idk)

do you think more scaffolded chatting with a doc ("ask it anything" -> "here are some questions you can ask this doc you are reading") can be more effective? like do you feel that chatbots are inherently the "everything feature", or is it about the specifics how they are made and configured here?

some of those tutoring places ARE a bit of a cult, so that would track honestly

this is actually really funny, in India we used to have extra “doubt sessions” if we went to tutoring, where we could just show up and ask questions, like this is actually what we called them

I wasnt sure what this would be like but your review has me v excited!!