Discussion
TFW you get asked to join a science communication event (yay!) where people can ask any question (AMA, let's go!) but the format is that an AI answers first and then the 'expert' can discuss it. Why? 😭
I refused and said it was irresponsible: a crucial layer of the truth sandwich is missing here. I also mentioned that anchoring effect, automation bias, and bullshit asymmetry all work against the researcher in this format. 😩
#SciCom
Why would anyone say "yes" to that.
@SylviaFysica How could anyone possibly think that was a good idea?!
@lcwheeler I do have a theory about how they brainstormed it...
Also, the theme was health. I would recommend people to talk about their health with a medical professional. Not a chatbot, nor a random philosopher they met at a market! 😅
@SylviaFysica I might be too Autist for this, but isn't this an excellent format to dissect all the bullshit the AI is spewing?
@dragonfi I guess that's the intention, but for that goal it would be better to have expert answer first, then look at the generated answer, and point out problems with it (~truth sandwich).
Even then:
- person might believe AI answer more (sounds authoritative, supposedly based on sum of human knowledge, etc.)
- it is very difficult to correct deeply rooted/wide-spread beliefs, that the AI answer might incorporate
- even specialist might need more time to even see all that's wrong with output
In addition, it is like most proposals to get an #LLM to answer questions, just a huge irresponsible waste of everyone's time.
@SylviaFysica that's the dumbest comms formatting decision I've heard of, in 20 years of doing science comms.
@SylviaFysica Wow, that's really gross, yeah :-(
Install bonfire.cafe
Get the full app experience
