We find that the presence of #explanations increases reliance on both correct and incorrect LLM responses.
This isn't surprising, given prior HCI/Psychology work on other types of explanations, but raises the question whether LLMs should provide explanations by default 
4/7
First, through a think-aloud study (N=16) in which participants use ChatGPT to answer objective questions, we identify 3 features of LLM responses that shape users' reliance: #explanations (supporting details for answers), #inconsistencies in explanations, and #sources.
2/7
Here's a link to a 15 min talk I gave at a Bielefeld conf on #AI and #explanations.It's titled "A Defense of Inexplicability" mainly on something like metaphysical grounds. https://www.youtube.com/watch?v=Pp6z9UnAZBc
Like these fuckers on #socialmedia that claim they are #anarchist but can't #explain #why they are, hilarious, then they mention #nihilist #ideology in their #explanations thats far from fucking anarchist #beliefs. 
#cognition #understanding #epistemology @sociology @psychology
Leonid #Rozenblit and Frank #Keil: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3062901/
Abstract:
"People feel they understand complex phenomena with far greater precision, coherence, and depth than they really do; they are subject to an illusion—an illusion of explanatory depth. The illusion is far stronger for explanatory knowledge than many other kinds of #knowledge, such as that for facts, procedures or narratives. The illusion for explanatory knowledge is most robust where the environment supports real-time #explanations with visible mechanisms. We demonstrate the illusion of depth with explanatory knowledge in Studies 1–6. Then we show differences in overconfidence about knowledge across different knowledge domains in Studies 7–10. Finally, we explore the mechanisms behind the initial #confidence and behind #overconfidence in Studies 11 and 12. Implications for the roles of intuitive theories in models of #concepts and cognition are discussed."
We gave a gyroscope to our 7yo grandchild for Hanukkah. Great, except it's the fastest route to "I don't know! It just is!" when they ask how it works. Likewise for magnets. Prisms let you get one step further. Yay? #science #explanations #STEM
I don't understand why one would prefer a simpler theory rather than a less simple one independent 1) aesthetic preference and/or 2) computational limitations. I am convinced that our finite powers are something we should take seriously in #theory choice (esp. since hearing a good talk by @bweatherson about this), but a friend really wants me to expect that simpler #explanations are more likely to be true than less simple ones. I can't see why that would be.
Perhaps a word of explanation is in order : the posts I mark with #LuminousDiary are "very short stories" that I'm writing esp. to post on Mastodon as a way to contribute the literary culture here.
Most of these very short stories are fictional, but "Getting out of the house" is based on a trip I took to the Imperial Valley with "P", a trip that is the subject of a novel I wrote nearly fifteen years ago. #Explanations
Next was an interesting talk by Michael Pazzani at the #USC Information Sciences Institute on creating #algorithmic system #explanations with expert input. The ideas here are broadly applicable beyond the #image #classification tasks discussed, and I love the approach to working with experts at all stages of the system design process. Highly recommend https://www.youtube.com/watch?v=iiBxFjL82lU (5/7)
