L.S.ZwarensteinJan 16, 2023
Eli Pariser
Fascinating and disturbing finding on how AI co-writing can influence opinion from Maurice Jakesh thesis research (h/t @natematias ). Jakesh had people write short essays using an AI tool like Google Autocomplete on whether social media was good or bad. When the tool pre-generated language supportive of social media (top bar), people were more likely to write essays with that opinion and vice versa (bottom bar).
Jan 16, 2023 at 3:35pmWeb
Show threadEli PariserJan 16, 2023
Full thesis here: https://mauricejakesch.com/assets/pdf/thesis_jakesch_cornell_phd.pdf
Show threadHobson LaneJan 16, 2023
@elipariser this is the thing that people don't want to admit to themselves, we're letting algorithms brainwash us. And we have zero control over the values and biases of those algorithms.
Show threadMudManJan 16, 2023

@elipariser There's a lot of interesting stuff in the thesis, but also a lot of shaky bits.

That particular experiment design seems very flawed to me. It gave people a tool that is simply not widespread and told them to write about a thing they weren't necessarily knowledgable or passionate about, which the thesis acknowledges and also admits only partially mitigates or controls for. And there is no control group whatsoever in terms of other persuasion tools, which would have been interesting.

Show threadMudManJan 16, 2023
@elipariser A lot more research is obviously needed, but I see a lot of weird echoes of studies linking other forms of neomedia with behaviors. It's just hard to estimate those kinds of influence and at best you're only ever taking a snapshot.
Show threadBert LatamoreJan 17, 2023
@elipariser @natematias Last Week's "On the Media" podcast focused on AI. Apparently some new AI tools are close to being able to research and write reports on demand. I can imagine it writing white papers to order, which would replace me in what I was doing before retirement.
Show threadMr. TeatimeJan 20, 2023
@elipariser
I exoeriences had something similar in a language test, where I was supposed to comment on a set of presentation slides about sugar.
The slides focussed on natural sugars in fruit etc. and how that was absolutely no problem, so trying to avoid all sugar was silly.
I agreed.
...and was then told the slides came from the sugar industry, and remembered that I actually avoid refined sugar myself, but had excluded that aspect from the discussion.
@natematias @vicgrinberg