New @BrightLineWatch on state of US democracy brightlinewatch.org/still-miles-ap…
-Experts: Overturning pop. vote key risk; legal changes to enable are grave/serious threats
-Public: Secession support ↑ in partisan strongholds but info on official AZ recount ↑ R trust more than "audit"
🧵
Hardball tactics like gerrymandering, packing the Supreme Court or blocking nominees, voter suppression, abolishing the filibuster, adding new states, or refusing to certify election results attract little support from public or experts (w/few exceptions) brightlinewatch.org/still-miles-ap…
Experts expect hardball tactics to be used more frequently, rating obstruction of Supreme Court nominations as highly likely and refusal to certify popular vote totals as a likely outcome as well. Hardball tactics more favored by Ds are seen as less likely brightlinewatch.org/still-miles-ap…
Experts overwhelmingly rate state bills that encroach on the independence of local election officials and restrict mail voting as grave or serious threats to democracy. Other changes were seen as less consequential, though still potentially harmful. brightlinewatch.org/still-miles-ap…
Information about the official audit in Maricopa County more than doubled GOP confidence in the vote count there. Surprisingly, news about partisan “audit” there also increased R confidence, though effect seems unlikely to persist if "results" claim fraud brightlinewatch.org/still-miles-ap…
A candidate choice experiment shows that Republicans continue to punish hypothetical GOP candidates for supporting Trump's 2nd impeachment or the certification of election results showing Biden's victory. Results are unchanged since Jan./Feb. of this year. brightlinewatch.org/still-miles-ap…
37% of Americans supported regional secession from the union - up from 29% earlier this year. Support was highest in the dominant partisan group in each region and increased most in partisan strongholds - Democrats in the West and Republicans in the South brightlinewatch.org/still-miles-ap…
Experts rate Trump’s continued refusal to accept the 2020 election results as the most abnormal and important event of recent months. However, unlike 2017-2020, most actions experts rated as both abnormal and important were taken by other Republicans. brightlinewatch.org/still-miles-ap…
New @nature: Like-minded sources on Facebook are prevalent but not polarizing
(open access!)
Our key findings:
-Median FB user gets 50.4% of content from like-minded sources
-But reducing exposure by ~1/3 for 3 months had no measurable effect on attitudes https://t.co/IpaHQ0bnGznature.com/articles/s4158…
Increased polarization and hostility are often blamed on echo chambers on social media.
We therefore look at the sources people are exposed to on Facebook to see (a) how slanted they are toward people's political leanings & (b) what effect changing those sources would have.
We estimate user political leanings with a validated classifier and categorize other users, Pages, & groups as "like-minded" (sharing user's leanings), "cross-cutting" (on other side), or neither (in the middle). Exposure data are June-Sept. '20, experimental data Sept.-Dec. '20.
Next I pressed ChatGPT to make synthetic @j_a_tucker, @AdamBerinsky, @emilythorson, @AliceMarwick, & @duncanjwatts more critical & to draw on specific research. Again not nearly as good as the real folks but how long until we're evaluating R2 on value above ChatGPT (VOCG)?
Here's a review written by synthetic me. I gave my paper an R&R, which I guess is a good sign?
Academics: Papers that say they are preregistered must (a) attach the prereg as filed and/or (b) provide a working link to the full prereg. Anything else should be bounced by the journal or rejected by reviewers. Otherwise prereg just becomes a totem people invoke w/no content.
I ran out of room but this too! Reviewers HAVE to look at the prereg. Writing down a few vague hypotheses and then saying the whole study is preregistered is not a preregistered study. Telltale signs: lack of description of deviations & what is exploratory
If you are an editor or on an editorial board, you need to include these checks in your post-submission workflow. Don't let papers get to the review stage that don't provide full preregistration information if they claim it!