Professor of Sociology, Psychology, & Org Behavior
@Stanford
. Director of Polarization and Social Change Lab, Co-Direct
@StanfordPACS
. Obsessive basketball fan
🚨New WP: Can LLMs predict results of social science experiments?🚨
Prior work uses LLMs to simulate survey responses, but can they predict results of social science experiments?
Across 70 studies, we find striking alignment (r = .85) between simulated and observed effects 🧵👇
🚨New WP: How can we reduce partisan animosity & anti-democratic attitudes in US?🚨
We share results of the Strengthening Democracy Challenge: N=32k megastudy testing 25 depolarization interventions
Shows effective treatments for many neg outcomes!
🧵👇
🚨Come work with us!🚨 The Polarization and Social Change Lab
@Stanford
has an open Research Coordinator position. We invite applicants from psychology, poli sci, sociology, econ, data science, et al. Position begins in Spring or Summer of 2022. Please forward to any and all!
🚨🚨Come work with us!🚨🚨
I'm excited to announce that Stanford's Polarization and Social Change Lab is recruiting two new, post-doctoral positions: a Postdoctoral Scholar and a Research Scholar! See full thread below for info on both. Please share!
🚨Results are in for the Strengthening Democracy Challenge. Winners will be announced this week!🚨
ITT we announce the 25 submissions we selected to test. We think these submissions are awesome & hope you do too.
But first, how we got here…👇🧵
🚨🚨Do Americans accurately perceive union benefits? 🚨🚨
Our new paper (led by
@JonneKamphorst
) out this week
@PNASNews
finds:
Americans significantly and systematically underestimate material benefits associated with union benefits. 🧵👇
🚨🚨New paper “Incivility is Rising Among American Politicians on Twitter” out now in SPPS! 🚨🚨
Is political discourse on Twitter growing more toxic? Is the platform responsible? If so, what can be done about it? 🧵👇
🚨Come work with us!🚨
The Polarization and Social Change Lab at Stanford has an open post-doctoral position. We invite applications from psychology, poli sci, sociology, econ, data science, et al. Review begins 1/21/2021. Please forward to any and all!
🚨New WP🚨 “Elite Party Cues Increase Vaccination Intentions among Republicans”
How to convince Republicans-biggest holdout group-to get vaccinated? We find pro-vax messaging from Trump+Rep elites works, whereas Biden+Dems message reduces enthusiasm
🚨Come work with us!🚨
The Polarization and Social Change Lab, in partnership with Stanford Impact Labs, invites applications to a NEW post-doc program for scholars looking to make an impact. We invite applicants from sociology, psychology, poli sci, econ, data science, etc 1/X
🚨Call for Submissions🚨 “The Strengthening Democracy Challenge,” a large-scale project testing interventions to reduce (a) anti-democratic attitudes, (b) support for partisan violence, and/or (c) partisan animosity, is open for submissions NOW 1/
How can progressive candidates build support beyond their base? By explaining how their policies fit well with widely held values like family, security, and the American Dream. My new piece with
@jgvoelkel
in the
@nytimes
🚨Out now in
@PNASNews
🚨
Are women politicians disadvantaged by perceptions of their electability? How can these perceptions be overcome?
We report results of 6 studies (N=7,895) conducted during the 2020 Dem primaries. 🧵
Before arguing with friends or family around the Thanksgiving table, take a look at the science behind arguing better. And it'll never hurt to try this: "Listen to people, get them to think about their own experience, and highlight your common humanity."
Stanford’s new Polarization and Social Change Lab has an open postdoctoral position. We invite applications from political science, psychology, sociology, & beyond. Review starts Dec. 15. Please forward to any and all who might be interested!
#GPT3
and
#ChatGPT
can produce appeals that persuade humans on political issues (even highly polarized ones), with effect sizes similar to appeals written by humans. New research from our lab, with
@baixx062
,
@jgvoelkel
, and
@JEichstaedt
. 🧵👇
New WP: 🚨Artificial Intelligence Can Persuade Humans on Political Issues🚨Large language models (like
#ChatGPT
and
#GPT3
) are changing the world. Here we show that GPT3 can write messages that change people’s attitudes on important policy issues 🧵1/11
🚨New WP🚨 “Correcting Inaccurate Metaperceptions Reduces Americans’ Support for Partisan Violence”
We find correcting inaccurate perceptions of outpartisans’ support for partisan violence significantly and durably reduced support for partisan violence
Academic Twitter, what are ways that we could apply social & behav science to help address the Coronavirus pandemic? E.g., ideas for effective public health messages, strategies for nudging leaders to quarantine, social distancing, things social media platforms can do, etc.
I had the great pleasure of chatting with
@ShankarVedantam
on
@HiddenBrain
, discussing political persuasion, empathy, moral reframing, and my unforgivable behavior towards roommates throughout my early 20’s. Of poss interest 👇
“If the goal is persuading somebody, you may need to…focus on getting into that person's head in order to construct an argument that would be persuasive to them” -
@RobbWiller
This week, we continue our US 2.0 series with a look at how to have more productive disagreements.
Republicans are the largest vaccine hesitant group in the US and their hesitancy is not declining. What can we do about it?
In this new piece in the
@LAtimes
, I &
@jayvanbavel
discuss strategies and some of the relevant behavioral science in this space
It was a pleasure working on this project - out today in
@Nature
- with
@kairuggeri
,
@jayvanbavel
, and many other terrific scholars. And a big h/t to Kai for his impressive leadership.
Our new paper in
@Nature
evaluates the accuracy of social and behavioral policy recommendations during COVID19 (from )
We reviewed 747 papers (mean N = 16,848) and found evidence for 89% of our central claims:
The strongest
Honored that
@pascl_stanford
was featured on PBS
@NewsHour
today! Heartfelt thanks to
@JudyWoodruff
for visiting to talk with our passionate team. Watch the segment here (starts @ 38:50):
Many 2022 candidates do not acknowledge results of 2020 election. This is a huge problem.
How can we persuade people not to support undemocratic candidates in the 2022 midterms?
@jgvoelkel
& I offer insights from the Strengthening Democracy Challenge.
🚨🚨Do you have social science research to share on COVID-19 vaccination?🚨🚨
If yes, pls add your research to the COVID-19 Behavioral Research Repository! Details below...
1/3
Very excited to see this in-depth coverage of the Strengthening Democracy Challenge in
@TheAtlantic
! When you have a chance, please check out this piece written by the outstanding
@GrahamDavidA
Our new paper on the effects of racial status threats on white Americans' opposition to welfare programs is out this week in
@SF_Journal
. I'm very proud of Rachel Wetts - sociology PhD candidate
@UCBerkeley
- for her excellent work on this paper!
Come work with us! The Polarization and Social Change Lab
@Stanford
has up to 2 open Research Coordinator positions. If you’re interested, please apply now! We will begin reviewing applications this week.
🚨Come work with us!🚨 The Polarization and Social Change Lab
@Stanford
has an open Research Coordinator position. We invite applicants from psychology, poli sci, sociology, econ, data science, et al. Position begins in Spring or Summer of 2022. Please forward to any and all!
🚨🚨Call for Submissions🚨🚨
For the “Structural Democratic Reforms Project” we are gathering ideas from a diverse range of academics, practitioners, and concerned citizens.
Submit your ideas for reforms to improve our political system NOW!
1/
When
@kairuggeri
proposed to
@jayvanbavel
and I to conduct a rigorous evaluation of claims made in the 2020
@NatureHumBehav
article we co-organized, I was excited but also a bit wary. How could we be involved in a fair evaluation of our own claims?
Did behavioral science offer anything during the pandemic? 76 authors (23 countries) assessed evidence from 3,000+ behavioral science papers during COVID, testing 19 claims by Van Bavel et al 2020.
Preprint:
Interactive database:
To evaluate predictive accuracy of LLMs for social science experiments, we used
#GPT4
to predict 476 effects from 70 well-powered experiments, including:
➡️50 survey experiments conducted through NSF-funded TESS program
➡️20 additional replication studies (Coppock et al. 2018)
🚨New WP🚨“Religious Identity Cues Increase Vaccination Intentions and Trust in Medical Experts among American Christians”
How to promote vaccination among American Christians? Highlighting Christian identity of experts increases trust & vacc intentions
@jimtankersley
Also unfortunate is a culture of contemputousness that fosters intellectual incuriosity and a posture of "how can I cut this down" rather than "what can I learn here" that I think can be unscientific.
This is an excellent, excellent thread. Fraud (& the use of QRPs to an extent that borders on fraud) directly and indirectly harms scientists who do it the right way. Things have improved a lot, but we need more progress, reforming institutions to align incentives w/ good science
Ugh: A 🧵. This pisses me off. I have spent a TON of time getting null results published - including one very much related to the fraudulent paper in question - here is my field exp null on honesty pledges: …
FYI, we are recruiting for at least one new Research Coordinator position in our lab (details 👇).
We encourage a culture of rigor, discovery, fun, and solidarity. We aspire to do work that has meaningful social impact. If this sounds appealing to you, please apply!
🚨Come work with us! 🚨The Polarization and Social Change Lab
@Stanford
has an open Research Coordinator position. We invite applicants from psychology, poli sci, sociology, econ, data science, et al. Position begins in Spring or Summer of 2024. Please forward to any and all!
But what if LLMs are simply retrieving & reproducing known experimental results from training data?
We find evidence against this: analyzing only studies *unpublished* at time of GPT4’s training data cut-off, we find high predictive accuracy (r = .90, adj. r = .94).
🚨🚨Come work with us!🚨🚨
Announcing a 3rd new postdoc opening in our lab, this one focused on large language models mentored by
@JEichstaedt
& me.
Please share widely!
@bubbaprog
And now the politicization of non-COVID vaccination begins. We can expect the association between partisanship and all forms of vaccination to rise.
Just so terribly sad. It didn't have to be this way.
Check out this new paper, which leverages a unique sample of state legislators to measure outparty perceptions among US political elites.
(Kudos to
@suji_kang_
and Jamie Druckman for leading this project!)
🧵👇
(1/n) Now out
@PNASNews
, we show that correcting legislators' misperceptions of out-partisan voters reduces legislators’ own support for undemocratic practices (SUP) and animosity toward out-partisans.
Check below what we found! ⬇️
🚨🚨Big changes are happening in our lab (
@pascl_stanford
) as we celebrate exciting transitions for several lab members!!🚨🚨
Here are some of the wonderful people who will be headed to new schools (and a sneak peek into the bright minds joining us this Fall)! 🧵👇
#LabUpdates
Now hiring a full-time research coordinator in the Stanford Polarization & Social Change Lab. Applications welcome from social psych, poli sci, sociology, et al. Projects on persuasion, civility, social movmnts, discrimination, inequality, & more. Join us!
Can confidently report that when an aging hipster dad steps on a child's Lego in the night, the pain is 40% physical & 60% the knowledge that one has become a cultural cliche.
Religious identity cues can increase vaccine intentions and trust in medical experts among American Christians.
This research - headed up by
@jchu1225
and Sophia Pink - out now in
@PNASNews
.
🚨New WP🚨“Religious Identity Cues Increase Vaccination Intentions and Trust in Medical Experts among American Christians”
How to promote vaccination among American Christians? Highlighting Christian identity of experts increases trust & vacc intentions
The deadline to apply for a post-doc in our lab is in one week. If you are interested in political dimensions of poverty, social movements, gender, race, and/or environmental protection, please consider applying!
🚨Come work with us!🚨
The Polarization and Social Change Lab at Stanford has an open post-doctoral position. We invite applications from psychology, poli sci, sociology, econ, data science, et al. Review begins 1/21/2021. Please forward to any and all!
Our results reveal that the efforts of those interested in reducing partisan animosity have yielded an impressive base of knowledge. 23 interventions reduced partisan animosity, often sizably (up to 10pts on a 100-pt scale), across survey and behavioral indicators, and durably.
In line with claims that American democracy is in crisis, we found concerning baseline levels of potentially problematic attitudes, e.g.:
➤Partisan animosity
➤Support for undemocratic practices
➤Support for undemocratic candidates
➤Biased evaluation of politicized facts
A big congratulations to
@AliaCrum
being promoted to Associate Professor! This is so richly deserved. You couldn't ask for a better colleague.
It's wonderful to see good things happen to great people. Go Ali!
For example, we found evidence that six interventions reduce support for undemocratic practices. Reducing misperceptions about outpartisans’ anti-democratic attitudes and highlighting the disastrous consequences of democratic collapse were the most promising approaches.
We find:
➤Partisan animosity is largely driven by other factors than support for undemocratic practices or partisan violence.
➤Partisan animosity is driven by similar factors as biased evaluation of politicized facts and general social distrust.
New joint-lab project analyzing levels of representativeness and participant attentiveness across a range of survey platforms. (Hats off to Nick Stagnaro for leading this!)
We hope there is lots of valuable information here for those using online participants!
🚨WP🚨
We test 9 online samples and find clear tradeoffs between attentiveness and representativeness - which sample is best depends on research q and priorities. For social/political qs I rec Bovitz/Lucid, for complex designs I rec Cloud/Prolific.
I had a great and wide-ranging conversation about effective paths to social change, persuasion, polarization, democracy and lots of other stuff on The Great Battlefield podcast.
Political violence is back in the spotlight following the appalling attack on Paul Pelosi.
In a
@monkeycageblog
piece in the
@washingtonpost
,
@jsmernyk
, Sophia Pink, & I describe recent research on ways to increase Americans’ opp to partisan violence:
@profcikara
Send an email to yourself explaining your trouble writing, explaining in detail what you would write if you were not having this trouble. (Alternatively, draft some poor relative to be the email recipient/unwitting muse.)
Below are short descriptions of the 25 interventions we selected (with short names we made). We hope you're as interested in them as we were!
More detail on our website, including test links to try them out yourself.
(& results coming soon!)
We also found five interventions that significantly reduced support for partisan violence. Engaging videos showing partisans’ reactions to misperception corrections and a joint pro-democratic statement by Utah’s candidates for governor in 2020 were most effective.
My research with
@jgvoelkel
in today's
@nytimes
. Our work is presented along with others' (e.g.,
@IanHaneyLopez
@anatosaurus
) in a discussion of how progressives can effectively overcome their opponents' racial divisiveness. A characteristically careful overview from
@Edsall
.
We test 25 interventions to reduce such attitudes, submitted by social scientists & practitioners. Most targeted partisan animosity, but many also aimed to reduce support for undemocratic practices or partisan violence. Walk-through of interventions here👇
🚨Results are in for the Strengthening Democracy Challenge. Winners will be announced this week!🚨
ITT we announce the 25 submissions we selected to test. We think these submissions are awesome & hope you do too.
But first, how we got here…👇🧵
Overall, our results provide a toolkit of promising interventions for practitioners & shed new theoretical light on challenges facing American democracy. Problematic attitudes are related in complex ways, and effectively treating them requires understanding those relationships!
I had such a great time talking with
@AndyLuttrell5
on his
@OpinionSciPod
. It was fun to talk about such a range of topics: persuasion, democracy, psychology, group perceptions, social cog...
(And kudos to Andy who always runs this show like a total pro!)
We're back!
@RobbWiller
shares his work on political persuasion + a giant new study testing strategies for reducing partisan animosity and enhancing commitments to democracy.
Apple:
Spotify:
Web:
I joined more than 200 scholars of democracy in calling for Congress to make elections more competitive by adopting proportional representation for the U.S. House.
We also find that predictive accuracy improved across generations of LLMs, with GPT4 surpassing predictions elicited from an online sample (N = 2,659) of Americans.
Paper 👉
Had the good fortune to discuss the Strengthening Democracy Challenge on
@cspan
's Washington Journal....& somehow did not appreciate it's primarily a call-in show until I was on (!).
It was a great experience, thanks to host
@cspanMcArdle
's stewardship.
Postdoctoral Scholar with interest in persuasion, inequality, polarization, messaging, social movement mobilization, climate change attitudes, public opinion regarding major social issues, &/or online political behavior. Application Deadline: Jan 9.
This project was a true team effort. Major thanks to the leaders of the project:
@jgvoelkel
, Nick Stagnaro, and
@jchu1225
. And huge kudos to Sophia Pink, Joe Mernyk, Chrystal Redekopp, Jamie Druckman, &
@DG_Rand
for their brilliance and hard work.
Overall our results show high accuracy of LLM-derived predictions for experiments with human participants, generally greater accuracy than samples of lay and expert humans.
Paper👉
We prompted the model with (a) demographic profiles drawn from a representative dataset of Americans, and (b) experimental stimuli. The effects estimated by pooling these responses were strongly correlated with the actual experimental effects (r = .85; adj. r = 0.91)!
Returning to Congressional toxicity, here are 2 examples:
A civil tweet doesn’t get a lot of “action,” but is liked on avg
A more toxic tweet is seen by way more people making it seem more liked, when actually it’s liked less by the avg person seeing it (on- and off-platform)
We also find promising interventions that yield significant improvements for each of the other potentially problematic outcomes we examine - although effects on these outcomes were generally less common, smaller, and less durable than for partisan animosity.
⚡️Common Identity
@alijaveed_
@kim_doell
@steverathje2
@jayvanbavel
Participants read about how American democracy is crucial to country’s success and at risk from extreme partisanship & then learn most Americans support democracy, dislike violence, & like members of both parties
⭐️Our lab’s preprint “Short Messages Encouraging Compliance with COVID-19 Public Health Guidelines Have Minimal Effects” now online⭐️
How can we persuade people to comply with COVID guidelines? Results of 5 experiments (not yet peer-reviewed!)
1/12
🚨Join us for a workshop on how to run megastudies at
#SPSP2024
🚨
@RobbWiller
and I will share our experiences and recommendations for conducting crowdsourced megastudies. Please sign up 👇:
I had a lot of fun talking with
@EricNeumannPsy
about our lab's research in this episode of the
@StanfordPsyPod
podcast. Our conversation went to some interesting places I didn't expect in the 2nd half. Hats off to Eric who is a phenomenal interviewer!
One-stop shopping for resources related to the Strengthening Democracy Challenge (press release, paper, details of interventions, etc.) on our website:
As we bid adieu to these terrific outgoing scholars, we're thrilled to welcome a new folks this Fall:
@AliaBraley
,
@blakey_will
,
@Izzy_Gainsburg
,
@JonasP_Schoene
, as well as new undergrad RAs: Sebastian Kane, Ben Kinder, Lavanya Sundar & Tiffany Saade.
Some professional news: After five amazing years, I'm leaving the world's prettiest university
@EUI_EU
. I'm incredibly excited to share that next academic year, I will be starting a Postdoc at
@Stanford
@pascl_stanford
and
@StanfordHAI
. 🌲☀️🏄
Finally, we find LLMs can accurately predict effects on socially harmful outcomes, such as the impact of antivax FB posts on vax intentions (
@_JenAllen
et al., 2024). This capacity may have positive uses, such as for content moderation, though also highlights risks of misuse.
This is a must-read for any/all folks working on messaging, persuasion, etc. for the campaign right now. Helps to answer some of the highest level strategic questions. Results are intuitive from a Bayesian perspective, but that's def not the one that predominates in this space.
When we designed the challenge, we realized how little we actually know about how connected the many important political attitudes are.
Using effect sizes from the many interventions on the many outcomes we collected, we can shed unique light on this question.
Our new paper on the effects of morality on perceptions of competence out now in JPSP. Very proud of Jennifer Stellar of
@UofT
for her outstanding work on this paper (esp the 2 preregistered experiments).
@AshuAshok
, an awesome scholar of identity, politics, & computational social science is joining the faculty of
@NYUPsych
. Her sharp mind and creativity will leave a lasting mark on all of us here!
Spotlight: Support for undemocratic candidates
Combination of how much voters endorse undemocratic practices, and how much they despise rival partisans - most effective interventions for reducing SUC are the top 2 for interventions for reducing SUP and top 4 interventions for PA
As the
#Deltavariant
drives surging caseloads across the US, a growing number of Republican politicians are now publicly supporting vaccination. Will it matter? Our research suggests that it could.
New op-ed with
@DG_Rand
in
@washingtonpost