Bernhard Clemm
@BernhardClemm
Followers
643
Following
823
Statuses
691
computational social scientist @gesis_org | data editor @polcommjournal
Amsterdam
Joined October 2014
Thanks to @MJHeseltine for the vision & leadership on this project! LLMs obviously come with many issues – but there is increasing evidence that they can facilitate social-science research. A (non-exhaustive) 🧵 with examples I have come across:
This paper tests the performance of GPT4 in text coding across a range of common political science coding tasks in 4 languages. @MJHeseltine & @BernhardClemm find that GPT4 coding is highly accurate, especially for shorter texts. Read more here:
2
4
16
Great honour to have contributed to this impressively large collaboration!
** Source Credibility @BernhardClemm
0
0
1
Very excited about this working paper, which @_Tiagoventura is going to present at @APSAtweets!
This work is coauthored work with @BernhardClemm @Jonathan_Nagler @erickaakcire and @mwojcieszak and another of the projects I started as a postdoc at @CSMaP_NYU. Link for the pre-print here:
0
0
1
Highly recommend applying for this position (which I was lucky to hold since 2023). If you want to shape open-science practices in the social sciences & work with a dedicated team of great scholars, don't miss it!
Join the PolComm team! We are accepting applications (CV, cover letter, 3 references) for our Data Editor position until September 15th. Read more here:
0
8
6
RT @polcommjournal: Join the PolComm team! We are accepting applications (CV, cover letter, 3 references) for our Data Editor position unti…
0
23
0
True, @europsa can feel lonely, but I wouldn't go so far as comparing it to a penal institution!
0
0
5
RT @HeikeKluever: Can citizens' assemblies strengthen democratic resilience? In a new @scriptsberlin paper that I present at #MINORITYRULE…
0
15
0
RT @th_ges: Finishing up our section with Survey Research & Survey Experiments, chaired & discussed by @klrmllr and @BernhardClemm
#EPSA20…
0
3
0
RT @TimWapps: It‘s @europsa-time 🥳 and I’m really excited to present our (@BernhardClemm, @FeIiX_H, @HeikeKluever) first results from a la…
0
8
0
RT @ba_lyons: Cool paper: "Most previous studies use a sample of information that is balanced on ideological valence. My study, in contrast…
0
2
0
@EikoFried Next time, try convincing the copy editors to include images in high resolution. They seem to be fans of big pixels!
1
0
1
Not wrong
This is what the credibility revolution did to social science. A generation stuck in the basement thinking exclusively about absurd situations so they can win a fake assumptions contest and a generation of gullible readers who want an easy button for believing a result.
1
0
0
SQL is the easiest coding "language" out there as far as I see it. Used across academia, industry and government. If that's boring / too technical, perhaps one shouldn't report on "Technology"?
0
0
0
These exchanges are what makes Twitter still worthwhile. How is Mastodon doing btw?
To explain away smoking & lung cancer link association, you need a confounder that: - is associated with a 20x increase in lung cancer - is 20x as prevalent among smokers What's the equivalent for social media & mental health?
0
0
3