Chair for Data-Science in the Economic and Social Sciences at @MannheimBSchool (Germany), affiliated with @gesis_org and @cshvienna
Joined March 2008
- Tweets2.8K
- Following653
- Followers2.7K
- Likes6K
Markus Strohmaier retweeted
My reading of the recent literature on AI writing:
People generally cannot reliably tell AI from human writing. Under blind conditions, they often actually prefer the AI writing. But tell them something was written by AI, and suddenly they prefer the โhumanโ version.
1/
Markus Strohmaier retweeted
๐จNew paper alert๐จ
๐ค Ever wondered how the way you write a persona prompt affects how well an LLM simulates people?
In our #EMNLP2025 paper, we find that using interview-style persona prompts makes LLM social simulations less biased and more aligned with human opinions.
๐งต1/7
LLMs caused drastic vocabulary shift in biomedical publications science.org/doi/10.1126/sciaโฆ
Markus Strohmaier retweeted
This @icwsm paper fine-tunes Temporal Adapters for Llama 3 8B on full timelines from a panel of British Twitter users, finding that the fine-tuned LLMs retrieve collective emotions! Cool extension of simulation work to the longitudinal scenario:
ojs.aaai.org/index.php/ICWSMโฆ
Markus Strohmaier retweeted
How much code now comes from AI? In new work with @simone_daniotti, @xiangnan_feng & @FrankNeffke we estimate that by end-2024 about 30% of Python functions pushed by US devs on GitHub are AI- generated. Adoption is rapid but diffusion lags globally. How did we do it?
Markus Strohmaier retweeted
Agreed. The behavioral and social sciences should rely more on open LLMs: doi.org/10.31219/osf.io/ybvzโฆ
Markus Strohmaier retweeted
Proud to present the first benchmark for representational similarity measures today at @iclr_conf!
If you also attend #ICLR, come see us this afternoon, Hall 3 #456!
Otherwise, you can check out our paper here:
openreview.net/forum?id=PRvdโฆ
Markus Strohmaier retweeted
Update: our survey on functional and representational similarity measures for neural networks has just been accepted to ACM CSUR!
We present & categorize 53 measures, discuss their properties & application, but also identify key research gaps:
dl.acm.org/doi/10.1145/37284โฆ
Markus Strohmaier retweeted
Excited to share the materials for our course on open LLMs for science of science research @ZakASHussain
and I offered at the recent meeting of @euroscisci
at @LMU_Muenchen.
github.com/zak-Hussain/LLM4Sโฆ
Markus Strohmaier retweeted
Want to learn about computational social science *for free* and launch interdisciplinary research projects? We are so excited to announce there will be *26* Summer Institutes in Computational Social Science this year! Apply to one of them here: sicss.io/locations
Markus Strohmaier retweeted
Job announcement: Junior Professorship in Computational Social Science at the University of Konstanz.
Join us at the Center for Data and Methods! Feel free to reach out if you have questions.
stellen.uni-konstanz.de/jobpโฆ
Markus Strohmaier retweeted
Two exciting positions as postdoc / senior researcher with tenure option at KTS @gesis_org (gesis.org/en/kts) on computational social science and #NLP. Details at: hidden-professionals.de/HPv3โฆ and
hidden-professionals.de/HPv3โฆ.
Markus Strohmaier retweeted
"The increasing reliance on AI tools for social and emotional tasks heightens the risk of #autonomy #loss. As synthetic relationships become more prevalent, companies that operate these AI systems may gain substantial influence over usersโ lives. "
from Nature Human Behaviour
๐จ Thrilled to share our latest paper on the risks & protective measures for #SyntheticRelationships, published in @NatureHumBehav๐จ
We explore how SRs between humans and AI can pose serious risks to individuals & society. Key concerns include...
nature.com/articles/s41562-0โฆ
๐งต1/5
Markus Strohmaier retweeted
Weโre hiring a Team Lead @gesis_org Computational Social Science department! ๐
Passionate about open science & innovative methods for digital behavioral data?
๐ Link: hidden-professionals.de/HPv3โฆ
โ Questions? Contact me via DM or mail
Markus Strohmaier retweeted
Do you need to copy art to make art?
That question is the core of the tension between generative AI and copyright.
Check out @rhfeiyang and @materzynska's new work that avoids slurping up art.
Their Art-Free Diffusion & Art-Free data let you make "imitation-free" AI art๐งต
Markus Strohmaier retweeted
We just opened a new team leader position at the Computational Social Science department at @gesis_org please spread the news and contact me or @SebStier for questions hidden-professionals.de/HPv3โฆ
Markus Strohmaier retweeted
Where does gender bias manifest in model parameters? ๐ค Excited to present our work on Local Contrastive Editing of Gender Stereotypes at @emnlpmeeting on Thursday at 10:30am! Drop by our poster to chat about new insights into bias control. #EMNLP2024
Markus Strohmaier retweeted
New preprint: ๐ง๐ต๐ฒ ๐๐๐ป๐ฎ๐บ๐ถ๐ฐ๐ ๐ผ๐ณ ๐๐ผ๐ฐ๐ถ๐ฎ๐น ๐ฐ๐ผ๐ป๐๐ฒ๐ป๐๐ถ๐ผ๐ป๐ ๐ถ๐ป ๐๐๐ ๐ฝ๐ผ๐ฝ๐๐น๐ฎ๐๐ถ๐ผ๐ป๐: ๐ฆ๐ฝ๐ผ๐ป๐๐ฎ๐ป๐ฒ๐ผ๐๐ ๐ฒ๐บ๐ฒ๐ฟ๐ด๐ฒ๐ป๐ฐ๐ฒ, ๐ฐ๐ผ๐น๐น๐ฒ๐ฐ๐๐ถ๐๐ฒ ๐ฏ๐ถ๐ฎ๐๐ฒ๐ ๐ฎ๐ป๐ฑ ๐๐ถ๐ฝ๐ฝ๐ถ๐ป๐ด ๐ฝ๐ผ๐ถ๐ป๐๐
arxiv.org/pdf/2410.08948v1 1/N
Markus Strohmaier retweeted
In the beginning of 2023 I expressed my fear that AI would deforest the knowledgeโs ecosystem spinellis.gr/blog/20230316/. Now a PNAS Nexus paper shows that large language models reduce public knowledge sharing on online Q&A platforms (StackOverflow) doi.org/10.1093/pnasnexus/pgโฆ