Sign in

Thomas Klebel

@tklebel.bsky.social
93 followers 179 following 29 posts

Senior Scientist @ IDea_Lab / University of Graz. I'm interested in what AI does to science, how we can be more reproducible when conducting studies, and how we can make studies more robust and informative by using causal inference techniques.

PostsRepliesMedia
Reposted by Thomas Klebel
jon ben-menachem @jbenmenachem.com · 29/09/2026
"Replication crisis" reforms like preregistration & pre-analysis plans were developed to control false positives in experimental research. Does that work in observational studies? Our new paper in Sociological Methods & Research argues for a different approach. 🧵 journals.sagepub.com/doi/10.1177/...
A Replication Framework for Observational Research: From Specification Restriction to Specification Expansion
Jonathan Ben-Menachem https://orcid.org/0000-0002-8092-0848 jbenmenachem@gmail.com, Ari Galper https://orcid.org/0000-0001-7909-8837, and Nic Fishman
Abstract
Sociology has remained relatively insulated from debates about the “replication crisis.” Heeding calls to consider replication more deeply, we introduce a distinction between two types of research reforms that have emerged in the wake of the crisis: specification-restricting and specification-expanding reforms. Specification restriction is often promoted as a way to increase the repeatability of research findings by controlling false positives. We argue that this statistical justification is less compelling in observational contexts—where the assumptions licensing false-positive control are harder to sustain and rigid ex ante analytical commitments can instead enshrine fragile specifications—than in experimental contexts. We further argue that apparently low replication rates do not necessarily imply a high prevalence of false positive findings. Instead, we propose a replication framework centered on specification-expanding reforms, stronger incentives for confirmatory research, and meta-analysis. This approach equips sociologists to assess the repeatability of findings and build a more cumulative discipline.
28129
Reposted by Thomas Klebel
Nature Portfolio @natureportfolio.nature.com · 06/09/2026
Nearly half of all polarization research focuses on the US, yet many American findings are global outliers. A Review in Nature Communications argues that a globally equitable understanding requires bridging traditional social science with new approaches leveraging social media and AI. 🧪
go.nature.com
Toward an integrated and globally equitable understanding of political polarization - Nature Communications
Nearly half of all polarization research focuses on the United States, yet many American findings are global outliers. A globally equitable understanding requires bridging traditional social science with new approaches leveraging social media and AI.
0239
Reposted by Thomas Klebel
Paul Hünermund @p-hunermund.com · 27/08/2026
📢 The Causal Data Science Meeting 2026 is back! Join us virtually on Nov 4–5 for two days of discussion on #causality in ML & AI, with Teppo Felin as our first keynote speaker. No fees, no proceedings—just research, feedback & exchange. 📅 Submit by Sept 30 ✉️ submission@causalscience.org
0104
Reposted by Thomas Klebel
Jacob T. Levy @jacobtlevy.bsky.social · 24/08/2026
On this day in 2018. Kieran @kjhealy.co ‘s obituary will include the MacArthur essay and “Fuck Nuance,” but for a particular niche audience of connoisseurs, this is his masterpiece.
1035985
Reposted by Thomas Klebel
Olivia Guest · Ολίβια Γκεστ @olivia.science · 15/08/2026
Watch this short clip to the end to understand what level (maximum) of white supremacy we're dealing with here. 1/n
24772251
Thomas Klebel @tklebel.bsky.social · 13/08/2026
Job alert: there is an open position for a full professor in data analysis (focus on natural language processing) at Uni Graz. I've recently started in the same "department" (the IDea_Lab), and it's a great place to work. You can find all relevant info below - reach out if you have any questions!
010
Reposted by Thomas Klebel
Jana Lasser @janalasser.eurosky.social · 24/07/2026
The University of Graz invites applications for a full professorship in Data Analysis, with a focus on Natural Language Processing! For details, see ▶️ job ad jobs.uni-graz.at/en/jobs/f38f... ▶️ fact sheet cloud.uni-graz.at/s/RA92YSf29z...
jobs.uni-graz.at
Universität Graz
1915
Reposted by Thomas Klebel
Copim @copim.bsky.social · 23/07/2026
NEW POST: 'Wiley’s Acquisition of Emerald: Some observations from Copim': copim.pub/wileys-acqui... The post explores issues of scale and corporate consolidation in academic publishing raised by the acquisition, heightened by the emphasis on the anticipated role of #AI. #Scale #AcademicPublishing
copim.pub
Wiley’s Acquisition of Emerald: Some observations from Copim - Copim
Copim reflects on Wiley’s acquisition of Emerald and the issues of scale and corporate consolidation in academic publishing.
01414
Reposted by Thomas Klebel
Alina Herderich, PhD @hailina.bsky.social · 22/07/2026
New tutorial on a core CSS problem: Text classification often treats "is this valid" and "does this model work" as separate questions hurting measurement quality. We propose "computational social mixed methods" to unify annotation, ML & stats into one coherent pipeline: doi.org/10.3758/s134... 🧵
1124
Reposted by Thomas Klebel
Benjamín Schultz-Figueroa @benjaminschultzfig.bsky.social · 15/07/2026
A really interesting piece on reducing AI cheating in the classroom from a Professor of English and former Dean of Humanities at University of Utah. The solutions? Smaller classes, lower teaching loads for faculty, and in-person classes.
hollisrobbinsanecdotal.substack.com
How to limit unauthorized AI use in the classroom
A handy rule
17355103
Reposted by Thomas Klebel
ruggsea @ruggsea.eurosky.social · 01/07/2026
Here's my new blogpost on the dataset I have created alongside @ksolovev.com: we decided to solve the issue of data availability for research on news articles by processing, cleaning, tagging and indexing the whole news subset of the Common Crawl, resulting in cs2.uni-graz.at/blog/infini-...
12011
Reposted by Thomas Klebel
Eric Schares @eschares.bsky.social · 24/06/2026
OpenAlex announces an improvement to their corresponding author assignment. Haven’t checked it out in detail yet, but big if true. This data field is one of the main gaps that cause me pain.
blog.openalex.org
A big improvement to our corresponding-author data - OpenAlex blog
We’ve made a major, systematic improvement to how OpenAlex finds and assigns corresponding authors, and the corresponding institutions tied to them. On a hand-checked gold standard, precision rose fro...
075
Reposted by Thomas Klebel
Jason Koebler @jasonkoebler.bsky.social · 15/06/2026
New: Researchers have quantified how easy AI search is to manipulate. Just 13 words buried in a random Reddit comment can poison AI search results. They suggest this is not easy to stop: "The way you can attack these systems is so much dumber than you think it is" www.404media.co/it-is-trivia...
404media.co
It Is Trivially Easy to Use Reddit to Manipulate AI Search, Research Suggests
"We show that a tiny snippet—just 13 words—of retrieved text on a UGC website like Reddit, Wikipedia, Quora, or Facebook can change AI agents to output spam / scam content pretty consistently."
5029771170
Reposted by Thomas Klebel
Bart Penders @penders.bsky.social · 12/06/2026
Join us as a postdoctoral researcher! "Invisible Knowledge Work in Open Science Practices", funded by OSNL #STS #metascience Come help us map and understand the invisible labour that enables, allows, supports and shapes #openscience, and lets it grow. Mostly #qual work, with small #quant pockets.
academictransfer.com
Postdoctoral Researcher Invisible knowledge work in open science practices
Welcome to Maastricht University! Are you an ambitious post-doctoral researcher with an Open Science mindset who seeks to unravel the threads of visibility and invisibility of staff across research pr...
13840
Reposted by Thomas Klebel
Berna Devezer @devezer.bsky.social · 04/06/2026
Wonderful news! Journal of Research on Research is launched. "if you are a researcher who researches any aspect of research, then this is the journal and community for you." Worth reading the whole editorial for a transparent and thoughtful account of this amazing community-driven effort.
111042
Reposted by Thomas Klebel
Samuel Moore @samuelmoore.org · 15/05/2026
bsky.app/profile/adam...
001
Reposted by Thomas Klebel
Samuel Moore @samuelmoore.org · 15/05/2026
"SSRN, a preprint server for social sciences research, had the highest rate of hallucinated citations at nearly 2%, almost five times higher than any other major repository." www.nature.com/articles/d41...
nature.com
Hallucinated citations highest in social sciences preprints site
More than 140,000 fake citations across four research repositories were identified in papers and preprints published in 2025 alone.
177
Reposted by Thomas Klebel
Ethan Plaut @ethanplaut.bsky.social · 15/05/2026
“…hallucinated references disproportionately assign credit to already prominent and male scholars, suggesting that LLM-generated errors may reinforce existing inequities in scientific recognition…” 🧪 arxiv.org/abs/2605.07723
arxiv.org
LLM hallucinations in the wild: Large-scale evidence from non-existent citations
Large language models (LLMs) are known to generate plausible but false information across a wide range of contexts, yet the real-world magnitude and consequences of this hallucination problem remain p...
13429160
Reposted by Thomas Klebel
Bart Penders @penders.bsky.social · 06/05/2026
Under the banner "The winding road to better science", we are joining forces with our metaresearch centre and are offering ourselves as hosts for potential MSCA postdoctoral fellows. If you are interested in developing a project on or across the boundaries of #metascience and #STS, reach out! 1/
11715
Reposted by Thomas Klebel
Berna Devezer @devezer.bsky.social · 30/04/2026
📣 The difference between replicable and not replicable is not itself scientifically replicable. 📣 New work with Erkan Buzbas, showing that verdicts such as "X% of results replicated" are based on an inferential machinery that doesn't work. arxiv.org/abs/2604.26268
Screenshot of a paper's title page. Title: "The Difference Between 'Replicable' and 'Not replicable' is not Itself Scientifically Replicable". Authors: Berna Devezer and Erkan O. Buzbas, both at the University of Idaho — Devezer in the Department of Business and the Institute for Modeling Collaboration and Innovation, Buzbas in the Department of Mathematics and Statistical Science. Authors contributed equally. Corresponding author: bdevezer@uidaho.edu.
Abstract: Replication studies estimate the replicability rate of scientific results by aggregating binary verdicts of experiments. Exact replications are rarely attainable, so most replication sequences are non-exact. Experiments differ in ways that matter and do not share a single common data-generating process. We formalize two statistical interpretations of this non-exactness. In a shared latent rate model (benchmark), experiments are exchangeable and depend on a common random replicability rate. In a conditionally independent rates model (operational), each experiment has its own replicability rate drawn independently from a population distribution. Under the shared latent rate model, even small variability among replicability rates induces an irreducible variance floor on the estimated mean replicability rate that cannot be eliminated by adding more replications. Under the conditionally independent rates model, the degree of non-exactness is not identifiable from standard replication data, because one binary verdict per experiment contains no information about between-experiment heterogeneity. Researchers therefore cannot tell which precision regime they are operating in or whether high- and low-replicability sequences can be distinguished in principle. As a result, the usual data structure of one binary verdict per experiment cannot support reliable demarcation between "replicable" and "not replicable" results and systematically understates uncertainty, making high- and low-replicability sequences appear discrim…
15543204
Reposted by Thomas Klebel
Fenner Tanswell @fennert.bsky.social · 22/04/2026
A long read about the state of AI and mathematics. davidbessis.substack.com/p/the-fall-o...
davidbessis.substack.com
The fall of the theorem economy
How AI could destroy mathematics and barely touch it
1126
Reposted by Thomas Klebel
Julia M. Rohrer @dingdingpeng.the100.ci · 22/04/2026
Good news everyone 🥳 Our (w @vincentab.bsky.social) primer on models as prediction machines (with the marginaleffects package) is finally officially published!> journals.sagepub.com/doi/10.1177/...
journals.sagepub.com
Models as Prediction Machines: How to Convert Confusing Coefficients Into Clear Quantities - Julia M. Rohrer, Vincent Arel-Bundock, 2026
Psychological researchers usually make sense of regression models by interpreting coefficient estimates directly. This works well enough for simple linear model...
11366131
Reposted by Thomas Klebel
Daniel Lakens @lakens.bsky.social · 16/04/2026
We are inviting applications for a two-year postdoctoral position in a collaborative meta-science project on the effectiveness of data and code sharing policies in research-performing organizations. www.tue.nl/en/working-a...
tue.nl
Postdoc In Meta-science
Personal type: Scientific staff
15265
Reposted by Thomas Klebel
Charlie Warzel @cwarzel.bsky.social · 26/03/2026
The AI economy looks...really precarious. So @matteowong.bsky.social & I did a bunch of reporting to try to figure out what happens when a potential bubble collides with a war in Iran and a potential resource shortage. The answer is...arguably the most dire stuff i've heard from smart ppl in a while
theatlantic.com
The AI Boom Wasn’t Built for the Polycrisis
“There are too many ways for it to fail for it not to fail.”
632169774
Reposted by Thomas Klebel
Jessica Hullman @jessicahullman.bsky.social · 09/03/2026
Interesting paper. When science is cheap, multiverse analysis becomes the key abstraction for authors & journals to strategize for. These results suggest its the new unit of evidence authoring & review should accommodate. Related to some of my comments here: substack.com/@jessicahull...
1175
Thomas Klebel @tklebel.bsky.social · 02/03/2026
Today I had the pleasure of starting my new role at the Complex Social & Computational Systems group of @janalasser.eurosky.social at University of Graz as a Senior Scientist, to support the group in research data and software engineering tasks.
282
Reposted by Thomas Klebel
Christopher Barrie @cbarrie.bsky.social · 25/02/2026
🧵on my new paper "Synthetic personas distort the structure of human belief systems" w Roberto Cerina I'm v excited about... 🚨 Do synthetic samples look like human samples? We compare 28 LLMs to the 2024 General Social Survey (GSS) to find out + develop host of diagnostics...
618282
Thomas Klebel @tklebel.bsky.social · 22/02/2026
Raises important questions: how will #scholcomm adapt, which norms around publishing will emerge? How will research assessment work in the future? What do we want "research" to look like? Curious to see where @socarxiv.bsky.social will end up with their policy.
012
Reposted by Thomas Klebel
Simon Willison @simonwillison.net · 18/02/2026
Gift link (hope it works) for the @ftrain.bsky.social NY Times piece on the impact of post-November-2025 coding agents (like Claude Code) on the cost of developing software - it's very worth a read www.nytimes.com/2026/02/18/o...
nytimes.com
Opinion | The A.I. Disruption We’ve Been Waiting for Has Arrived
67718
Thomas Klebel @tklebel.bsky.social · 14/02/2026
"The 2026 International Conference on Machine Learning (ICML) has received more than 24,000 submissions — more than double that of the 2025 meeting." That's just absurd. Maybe there was substantial growth before, but this is clearly unsustainable.
100
Thomas Klebel @tklebel.bsky.social · 13/02/2026
Great thread, with lots of important considerations and implications for the metascience community.
110
Thomas Klebel @tklebel.bsky.social · 11/02/2026
Good to see ubiquitous publication bias documented that well for an entire field. I'd expect the same to be true for sociological research, although methods are probably more diverse there.
110
Reposted by Thomas Klebel
Julia M. Rohrer @dingdingpeng.the100.ci · 11/02/2026
A variation: Scientists who claim they’re “not interested in causality” because they assume the term only applies to deterministic, law-like relationships that are unrealistic in their field. Instead, they’re interested in how “X drives Y”, the effects of X, the “extent to which X matters for Y”>
89317
Reposted by Thomas Klebel
Bart Penders @penders.bsky.social · 04/02/2026
The stories we tell build our world, so we should be careful about the stories we tell ourselves and we should be wary of the stories that others tell. This goes for politics and for science alike.
023
Reposted by Thomas Klebel
Ludo Waltman @ludowaltman.bsky.social · 02/02/2026
"Honest, dedicated, skilled researchers may investigate a topic and come to opposite conclusions because of variation between how they conduct their analysis" My sense is almost all experienced researchers are intuitively aware of this, but nevertheless we tend to ignore this inconvenient truth!
022
Reposted by Thomas Klebel
Peder M Isager @isager.bsky.social · 28/01/2026
New blog post introducing Causion - a web app for causal inference teaching and learning: pedermisager.org/blog/causion....
pedermisager.org
Introducing Causion: A web app for playing with DAGs | Peder M. Isager
Personal website of Dr. Peder M. Isager
514567
Reposted by Thomas Klebel
Berna Devezer @devezer.bsky.social · 28/01/2026
Just like democracies are being put to the test around the world (and succumbing sooner than we hoped), our ideals of research integrity and scientific rigor are being tested by ever-invasive AI tools. I am not hopeful science will survive this unscathed. We'll need to rethink a lot of what we do.
26819
Thomas Klebel @tklebel.bsky.social · 27/01/2026
Happy to share our recent article on causal inference in science studies. It aims to introduce causal thinking to the science of science community with an example from Open Science.
182