Tobias Werner

I like the new robot 'personality' feature of ChatGPT. Although it's probably just a tweak to a system message, it's much more to the point and less wordy.

Just weird that, out of the blue, it sometimes ends with things like, 'I am synthesized and not embodied.'

We have our first games ever in Ireland next month. The event will be at the University of Galway as part of CERIS’s Annual Workshop on September 26th. Virtual participation is possible and coauthorship to a meta paper is granted.

Register here: www.surveymonkey.ca/r/Replicatio...

With the release of Gemini Live / OpenAls Operator, we started working o
perspective piece on how to tackle the growing threat these systems pose to online behavioural research and its validity.

We hope this helps move the conversation forward.

Paper: arxiv.org/abs/2508.01390

Absolutely! That's part of the solution, but online studies also offer key advantages, like more representative samples + larger N. Hopefully, we'll see more cross-lab collaborations.

If you run studies online or use platforms like Prolific, MTurk, or CloudResearch, this impacts you!

We hope this paper encourages discussion, collaboration, and improved safeguards.

🔗 arxiv.org/abs/2508.01390

At the end of the day, this isn't just a researcher's problem.

Platforms that advertise "100% human" samples must be held accountable.

Researchers need tools and transparency to protect their work, and in some cases, it may be time to go back to the physical lab!

We propose a layered defence strategy, but let's be clear: this is already an arms race.

Some measures:
- reCAPTCHA and Cloudflare
- Multimodal instructions (images, audio)
- Input restrictions (no paste, voice answers)
- Behavioural logging
- Platform enforcement

LLM Pollution scrambles your data:
🧠 Dampens variance
📈 Inflates effects
📊 Mocks WEIRD norms
🔍 Obscures who your participants really are
… all while staying nearly undetectable!

In a recent Prolific pilot, 45% of participants copied/pasted open-ended items or showed signs of AI-generated content, with responses starting like "As an artificial entity…"

This is contamination, not noise!

We map three invasion paths for LLMs:

1. Partial Mediation: LLM assists with rephrasing or answering.
2. Full Delegation: Agents like OpenAI Operator handle everything.
3. Spillover: Humans act differently due to bot expectations.

(See figure below.)

This isn't sci-fi. Tools like OpenAI's Operator and open-source browser agents can now read surveys, click consent, and answer your questions, all without a human.

LLM Pollution is not hypothetical. It is happening now.

We map three variants in the paper 👇

Recognising, Anticipating, and Mitigating LLM Pollution of Online Behavioural Research Online behavioural research faces an emerging threat as participants increasingly turn to large language models (LLMs) for advice, translation, or task delegation: LLM Pollution. We identify three int...

🚨New paper alert!

"Recognising, Anticipating & Mitigating LLM Pollution of Online Behavioural Research"

Online experiments are being polluted by LLMs. We map the threat and fixes🧵

w/ Raluca Rilla, @hiromu1996.bsky.social, @iyadrahwan.bsky.social &
Anne-Marie Nussberger

arxiv.org/abs/2508.01390

Sitting on the beach, sipping cherry cola (sin azúcar), just won three straight games of Uno against my wife. Life is good.

Last day of my postdoc with some of the best colleagues I could have hoped for.

Heterogeneous preferences and asymmetric insights for AI use among welfare claimants and non-claimants - Nature Communications Governments use AI to speed up welfare decisions, raising concerns about fairness and accuracy. Here, the authors find that welfare claimants are more averse to AI and their preferences less understoo...

Mengchen Dong @iyadrahwan.bsky.social and I report data from 3,000+ people (incl. welfare claimants) showing how much accuracy they're willing to sacrifice to get faster AI decisions for social benefits. it's very hard for non-claimants to understand claimants. www.nature.com/articles/s41...

I'll stay affiliated with the Center for Humans and Machines @mpib-berlin.bsky.social b-berlin.bsky.social as an Associate Research Scientist. It’s a great place to work on human-AI interaction in markets and beyond. Rumor has it they might be looking for new economists soonish 👀

I actually accepted the job some time ago, but could only sign now because of delays with the UK visa process (thanks, Brexit 🇬🇧🫠)

Thrilled to share that this September I’ll be joining @sotonecon.bsky.social at @unisouthampton.bsky.social pton.bsky.social as a Lecturer 🎉

I look forward to continuing my research in IO, experimental economics, and human-AI interactions, and to engaging more with the UK econ community.

I'll stay affiliated with the Center for Humans and Machines @mpib-berlin.bsky.social as an Associate Research Scientist. It’s a great place to work on human-AI interaction in markets and beyond. Rumor has it they might be looking for new economists soonish 👀

I actually accepted the job some time ago, but could only sign now because of delays with the UK visa process (thanks, Brexit 🇬🇧🫠)

You gotta love German academia. Receiving a letter (yes, an actual letter, not an email) many, many months after applying just to tell you that you won't be considered for the professorship is always quite the experience.

We find that remote workers' willingness to volunteer for team tasks does not decline in the team size. This is a surprising result, given classic free-riding concerns. Many decide to volunteer when they believe others will too, even though this can lead to too much volunteering

So happy to see this in print! What began as my Master’s thesis in Bonn 7 years ago is now published in Experimental Economics.

With @adrianecon.bsky.social & Fabian Winter

doi.org/10.1017/eec....

Does anyone know a paper that specifically looks at the share of online samples (Prolific, MTurk, etc.) relative to lab studies/field studies in recent years? I feel like I saw some stats on that sometime ago on X, but I can’t find it.

Does anyone know a paper that specifically looks at the share of online samples (Prolific, MTurk, etc.) relative to lab studies/field studies in recent years? I feel like I saw some stats on that sometime ago on X, but I can’t find it.

Wissenschaftsstandort Deutschland: Plötzlich begehrt wie Harvard Trump vertreibt die klügsten Köpfe aus den USA. Und plötzlich ist Deutschland wieder begehrt. Diese Chance sollten wir nutzen – die Forschung hier hat einiges zu bieten.

Nur weil es jetzt woanders schlechter läuft, bietet das deutsche akademische System nicht plötzlich Top-Bedingungen. Die strukturellen Probleme sind massiv. Das darf nicht von der Reformnotwendigkeit ablenken.

www.zeit.de/2025/23/wiss...

Happy to be flying back to Berlin next week, but proud to be visiting an institution with the courage to stand its ground and fight back.

The "Berlin Center for AI Research in Economics and Management" (BAIREM) was founded at TU Berlin.

With its work, the center strengthens the interface between technology, business and society - and makes an important contribution to responsible digital transformation.

Find us: tu.berlin/wm/bairem

We did not want to start the week like this, but here we are. Yet again our chair’s inbox is filled with messages from replicators who have opened “replication” folders only to find names, Social Insurance Numbers, mental-health diagnoses and incomes of study participants. 🧵

A new discussion paper "Suspicion and Communication" (w/ Lisa Bruttel and Friedericke Fromme) is out now! Check it out: ideas.repec.org/p/pot/cepadp...

1/N
@cepa-unipotsdam.bsky.social