If you are feeling unsafe, I would encourage you to contact either 988 or 911. They will be able to help you.

Perfect Holiday gift! 🎁

Worth every penny! (It’s open access 😉)

No, sorry - in person only!
Really happy you liked the book! :)

Featured article by Anton Törnberg and @pettertornberg.com: www.tandfonline.com/doi/full/10....

To be fair, the fact that my paper has a higher Altmetric than Marx's Das Kapital might be taken to imply that the limitations in the methodology are to my paper's benefit... ;)

Thanks Jonathan.
Democrats still have users but rarely visit the site.
I would point you to the preprint I put up for more details and better versions of the figures: www.arxiv.org/abs/2510.25417

Find my co-authors on Bluesky: @chrisbail.bsky.social @cbarrie.bsky.social

Colleagues who do excellent work in this field, and might find these results interesting:
@mbernst.bsky.social
@robbwiller.bsky.social
@joon-s-pk.bsky.social
@janalasser.bsky.social
@dgarcia.eu
@aaronshaw.bsky.social

This has been carried out by amazing Nicolò Pagan, with Chris Bail, Chris Barrie, and Anikó Hannák.

Paper (preprint): arxiv.org/abs/2511.04195

Happy to share prompts, configs, and analysis scripts.

Takeaways for researchers:
• LLMs are worse stand-ins for humans than they may appear.
• Don’t rely on human judges.
• Measure detectability and meaning.
• Expect a style–meaning trade-off.
• Use examples + context, not personas.
• Affect is still the biggest giveaway.

We also found some surprising trade-offs:
🎭 When models sound more human, they drift from what people actually say.
🧠 When they match meaning better, they sound less human.

Style or meaning — you have to pick one.

So what actually helps?
Not personas. And fine-tuning? Not always.

The real improvements came from:
✅ Providing stylistic examples of the user
✅ Adding context retrieval from past posts

Together, these reduced detectability by 4-16 percentage points.

Some findings surprised us:
⚙️ Instruction-tuned models — the ones fine-tuned to follow prompts — are easier to detect than their base counterparts.
📏 Model size doesn’t help: even 70B models don’t sound more human.

Where do LLMs give themselves away?

❤️ Affective tone and emotion — the clearest tell.
✍️ Stylistic markers — average word length, toxicity, hashtags, emojis.
🧠 Topic profiles — especially on Reddit, where conversations are more diverse and nuanced.

The results were clear — and surprising.
Even short social media posts written by LLMs are readily distinguishable.

Our BERT-based classifier spots AI with 70–80% accuracy across X, Bluesky, and Reddit.

LLMs are much less human-like than they may seem.

We test the state-of-the-art methods for calibrating LLMs — and then push further, using advanced fine-tuning.

We benchmark 9 open-weight LLMs across 5 calibration strategies:
👤 Persona
✍️ Stylistic examples
🧩 Context retrieval
⚙️ Fine-tuning
🎯 Post-generation selection

We use our Computational Turing Test to see whether LLMs can produce realistic social media conversations.

We use data from X (Twitter), Bluesky, and Reddit.

This task is arguably what LLMs should do best: they are literally trained on this data!

We introduce a Computational Turing Test — a validation framework that compares human and LLM text using:

🕵️‍♂️ Detectability — can an ML classifier tell AI from human?

🧠 Semantic fidelity — does it mean the same thing?

✍️ Interpretable linguistic features — style, tone, topics.

Most prior work validated "human-likeness" with human judges. Basically, do people think it looks human?

But humans are actually really bad at this task: we are subjective, scale poorly, and very easy to fool.

We need something more rigorous.