Jan Zilinsky

The mimicking behavior is also interesting: I listened to Slovak news this morning on the radio and the reporter also said who "is expected to win".

The video is from July, so "announces" isn't quite right.

Also, a response to that video at the time seems to contain a small hint of paranoia

Hard to distinguish from opportunism...

- Will certain people always continue "just asking questions"? Yes.

- But citizens have the right to ask questions and, in a tolerant society, they could ask them without getting mocked

- It's a bad case here, because how could anyone prove a negative? (Can't show "we're not altering the weather")

Came here to write that but you've saved me the typing

Liked this passage in @smotus.bsky.social's book - though there are places were non-confusing design is the exception and a cryptic design is the default...

Opinion | The A.I. Prompt That Could End the World

Latest piece centering scenarios where AI could autonomously develop harmful capabilities like engineering lethal pathogens

But lowering the barrier for malicious actors to cause harm is a risk that's real today

So, of course we get stories about hypothetical risks:
www.nytimes.com/2025/10/10/o...

On morning Munich commutes, I see surprisingly little tech use. People read newspapers, nap, or chat. Occasionally someone knits, and I watch a bit wistfully.

Kids seem perfectly well-adjusted: those in groups chat and tease each other, those commuting alone do homework or use phones.

Opinion | We Really Want to Trust Crypto Interests With the Future of Money?

"the crypto lobby is pushing Congress to pass the Digital Asset Market Clarity Act, which ... would bar the Federal Reserve from issuing or even piloting a government-issued digital currency ... without congressional approval" www.nytimes.com/2025/09/29/o...

Sage Journals: Discover world-class research Subscription and open access journals from Sage, the world's leading independent academic publisher.

"Acceptance of the scientific consensus was very high in the sample as a whole (95.1%), but also in every sub-sample (e.g. no trust in science: 87.3%) ... [P]eople are motivated to reject specific scientific beliefs, and not science as a whole."

journals.sagepub.com/doi/abs/10.1...

Motivated reasoning (and responding!) is a huge issue for polling - BUT, the magnitude of the problem also depends on what else is asked within a given survey: www.cambridge.org/core/service...

I thought GPT-5 was supposed to tone down the flattery 😂

A friend’s kid told me in 2015: “I can learn anything I want by finding a tutorial on YouTube.”

Did I find it credible? Not quite, but the sentiment was fascinating.

Still think about it sometimes and also wonder how many people now use video-sharing platforms as their search engines

Very cool that people can be open-minded

If you've felt that broadcast news is less informative, you're probably right.

Information density of ABC, CBS, and NBC news segments has declined, according to LLM-based classifications: 5harad.com/papers/no-ne...

Picture of the Vancouver waterfront

Award from the ITP section for my paper with Jan Zilinsky

View from Cypress Mountain

Thanks to the folks at APSA for organizing a great conference, and Vancouver for being a great host city!

cc @janzilinsky.bsky.social

Axios article headline about Spencer Cox taking a quieter approach

Finally, as I found in research my on nasty politics, elite rhetoric matters—especially from national leaders.

Republican Governor of Utah, Spencer Cox, has sought to reduce the political temperature following the assassination of Charlie Kirk. www.axios.com/local/salt-l...

The social sciences face a replicability crisis. A key determinant of replication success is statistical power. We assess the power of political science research by collating over 16,000 hypothesis tests from about 2,000 articles in 46 areas of the discipline. Under generous assumptions, we show that quantitative research in political science is greatly underpow- ered: the median analysis has about 10% power, and only about 1 in 10 tests have at least 80% power to detect the consensus effects reported in the literature. We also find substantial heterogeneity in tests across research areas, with some being characterized by high power but most having very low power. To contextualize our findings, we survey political methodologists to assess their expectations about power levels. Most methodologists greatly overestimate the statistical power of political science research.

The pretty draft is now online.

Link to paper (free): www.journals.uchicago.edu/doi/epdf/10....

Our replication package starts from the raw data and we put real work into making it readable & setting it up so people could poke at it, so please do explore it: dataverse.harvard.edu/dataset.xhtm...

Used it last year, and wrote a LlooM-inspired article the other day (and could've run the concept through the Lloom infrastructure of course...) goodauthority.org/news/the-eps...

People in Denmark have varied concerns about tech giants:

- Impact on young people's development (83.5% concerned)
- Role in spreading misinformation (81%)
- Ability to influence public opinion (79.5%)
- Economic power (79%)
- Collection of personal data (78%)
- Impact on social cohesion (75%)

So why do most of us rarely switch on the airplane mode?

(Not dismissing what the respondents are saying, just wondering what exactly they are telling us about their unhappiness with the current tech)

Every generation discovers that the latest technology is destroying civilization.

Every generation is wrong.

Also... close to nothing for a few of these:

Different AI systems are optimized for different tasks.

Expecting precise instruction following from current image models shows a misunderstanding of their architecture and training objectives:

I asked 4 AI models to rate the quality of different ideas (all about changing jobs, but for different reasons)

1) Models seems to take users' happiness seriously
2) Gemini flatters users the most, Grok is harsher as an evaluator
3) Overall, it doesn't seem like LLMs were excessively deferential

GitHub - zilinskyjan/information-and-influence-course Contribute to zilinskyjan/information-and-influence-course development by creating an account on GitHub.

Here are my lecture slides on the behavioral economics of political persuasion

Slides: zilinskyjan.github.io/information-...
Repo: github.com/zilinskyjan/...

True for young adults and for many of us older adults too :)

But right, I should have said I hope 'many will choose the latter'.

As faculty it seems we need to think hard about how we encourage students to use the new tools well

"Using ChatGPT to write your essays is like bringing a forklift into the weight room"

Not quite.

It's like bringing a fitness trainer to the gym with you; she's *willing to lift weights* for you, but stands ready to give you excellent feedback on your form

It's up to students to choose the latter

Tried making an image of an economics seminar :)

How did I miss this amazing news?

Excited to keep learning from your research, Dominka!