Andres Karjus

Interesting work, but man what a missed opportunity to make use of the (for once totally justified) xkdc ggplot theme here :D

Humanities can greatly benefit from ML/AI (esp with the wealth of data digitized now), and could & should be at the center of the AI revolution to help design smarter, less biased, more human and humanistic models.
Unfortunately, in many places this sort of dogmatic romantic primitivism prevails.

We just created an 1st year undergrad course on digital competencies (incl AI), 300 students signed up, and one of the first lectures was literally titled "what is a computer" (several profs thought it's needed now). From what I heard was well received too.

This does puzzle me. People have done interrater agreement, train-eval-test etc for ages, but now this one type of ML model is seen as some oracle (term from the Messeri et al typology which fits well I think). You're doing the good work as editor if you call them out though, thanks!

Out now in @springernature.com Journal of Cultural Economics:
"Socioeconomic factors of national representation in the global film festival circuit:
skewed toward the large and wealthy,
but small countries can beat the odds"
🔓 link.springer.com/article/10.1...

People into magical thinking (outside of their free time) should ideally have no business doing science :D but I guess it is what it is

And isn't the (limited but often only feasible) solution to both human/machine annnotations to just use expert test sets, assess error &take it into account in stats? Like what else is there to do? You seem to criticize DSL, but what's the better way? (besides just... not doing large scale research)

Nice work; but question: is the issue w LLMs really any different from any other annotation procedures? Including human annotators, can also be low quality/erroneous (eg lazy crowdworkers/students, bad instructions) &can also be "hacked" (data forgery/manipulation). Is this really an LLM problem?

Quantifying the global film festival circuit: Networks, diversity, and public value creation Film festivals are a key component in the global film industry in terms of trendsetting, publicity, trade, and collaboration. We present an unprecedented analysis of the international film festival ci...

This is, to our knowledge, the largest-scale global study of socioeconomic representation in film festivals to date.

It also expands & adds the cultural econ angle to our recent "Quantifying the global film festival circuit" paper: journals.plos.org/plosone/arti...

We find that while the festival circuit does favor a few large&wealthy countries (we suggest a few ways to measure&model this), some small ones (eg 🇮🇸 🇺🇾 🇱🇺 🇲🇳 🇪🇪) perform way better than expected by their socioeconomic & geo profile (location matters btw!)

We analyze country representation & diversity at 600 film festivals over a decade (20k films, 163 countries, 30k appearances), combining several databases and using various metrics, stats and simulation methods & mapping cultural flows & exchange parity of films across borders.

Out now in @springernature.com Journal of Cultural Economics:
"Socioeconomic factors of national representation in the global film festival circuit:
skewed toward the large and wealthy,
but small countries can beat the odds"
🔓 link.springer.com/article/10.1...

I love the smell of billion-dollar publisher typesetting fcking up our formulas in the morning

And last but not least, grateful to Estonian Business School for awarding me with the EBS Research Prize of 2025 just today!

Then also, in September I'm visiting Linz in Austria via the EdTech Talents program, collaborating with private sector and academic partners on related topics of media literacy education and AI.

3 little updates. Started as assoc prof of computational social science at Uni Tartu (part time) in the ERC funded RiDe project. We use large historical media corpora &LLMs to explore dynamics of industrialization &spread of related ideas/sentiments, among other things.

We wrote a little overview chapter on Evolutionary Linguistics with Jonas Nölle and @stefanhartmann.bsky.social for the upcoming Encyclopedia of Language and Linguistics (3rd ed), edited by Nesi&Milin. Preprint here: osf.io/preprints/ps... (with extra references! the handbook had a limit on those)

On 20th August we celebrate the 1991 restoration of independence from Soviet Union - restoration, because Estonia was already a legitimate independent republic before WW2, during and after which USSR (part of its pact with nazi Germany) tried to occupy us for half a century.

Got invited to the presidential rose garden celebration of the Estonian Restoration of Independence day; always nice when the work one does gets recognized in these ways too.

If you're also at the #ic2s2 and interested in how #genAI affects creative fields, come see
"Expertise Elevates AI Usage: Experimentally Comparing Laypeople to Professional Artists"

Wednesday 11AM Human-AI session, room Vingen-1.
Preprint: arxiv.org/abs/2501.12374
#AI #art @ic2s2.bsky.social

If you're also at the #ic2s2 and interested in how #genAI affects creative fields, come see
"Expertise Elevates AI Usage: Experimentally Comparing Laypeople to Professional Artists"

Wednesday 11AM Human-AI session, room Vingen-1.
Preprint: arxiv.org/abs/2501.12374
#AI #art @ic2s2.bsky.social

- built by students who had never written a line of code in their lives.

Shows that you can just take a few days, maybe get somebody like me to provide a few pointers, and just DO stuff.

For example, we had: an AI-driven study assist app, a cooking recipes app, couple of tourism related apps and brochures, a custom-GPT assistant for a bike repair service, and an entirely vibe-coded, small but working anime game, with generated visuals and music -

And sure, I was ready to just organize them into study groups, that’d be fine too. Only a couple people had some programming experience. But by day 4, all groups had built and presented functional prototypes or demos!

Finished teaching another practical generative AI course in the Tallinn Summer/Winter School series at Tallinn University. Like last time, lectures and teamwork time. Again an international mix of industry people and students. On day 1, most said they just want to learn a bit about AI...

Arrived at a linguistics conference (been a while, have done other stuff in the meantime), and the first plenary is... reading their entire talk off a paper, reciting their latest publication I guess. Feels weirdly nostalgic but also holy shit ppl still do that

Last change to apply for our postdoc in the deep transitions psotdoc! Come do big data analytics, work with large media corpora and enjoy a pretty good salary to living cost ratio in Unesco's City of Literature!

The PhD position deadlines have now passed, but these research fellow calls are still open:
Comp social science, big data, time series - come work with me! euraxess.ec.europa.eu/jobs/337755 (16.06)
DH, corpora - work with our wonderful colleagues in the humaities: ut.ee/en/job-offer... (02.06)

That is pretty much what we have planned already. The course is called "Digital skills and AI", so of course there will be content on how to use (&not use) AI, and we allow using AI when writing (just request short explanation). But then it got me thinking about assistance as such.

Agree agency is crucial. And it's definitely a gradient, where human assistants were also previously positioned, but indeed they wouldn't bother doing stuff for you (unless paid enough I guess). I don't know the answer either, but I'm writing course guidance right now myself and got me wondering