@SuneAuken En av grunnene er nok at formatet er veldig annerledes. Det er vanskelig å omstille seg til å ikke komme med originale bidrag men bare gjengi eksisterende kilder uten et forskningsspørsmål eller hypotese eller noen av elementene som er standard i akademisk forfatterskap. Jeg har hørt […]

Their findings echo personal experience: Working with hand-coded data on legal issues and doctrine in case law of the European Court of Justice (https://doi.org/10.1177/2755323X251337809) we made serious attempts to replicate our findings using machine learning, but did not manage to create […]

The variation cannot be explained away by the challenges inherent in text annotation: The authors "find no correlation between human inter-annotator agreement and LLM hacking risk, meaning that even for tasks where human experts perfectly agree, LLM-based annotations can yield unreliable […]

While LLMs can be easily abused, good intentions are not going to help us avoid these problems: When statistical significance of real effects creep up towards p=.05, LLM error rates approach 70%. Choices of prompt engineering "contribute less than 1% of explained variance", raising serious […]

Malicious actors can easily manipulate LLMs to reach pretty much any conclusion: In as many as 68% of cases, statistically significant effects could be reversed to display findings diagonally opposed with reality. With slight tweaks they could create false positives in 94% of cases, whereas true […]

@JosteinSN Det er vanskelig å se for seg ellevill jubel mellom de døde og de utsultede i ruinene. Konseptet fredsavtale spiller på illusjonen av en krig. Dette er bare en pause i et folkemord som Netanyahu har alle intensjoner om å fortsette.

Jeg kan puste mer normalt når nødhjelpen kommer […]

Sure, good thinking! https://doi.org/10.5281/zenodo.17207970

I compiled some data on the UKSC some time back for a presentation I was giving on web scraping.

The UKSC changed their web site pretty much immediately after, but I guess the data still holds in case anyone is interested in the UK.
https://brekke.it/?p=data&set=uksc

Det ser ut til at det fungerer fint med Bluesky også. Ikke verst.
@aslakr

Forhåndsstemte MDG fra utlandet, det føles godt! @slaeg

I guess this would have to be implemented by @anewsocial . Probably not on top of their priorities, but I'd love to heat it it has been considered.

@johank76 @bsky.brid.gy

@DetersHenning Where is the replication data available from? I can't seem to find it under supplementary materials on Cambridge or on your website under resources. Not that I would need it any time soon - just curious!

@DetersHenning Sad to hear you won't make it to Copenhogen! Hopefully we'll make up for it at some European conference next year.

It always fascinated me how much has been done on very limited information about observations in PR procedures, yet nobody seemed to care about information about […]

Great read! It plays directly into a side project I'm working on myself these days, so the timing is perfect for me: I was planning on writing a paragraph or two about the relationship between authority transfer and politicization today anyway. That should be a lot easier now.

The drop in […]

The take-home message: Overall, mobilization against EU legislation, implementation, and enforcement is growing, but the salience and polarisation of the respective Court proceedings depends on how much authority the EU wielded in these domains in different […]

[Original post on eupolicy.social]