@EclecTweetz That’s true, average score is computed as a trimmed mean. That means max- and min-scoring judges have a more indirect influence. That doesn’t mean a biased judge cannot effect the results (they are replacing an unbiased judge nonetheless).
The scoring scandal in Olympic ice dance bugged me enough to visualize the data.
The French judge was a clear outlier. She scored the French couple higher and the US couple lower than every other judge.
@DanW9 I’m really sorry I couldn’t live up to your integrity standards, and I’ll try to do better next time. But hey, thanks for giving me the opportunity to set the record straight!
@DanW9 If both scored the average of the other 7, the French couple would still keep gold by 0.19 points!
But “what if” analysis may be a red herring: consistent bias on both sides is very problematic and casts doubt on the integrity of judging.
If the French judge had scored like the average of the other eight in both segments, gold and silver would have flipped. No other single judge's removal changes the result.
Data from https://t.co/wO4CE0xcmp
French judge's bias was consistent. She favored the French couple and punished the US couple on nearly every scored item. The US judge, who only served on the Free Dance panel, showed a similar bias favoring the US pair, but it was more muted, especially on the anti-France side.
@cbarrie Thanks for sharing! Our psychometric replications also question LLMs as naive subjects. E.g, how can they truly simulate naive subjects on a social trust survey with the survey and literature itself leaking into their training data?
https://t.co/f1PFq83zkl
@SolomonMg@cbarrie Pre-print here: https://t.co/f1PFq83zkl
Some argue LLMs can replace naive subjects in psych experiments. We find their evidence unsound: models aren't just contaminated by training data—the entire relevant literature IS their training data..
@SolomonMg@cbarrie What a timely coincidence. Our new paper w/ Bogdan State argues (and shows why) that LLMs replicating psych phenomena doesn't prove naive subject capability, as a large portion of social science literature is already in their training data.
@xinmingd@DinaPomeranz@jenniferdoleac I was curious about the hypothesized actual, mechanistic aspects (e.g., reading tweets, FB group posts versus talking on the phone).
For example, how do we know "The transmission channel is the social media platforms" and not some other social process that links the two regions?
@xinmingd@DinaPomeranz@jenniferdoleac I may have missed it, but do you discuss what possible mechanisms may link air pollution to increased remote crime rates? Section 9 discusses mechanisms for local effects.
Is the idea ↑pollution -> ↑aggressive social media posts -> ↑crime in remote, socially linked regions?
Sokağa çıkma yasağının uygulandığı, eski genel nüfus sayımlarının bebek isimlerine olan etkisini merak ettiniz mi?
Ben ettim. Çıkan sonuçlar çok makul ve hemen inanacaksınız.
https://t.co/WLg2lD35IJ
Bir bebek ismi olarak Nisa'nın Nisan aylarında popülerleşmesi bize muhafazakar kesimin isim tercihlerinde sekülerleşme eğilimi hakkında ipucu verebilir mi? Bence verir. Yeni isim notum: https://t.co/DiJekk4gLx
Şampiyonluk maçı öncesi @karakurtebrar182'ın ismi Ebrar ile ilgili ufak bir araştırma yaptım: https://t.co/RzhxjZKSmo
1990'lardan sonra popülerleşmeye başlayan Ebrar ismi, 2017 Ocak ayında ani bir patlama yapmış (kırmızı çizgi 1 Ocak 2017).
Şampiyonluk maçı öncesi @karakurtebrar182'ın ismi Ebrar ile ilgili ufak bir araştırma yaptım: https://t.co/RzhxjZKSmo
1990'lardan sonra popülerleşmeye başlayan Ebrar ismi, 2017 Ocak ayında ani bir patlama yapmış (kırmızı çizgi 1 Ocak 2017).
Şampiyonluk maçı öncesi @karakurtebrar182'ın ismi Ebrar ile ilgili ufak bir araştırma yaptım: https://t.co/RzhxjZKSmo
1990'lardan sonra popülerleşmeye başlayan Ebrar ismi, 2017 Ocak ayında ani bir patlama yapmış (kırmızı çizgi 1 Ocak 2017).
Ebrar Alya etkisine güvenimi arttıran bir sebep aynı etkiyi Alya ismi için de görmemiz.
Ebrar'dan farklı olarak Alya'ya olan ilginin devam ettiğini görüyoruz.