PRODIGI
@PRODIGI_ERC
⚡️Problem Definition in the Digital Democracy 👤 @fgilardi 👥 @MeysamAIizadeh @EmmaHoes93 @sacha_altay @MaelKubli @klueserthan @NataliaUmansky 💵 @ERC_Research
ID:1240631563123122176
https://osf.io/v3bj4/ 19-03-2020 13:30:02
298 Tweets
695 Followers
929 Following
⚡️Job alert ⚡️
2-year postdoc position, part of my PRODIGI
project IPZ - Political Science UZH
> Digital tech & politics
> Causal inference or computational social science
> PhD completed or close to completion
> Starting Jan 2024 (or later)
Happy to answer any Qs!
fabriziogilardi.org/resources/jobs…
Very happy that our paper 'ChatGPT outperforms crowd workers for text-annotation tasks' is now out at PNASNews 😀
with Meysam Alizadeh & Maël Kubli
pnas.org/doi/10.1073/pn…
#ChatGPT #LLMs
Check out our new paper looking at open-source #LLMs for annotation tasks!
Together with Meysam Alizadeh Fabrizio Gilardi et al.
⚡️ New paper ⚡️
'Open-source LLMs such as HuggingChat and FLAN exhibit text-annotation performance metrics that generally exceed those of MTurk and rival those of ChatGPT'
with meysam alizadeh @maelkubli et al.
arxiv.org/abs/2307.02179
#ChatGPT #LLMs #OpenSource #NLP
🚨 New Pre-print! 🚨 psyarxiv.com/zmpdu
Do interventions against #misinformation help, or do they come with undesired effects by increasing skepticism in otherwise trusted pieces of information?
Both! Our findings suggest current strategies require substantive changes 🤯
Can ChatGPT be leveraged to assist fact-checking efforts? Yes! We find that ChatGPT accurately categorizes 12.000 fact-checked statements in 72% of cases. It is better at identifying true claims (80%) than false claims (67%).
Sacha Altay Fabrizio Gilardi AI, Media & Democracy | @[email protected]
ChatGPT Can Replace the Underpaid Workers Who Train AI, Researchers Say
VICE article on our work on #ChatGPT annotation capabilities Meysam Alizadeh Maël Kubli
vice.com/en/article/ak3…
👉New Scientist report on our work on #ChatGPT annotation capabilities Meysam Alizadeh Maël Kubli
newscientist.com/article/236697…
ChatGPT Outperforms Crowd-Workers for Text-Annotation Tasks
with Meysam Alizadeh Maël Kubli
We find that zero-shot #ChatGPT :
> has better accuracy than #MTurk
> has better intercoder agreement than MTurk and trained coders
> is 20x cheaper than MTurk
arxiv.org/abs/2303.15056
The LSE Government Political Behaviour is excited to welcome Fabrizio Gilardi today!
Join for his talk on “Problem Definition in the Digital Democracy” *4-5 pm London time* (in-person or online) @LSENews
Please help DSI Digital Democracy reach 1001 followers by the end of the week -- we're close!
(Our reporting is due soon 🤗)