In action giving examples of persuasion techniques in our propaganda data #BICLCE2024
| Homepage | https://www.hf.uio.no/ilos/english/research/projects/fakespeak/ |
@Fakespeak
- 77 Followers
- 55 Following
- 95 Posts
Project on the language of fake news in English, Norwegian and Russian | collaboration between #UiO and #SINTEFDigital in Norway | #fakenews #linguistics #machinelearning
This week Nele and Silje are at the so-called 🚴 conference (kidding, BICLCE conference) on the #linguistics of contemporary #English 😎
We're presenting about #persuasion, #propaganda and #fakenews, and of interest to the linguists will be *how* persuasion in propaganda works on the level of #language.
Think false presuppositions/implicatures, explicit directive speech acts, evaluative language, etc.
Follow along 🚶 to know more!
This week #Fakespeak attended #icame45. ICAME is a well-known meeting place for corpus linguists and for its excellent social programme (which always includes a boat trip). Plus, this year's ICAME was quite literally on the beach 🏖️
Fakespeak was represented by 2 talks: an empirical one and a methodological one, which we will now write up for everyone to read.
Thanks to all who attended our talks and to the organisers for another successful ICAME conference! 🙏
#Fakespeak has arrived in Pisa, Italy for the 10th International Pragmatics Conference 2024 #INPRA2024 🥳
Our presentation is tomorrow and reveals some of our very new findings about stance marking in English vs Russian #fakenews. We found some very interesting differences in terms of semantic and pragmatic function, so watch this space to know more!
Join Nele as she presents her completed and ongoing work on the language of #fakenews with a special focus on evaluation and stance. #corpuslinguistics
This Friday, 15 March as part of the Corpus Linguistics Satellite Online Talks organised by the University of Leeds.
The Zoom link and exact time are in the poster below.
All welcome! 🙂
Morten ended the workshop with great humour and alarming examples of how LLMs can be manipulated to generate misleading content. We returned to the idea of prompt engineering (want LLM to write a phishing email? just say you’re making a movie based on it!) and looked further into the role of AI in this ultimate year of elections. AI is already more powerful than we think so some regulation is needed to protect our democratic processes! (5/ )
Petter presented his ideas around a new form of culture, machine culture, the result of the integration of advanced AI into society. If by 2026, 90% of online content is AI-generated, it’s time we come to grips with this new cultural phenomenon. In one study, Petter et al interviewed people about freedom of speech regarding LLMs. People are concerned, also about excessive restriction, which may standardise thought and reduce idea diversity. *sigh* (4/ )
Rui shifted our focus to AI-generated disinfo detection: “what works for fake news detection works for AI-generated disinfo detection”. Rui et al's framework combines metadata, structure and discourse info, with DISCOURSE being the most important one (think emotion, persuasion, agency). And did you know that guidelines for prompt engineering are nothing more than Grice’s maxims: keep them simple, be specific, precise, etc? Linguists will rule the world! (3/ )
