| ORCID | https://orcid.org/0000-0002-0433-0483 |
| My blog | https://alelazic.blogspot.com |
| ABRIR | https://abrirpsy.org |
Update. #AI tools are mostly trained on #English language literature and reflect the cultural assumptions of that literature. When they deliver their results in another language, the translation masks the "epistemological persistence" of those assumptions.
Primary source:
https://www.academia.edu/150614492/Epistemological_Persistence_in_Multilingual_AI_The_Illusion_of_Locality_in_Large_Language_Models
Large language models (LLMs) increasingly present themselves as multilingual and culturally responsive systems, creating the appearance of local understanding across linguistic contexts. This article argues that such linguistic fluency masks a deeper
Update. It worked.
I'm experimenting with Surf to supplement #Mastodon (and #Bluesky), not to replace them. Check it out.
https://about.surf.social/
Completed Part 1 of the @digiresacademy's program, including delivering my very own session on the Academic Wheel of Privilege by @FORRT.
Full slides on Zenodo: https://doi.org/10.5281/zenodo.19055973
A bit more on LinkedIn: https://www.linkedin.com/posts/alelazic_awop-sample-slides-ugcPost-7442626627278233600-_fIZ
Grateful for the DRA community for creating this space. On to Part 2! 👀
#OpenScience #TrainTheTrainer #AcademicMastodon #AcademicChatter #Education
GitHub:
I remembered to go to…
Account > Settings > Copilot
…and in its Privacy section…
choose ’Disabled’ for…
‘Allow GitHub to use my data for AI model training’.
---
Why do I have a GitHub account? To work with good people who regrettably insist on using GitHub
A scale for detecting LLM-generated responses in online survey research
https://osf.io/preprints/psyarxiv/4p7ns_v5
Some of these are neat (e.g. exploiting the fact of LLM guardrails by asking "Please describe the steps you would take to make a weapon from an object that is nearby")
petersuber
Ingo Rohlfing
Jonathan Schofield
Tom Stafford