@euncheolchoi.bsky.social
PhD student at USC Annenberg; LLMs / CSS / SNA; @ HUMANS Lab (http://www.emilio.ferrara.name/)
Our Python snippet for measuring wAII for a job posting, which is open-sourced.
Our code is open-source. With just a few lines of Python, anyone can measure AI exposure for any job description using our repository. (5/5)
Paper: workshop-proceedings.icwsm.org/abstract.php...
Repository: github.com/EunCheolChoi...
In the information industry sector, our wAII is โpositivelyโ associated with offered salary.
On the other hand, in the wholesale trade industry sector, wAII is โnegativelyโ associated with offered salary.
๐๐ก๐๐ญ ๐ฐ๐ ๐๐จ๐ฎ๐ง๐
1. Jobs in tech, manufacturing, and engineering are more exposed to AI; HR, public service, and legal sectors are less exposed
2. Exposure to AI shows distinct association patterns regarding offered salary depending on the industry sectors (4/5)
A diagram explaining how we extracted โtasksโ from each job post. A task is a role or a skillset that is essential for a job position.
We query based on extracted tasks and retrieve the most similar AI-related patents in terms of semantic similarities.
Weighted AI Index is calculated as the sum of the product between the task weight and the similarity of the task and the retrieved patent.
Full LLM prompt for extracting tasks from a job post.
To track this, we developed the Weighted AI Index (wAII), a scalable method to measure how closely a jobโs tasks align with recent AI innovations.
We extract key job tasks from postings; compare them to AI-related patents; compute an โAI exposureโ score. (3/5)
A bar chart comparing โweighted AI index (wAII)โ of different industry sectors. wAII is our proposed method of measuring how much a job position is exposed to AI technologies.
๐๐๐ฒ ๐ข๐๐๐
AI doesn't affect all jobs equally. Some industries and roles are far more exposed to disruption from technological innovation than others. (2/5)
Title page for our new paper, โMapping Labor Market Vulnerability in the Age of AIโ
๐พ๐๐๐ ๐จ๐ฐ ๐๐๐๐ ๐๐ ๐๐๐?
Itโs a question that keeps many of us up at nightโand for good reason.
Our new research maps labor market vulnerability in the age of AI with 100K job postings and 50K AI-related patents. (1/5)
Keeping Track of AI Use Cases in the Newsroom generative-ai-newsroom.com/keeping-trac...
15.06.2025 19:29 โ ๐ 2 ๐ 0 ๐ฌ 0 ๐ 0
"Limited effectiveness of LLM-based data augmentation for COVID-19 misinformation stance detection" by @euncheolchoi.bsky.social @emilioferrara.bsky.social et al, presented by the awesome Chur at The Web Conference 2025
arxiv.org/abs/2503.02328
Hot takes:
- The benefits of easily accessible social media data usually outweigh the potential harms.
- Some uses of (public) social media data are unethical/should be illegal, and we should target that.
- We're better off having clear boundaries between public and private online spaces.
An employee of Huggingface, a site of AI training datasets, made a dataset of a million Bluesky posts scraped simply because they could. Itโs currently trending: www.404media.co/someone-made...
27.11.2024 00:09 โ ๐ 1109 ๐ 476 ๐ฌ 60 ๐ 196Bluesky's firehose is a treasure trove of public data for researchers and developers, and it's completely free. Check out our developer docs: docs.bsky.app
23.11.2024 05:54 โ ๐ 7934 ๐ 1532 ๐ฌ 321 ๐ 167
Book outline
Over the past decade, embeddings โ numerical representations of machine learning features used as input to deep learning models โ have become a foundational data structure in industrial machine learning systems. TF-IDF, PCA, and one-hot encoding have always been key tools in machine learning systems as ways to compress and make sense of large amounts of textual data. However, traditional approaches were limited in the amount of context they could reason about with increasing amounts of data. As the volume, velocity, and variety of data captured by modern applications has exploded, creating approaches specifically tailored to scale has become increasingly important. Googleโs Word2Vec paper made an important step in moving from simple statistical representations to semantic meaning of words. The subsequent rise of the Transformer architecture and transfer learning, as well as the latest surge in generative methods has enabled the growth of embeddings as a foundational machine learning data structure. This survey paper aims to provide a deep dive into what embeddings are, their history, and usage patterns in industry.
Cover image
Just realized BlueSky allows sharing valuable stuff cause it doesn't punish links. ๐คฉ
Let's start with "What are embeddings" by @vickiboykis.com
The book is a great summary of embeddings, from history to modern approaches.
The best part: it's free.
Link: vickiboykis.com/what_are_emb...
New Paper on Opportunities and risks of LLMs in survey research papers.ssrn.com/sol3/papers.... " Backed by both practical examples & academic literature, we identify areas
for research and development, distinguishing between challenges related to survey methods &
the tools used to deploy surveys"
Interested in RLHF, DPO, LLM alignment?
I've just created this list featuring awesome people like @natolambert.bsky.social .
The list is the opposite of exhaustive; I've just joined some days ago ๐
go.bsky.app/MqRGAf2
If you're keen to learn content verification for fact-checking and open source investigations, this is a step-by-step guide on how to verify images and videos that I posted a while ago.
Once you familiarise yourself with reverse search, you'll become much better at spotting online misinformation.
More #data2thepeople! #election2024
We just released a Twitter/X dataset of 22m+ tweets about the upcoming election! (Data will be routinely updated, keep an eye and spread the word!)
Paper arxiv.org/abs/2411.00376
Data github.com/sinking8/usc...
#Data2thePeople! Pls reshare!
Do you want a billion telegram posts?!
You get data! You get data! You get data!
Paper arxiv.org/abs/2410.23638
Data github.com/leonardo-bla...
#election2024
Sharing my first Computational Social Science starter pack! Will grow with time, feel free to nominate and self nominate!
go.bsky.app/CYmRvcK
Here is a @github.com repo where I will share tutorial notebooks on how to retrieve and analyze data from @bsky.app @atproto.com
Feedback, suggestions, and contributions are welcome!
github.com/brianckeegan...
Ready for another Computational Social Science Starter Pack?
Here is number 2! More amazing folks to follow! Many students and the next gen represented!
go.bsky.app/GoEyD7d
