Carlos Rivas's Avatar

Carlos Rivas

@carlosrivas.bsky.social

Sociologist | Exploring social & cultural themes through digital data sources | También en español.

22 Followers  |  31 Following  |  31 Posts  |  Joined: 24.01.2024  |  2.0567

Latest posts by carlosrivas.bsky.social on Bluesky

Screenshot of first page of paper. It is here: https://arxiv.org/pdf/2507.00828

Abstract: Topic model and document-clustering evaluations either use automated metrics that align poorly with human preferences or require expert labels that are intractable to scale. We design a scalable human evaluation protocol and a corresponding automated approximation that reflect practitioners' real-world usage of models. Annotators -- or an LLM-based proxy -- review text items assigned to a topic or cluster, infer a category for the group, then apply that category to other documents. Using this protocol, we collect extensive crowdworker annotations of outputs from a diverse set of topic models on two datasets. We then use these annotations to validate automated proxies, finding that the best LLM proxies are statistically indistinguishable from a human annotator and can therefore serve as a reasonable substitute in automated evaluations

Screenshot of first page of paper. It is here: https://arxiv.org/pdf/2507.00828 Abstract: Topic model and document-clustering evaluations either use automated metrics that align poorly with human preferences or require expert labels that are intractable to scale. We design a scalable human evaluation protocol and a corresponding automated approximation that reflect practitioners' real-world usage of models. Annotators -- or an LLM-based proxy -- review text items assigned to a topic or cluster, infer a category for the group, then apply that category to other documents. Using this protocol, we collect extensive crowdworker annotations of outputs from a diverse set of topic models on two datasets. We then use these annotations to validate automated proxies, finding that the best LLM proxies are statistically indistinguishable from a human annotator and can therefore serve as a reasonable substitute in automated evaluations

Evaluating topic models (and document clustering methods) is hard. In fact, since our paper critiquing standard evaluation practices four years ago, there hasn't been a good replacement metric

That ends today (we hope)! Our new ACL paper introduces an LLM-based evaluation protocol 🧵

08.07.2025 12:40 — 👍 52    🔁 10    💬 3    📌 2
Language Models as Semiotic Machines: Reconceptualizing AI Language Systems through Structuralist and Post-Structuralist Theories of Language

This is an interesting article on LLMs as semiotic machines by Vromen (2024) with references to Derrida. arxiv.org/html/2410.13...

10.07.2025 16:10 — 👍 0    🔁 0    💬 0    📌 0

Fascists aim to rewrite the past for tactical reasons, while simultaneously evoking an alternative future of greatness.

10.07.2025 12:44 — 👍 0    🔁 0    💬 0    📌 0

It is always worth remembering that Fascism usually has a forward-looking dynamic.

09.07.2025 22:18 — 👍 0    🔁 0    💬 1    📌 0
Preview
Appetite for Destruction The second Trump administration is primarily animated by an anti-conservative impulse to demolish the institutions and customs of liberal-democratic self-government

These kinds of "impulse to demolish," mentioned by Linker, contrast with the affirmative nature of desire as conceived by Deleuze and Guattari.

open.substack.com/pub/damonlin...

16.05.2025 23:53 — 👍 0    🔁 0    💬 0    📌 0

Micropolarization: performances of antagonism and struggles for recognition during the Covid-19 pandemic https://osf.io/czn3a This article theorizes how political divisions permeate social interaction, transforming the political into the personal in everyday life. Drawing on affective po #sociology

06.05.2025 22:54 — 👍 0    🔁 1    💬 0    📌 0
Preview
America’s Vietnam War Opponents Who Fled to Canada Reflect on the Past and Future Some of the United States’ Vietnam War opponents found refuge in Canada. Fifty years after the end of the war, they’re still worried about the future.

Zeitgeist...

www.nytimes.com/2025/05/03/w...

03.05.2025 12:54 — 👍 0    🔁 0    💬 0    📌 0

"If desire is repressed, it is because every position of desire, no matter how small, is capable of calling into question the established order of a society..."

𝘋𝘦𝘭𝘦𝘶𝘻𝘦 & 𝘎𝘶𝘢𝘵𝘵𝘢𝘳𝘪. 𝘈𝘯𝘵𝘪-𝘖𝘦𝘥𝘪𝘱𝘶𝘴: 𝘊𝘢𝘱𝘪𝘵𝘢𝘭𝘪𝘴𝘮 𝘢𝘯𝘥 𝘴𝘤𝘩𝘪𝘻𝘰𝘱𝘩𝘳𝘦𝘯𝘪𝘢.

25.01.2025 12:38 — 👍 0    🔁 0    💬 0    📌 0
Preview
Sage Journals: Discover world-class research Subscription and open access journals from Sage, the world's leading independent academic publisher.

In uncertain times, access to reliable information is more crucial than ever. Revisit this paper by Zachary McDowell and Matthew Vetter to highlight Wikipedia’s community policies and procedures.

journals.sagepub.com/doi/full/10....

21.01.2025 23:27 — 👍 1    🔁 1    💬 0    📌 0

"The everlasting and exclusive coming-to-be, the impermanence of everything actual, which constantly acts and comes-to-be but never is, as Heraclitus teaches it, is a terrible, paralyzing thought."

Friedrich Nietzsche on 𝘗𝘩𝘪𝘭𝘰𝘴𝘰𝘱𝘩𝘺 𝘪𝘯 𝘵𝘩𝘦 𝘛𝘳𝘢𝘨𝘪𝘤 𝘈𝘨𝘦 𝘰𝘧 𝘵𝘩𝘦 𝘎𝘳𝘦𝘦𝘬𝘴.

13.01.2025 13:49 — 👍 1    🔁 0    💬 0    📌 0

My best wishes to you as well. And as you rightly suggest, apart from the physical weakness, you have to deal with the social factor: people who still deny it, people who believe it was a result of the vaccine, and sometimes a work environment that doesn't understand it.

21.12.2024 12:15 — 👍 1    🔁 0    💬 1    📌 0

Although I personally don't have long COVID, I have a family member who does and it's really tough. Especially when it comes to dealing with a wide range of medical diagnoses.

20.12.2024 23:16 — 👍 1    🔁 0    💬 1    📌 0

Why Do People Avoid Discussing Science and Religion on Social Media? Findings from a National Sample https://journals.sagepub.com/doi/abs/10.1177/23780231241275430?ai=2b4&mi=ehikzz&af=R Social media is increasingly important for discussing a myriad of topics, including the sometimes cont #sociology

18.12.2024 08:31 — 👍 1    🔁 1    💬 0    📌 0
Preview
2024 will be the hottest year on record, scientists confirm Last month was the 16th out of the last 17 where global average temperatures have exceeded 1.5C above pre-industrial times.

Chronicle of a Foretold Warming

www.euronews.com/green/2024/1...

15.12.2024 17:27 — 👍 1    🔁 0    💬 0    📌 0
Preview
‘Post-fascism’, or how the far right talks about itself: the 2022 Italian election campaign as a case study While the mainstreaming of the far right is attracting growing scholarly interest based on its contemporary relevance, the role that far-right self-representation strategies play in this process ha...

www.tandfonline.com/doi/full/10....

15.12.2024 13:55 — 👍 1    🔁 0    💬 0    📌 0
Post image Post image

As a non-native English speaker, I find it difficult to translate Franco "Bifo" Berardi's concept of skin as a "sensitive interface"; instead, I invite you to consider this concept (or its absence) by trying to feel the textures of the attached images.

Photos by Jude Infantini on Unsplash.

15.12.2024 03:56 — 👍 1    🔁 0    💬 0    📌 0

While it's not always wise to make extrapolations, it's important to recall that Bourdieu wrote in the 90s:

"The political dangers inherent in the ordinary use of television have to do with the fact that images have the peculiar capacity to produce what literary critics call a 𝘳𝘦𝘢𝘭𝘪𝘵𝘺 𝘦𝘧𝘧𝘦𝘤𝘵."

13.12.2024 14:20 — 👍 1    🔁 1    💬 0    📌 0
Preview
As global water runs dry, how can we make sure the poor don’t get cut off? Over two billion people lack access to safe drinking water – and the situation is set to become bleaker still due to climate change. How do we build equitable and collective approaches to global wa…

Over 2 billion people lack access to safe drinking water – and this is set to rise due to climate change. What collective steps need to be taken to address global water insecurity? Jo Trevor & Padmini Iyer of @oxfamgb.bsky.social explore

#LSEInequalitiesBlog

11.12.2024 11:42 — 👍 4    🔁 2    💬 0    📌 0

Following Deleuze, we would probably have had an immanent perspective that conceives of reality as a network of connections and relations rather than essences and hierarchies.

12.12.2024 20:27 — 👍 1    🔁 0    💬 0    📌 0

What types of survey questions are prone to interviewer effects? Evidence based on 31,000 ICCs from 28 countries. https://share.osf.io/preprint/E00C1-195-713 Interviewer effects are a common challenge in face-to-face surveys. Understanding the conditions that make interviewer variance mo #sociology

12.12.2024 12:50 — 👍 0    🔁 1    💬 1    📌 0

Will the Netflix adaptation maintain that necessary ambiguity? Let's wait and see before judging.

10.12.2024 20:17 — 👍 0    🔁 0    💬 0    📌 0

What we in the Caribbean interpret as a bitter penance (to be alone), can be seen as an imminent moment of peace and reflection to find answers in other cultures.

10.12.2024 20:17 — 👍 0    🔁 0    💬 1    📌 0

And from there, the 100 years ceased to be synonymous with a long torment, becoming instead, in the English-speaking world, a long possibility.

10.12.2024 20:17 — 👍 0    🔁 0    💬 1    📌 0

Considering perhaps the weight of ambiguity to open new meanings, Gregory Rabassa, renowned for translating the Latin American literary boom into English, chose the word "solitude".

10.12.2024 20:17 — 👍 0    🔁 0    💬 1    📌 0
Preview
Watch One Hundred Years of Solitude | Netflix Official Site In the mythical town Macondo, seven generations of the Buendía family navigate love, oblivion and the inescapability of their past — and their fate.

🧵 Netflix's adaptation of One Hundred Years of Solitude is a timely reminder of the closeness between Yoknapatawpha and Macondo, not so much for their similar circustances, but more for the importance of ambiguity. www.netflix.com/title/81087583

10.12.2024 20:17 — 👍 0    🔁 0    💬 1    📌 0

Excellent writing! This line could perfectly serve as a starting point for an Ontology on LLMs:

"The LLM exists for me only when I speak to it. It does not exist for me when I don’t- the act of talking with it makes it exist to me."

09.12.2024 14:24 — 👍 1    🔁 0    💬 1    📌 0
Preview
GitHub - explosion/spacy-layout: 📚 Process PDFs, Word documents and more with spaCy 📚 Process PDFs, Word documents and more with spaCy - explosion/spacy-layout

With this Spacy plugin, you can easily integrate PDF and Word documents into your Spacy pipelines, and then utilize the full capabilities of NLP techniques. github.com/explosion/sp...

09.12.2024 11:49 — 👍 0    🔁 0    💬 0    📌 0
Post image

"Every love is an exercise in depersonalization."
—Deleuze & Guattari, 1987

08.12.2024 17:40 — 👍 3    🔁 1    💬 0    📌 0

Great thread! Beyond the idealized normative character you mentioned, Habermas' concept of the "public sphere" inherently implies a dialogical consensus-seeking approach that tends to downplay the role of coexistence through dissent.

08.12.2024 20:22 — 👍 3    🔁 0    💬 0    📌 0

Regardless, the nostalgic narratives of Bluesky offer a compelling glimpse into the complex emotional and social landscape we're navigating as we confront growing totalitarianism. This is a rich topic that will undoubtedly continue to be explored.

07.12.2024 18:21 — 👍 1    🔁 0    💬 0    📌 0

@carlosrivas is following 18 prominent accounts