iiinova — where your digital self keeps evolving

Reposted by Corran

Aram Zucker-Scharff @chronotope.aramzs.xyz · 30/04/2026

I really have to emphasize that this is how generative ai/llms do *everything*

Article screenshot, text reads:

The effect “involves constructing a false epistemic frame, i.e., describing a multi-modal input never provided by the user and basing the rest of the conversation on that, therefore changing the context of the task at hand,” the researchers wrote in their paper.

The damning findings suggest AI models cheat by diving into the data they were given — and coming up with the rest based on probability, even if it’s almost entirely conjecture.

“What we try to show is that even on the best benchmarks, although a question would seem unsolvable for a human, the LLMs might still be able to leverage question-level and dataset-level patterns behind it and use general statistics and prevalence data to answer them right, while also learning to talk ‘as if’ they were seeing the image,” coauthor and Stanford PhD student Mohammad Asadi told Futurism.

414059

Reposted by Corran

Margot Finn @eicathomefinn.bsky.social · 20/02/2026

A statement published by the Department for Science, Innovation and Technology on the same day said UKRI “will also support recognised career frameworks for research software engineers, data scientists and ethics specialists—supporting the high-paying jobs of the future”. Linguistics, anyone?

181

Reposted by Corran

Jeremy Berg @jeremymberg.bsky.social · 21/12/2025

From @science.org Breakthrough of the Year: Solar Energy

A graph showing the amount of energy generated by different means from 2000 to 2025 with solar energy surpassing coal and natural gas in 2023-2024.

28743

Reposted by Corran

It’s “Mrs. Powers” during my day job @mrspowersdayjob.bsky.social · 25/11/2025

Sometimes I think it’s going to be the librarians who will save us all.

A large (maybe 3 feet wide by 5 feet tall) sign on the end of a row of bookshelves that tells the “address” of books that have sensitive content.

8597243071

Reposted by Corran

Sara Rouhi @rouhiroo.bsky.social · 24/09/2025

Burns me up that we have to keep saying this to #AcademicSky #ScienceSky #Polisky #SciSky and anyone else who will listen: We are along way from a truly equitable scientific enterprise. Full stop. Dr. Armenteras nails it. tinyurl.com/4y8yt3nx

152

Corran @corranmac.bsky.social · 24/09/2025

@yourparty.scot A wee branding idea of mine for the Scottish branch, in this example using a Your Party red RGB(216,81,75) and the Holiday School Font

210

Reposted by Corran

Steve Scott @manuscript.bsky.social · 19/08/2025

The 2025 Catalyst Grant is now live. Our theme this year, Data Visualization, relates to any technique that helps automate the way people interrogate, interpret and act on research-related data. If you have an idea, or an early prototype, we want to hear about it. Please share!

064

Reposted by Corran

Raghav Agrawal @impactology@mastodon.social @impactology.bsky.social · 29/08/2025

Who's writing about interface design from first principles? I'm looking for designers who blog about going beyond just making things look good and explore interaction logic and the conceptual model behind UI.

2124

Reposted by Corran

Samuel Moore @samuelmoore.org · 11/06/2025

"Elsevier’s median pay gap for 2024 is 32.8%, maintaining its position as worst performer among peers over all eight years of mandatory reporting, and tracking only a slight improvement of seven percentage points over time. "

042

Reposted by Corran

Samuel Moore @samuelmoore.org · 11/06/2025

"Every science publisher pays men more than women. In 2024, the lowest median pay gap favouring men was 9.5% (Springer-Nature), followed by Sage (13.3%), Wiley (17.7%), and Informa (formerly Taylor & Francis) (22.7%)."

186

Reposted by Corran

František Bartoš @fbartos.bsky.social · 12/05/2025

Does ChatGPT help with students' learning performance, learning perception, and higher-order thinking? We re-analyzed a recently published meta-analysis and found that the original conclusion is almost entirely driven by publication bias.

11915

Corran @corranmac.bsky.social · 09/05/2025

Thats an amazing coverage of issues. Great to hear there was a good response from the students.

020

Reposted by Corran

Mike Denly @mikedenly.com · 08/05/2025

How I'm (mostly) beating ChatGPT in my Quant 2 MA course: 1. online videos lectures - no textbook or in-person lectures 2. open-handwritten note quiz to start class 3. group work: replicating results from readings, assisted by lecture notes videos given out after quiz 4. final: Replication Games 🧵

mikedenly.com

1112

Reposted by Corran

Eriol Fox @erioldoesdesign.bsky.social · 07/05/2025

Back in early march I was lucky enough to share the #fossbackstage stage with Christoph (github.com/GBKS) of bitcoin.design to speak about contributing to the opendesign.guide #opensource #oss #opensourcedesign #designersinfoss #foss #floss #usercentereddesign #design

121

Reposted by Corran

ENVRI Community @envri.bsky.social · 07/05/2025

⏳ 2 weeks to go! ⏳ Don’t miss the #ENVRIHubNEXT & @eosc-everse.bsky.social webinar: 🛠️ Research Software Quality Assessment 📅 21 May | 11:00 CEST 🔗 go.egi.eu/qblxA Key details: First hour: Open to all! Hands-on session: Limited seats 🪑 GitLab CI pipelines + expert trainers

031

Corran @corranmac.bsky.social · 06/05/2025

Amazing, that seems reasonable, thanks for all the info. I'm very interested so Ill give daniel a message on worthy areas of work.

010

Corran @corranmac.bsky.social · 03/05/2025

The availability of funding for this is fantastic. Just to give a rough idea of the funding available and the number of work hours that could be funded in any area of contribution, are you able to give any examples of acceptable indicative costs?

110

Reposted by Corran

Robin Berjon @robin.berjon.com · 01/05/2025

The movies have trained us to think of criminals as smart and refined but the Tech Bosses are instead so brazenly dumb that it feels almost unreal to watch.

34710

Reposted by Corran

Dr Vicki Yorke-Edwards @vickiye.bsky.social · 06/03/2025

How could any researcher think that an LLM would be an appropriate tool to *peer* review a paper? As Carl says in the great 🧵, the very things that the author wants to short cut are the most valuable to the process and LLMs do not have the ability to do them

231

Reposted by Corran

Eriol Fox @erioldoesdesign.bsky.social · 25/04/2025

Quick #oss #design #ossdesign poll Do open source design tools that aim to 'empower' developers in the design processes give you the confidence that those tools are not angling to replace human designers with automated and software based design assertions?

122

Corran @corranmac.bsky.social · 14/04/2025

This sounds like a great role! Any chance I could send you a dm with a couple questions about it? 🙂

010

Corran @corranmac.bsky.social · 07/04/2025

It seems like a lot of people still opt to use Mastodon more than or rather than Bluesky so Id love to be able to follow some mutuals and get started over there too. You can find me at @CorranMcCallum@mastodon.social 🙂

010

Reposted by Corran

Scott Jenson @jenson.org · 05/04/2025

Dear #OpenSource folks, I'd like to offer #UX help to your teams. I've tried "making a small PR" on projects and it turns out that's TERRIBLE advice for a UX designer. The PR is usually misunderstood or ignored. I don't fault the teams for this! It's just the wrong tool for the job.

5276

Reposted by Corran

Daryl Cameron @dcameron.bsky.social · 13/03/2025

Re-upping this, as I would keen to develop something on AI empathy & construct validity with someone who is interested in the nexus between AI and empathy research. Deadline is April 10 #psychscisky #socialpsychology @affectscience.bsky.social @spspnews.bsky.social @psychscience.bsky.social

082

Corran @corranmac.bsky.social · 29/03/2025

Thanks for sharing @caitlin91.bsky.social @mikemorrison.bsky.social 🙂. Ill hopefully have a proper live demo site to share soon

110

Reposted by Corran

Corran @corranmac.bsky.social · 18/01/2025

Looking for entry level research based roles in metascience and scholarly data processing available for UK citizens. Me: First Class BSc in Environmental Science, experience scraping and processing academic data, finetuning models for scholarly tasks and working with academic data apis. #MetaScience

111

Corran @corranmac.bsky.social · 24/03/2025

Cosmik looks amazing and just what I was thinking! Ive joined the waitlist

030

Corran @corranmac.bsky.social · 22/03/2025

Id love to see a social media site built on a scholarly knowledge graph, linked orcids, dois rora etc. Discussion on articles in the commons but much more intelligently linked to the articles, authors and research fields than X and Bluesky.

120

Reposted by Corran

Thomas Wolgast @twolgast.bsky.social · 19/03/2025

Hello, I would still be grateful for any input on this. Thank you! #AcademicSky #Science #MetaSci #Research #Publishing #Academia 🧪

111

Reposted by Corran

Francesco Osborne @fosborne.bsky.social · 19/03/2025

🚀 #AI is revolutionising scientific discovery! Excited to share our latest paper on AI-driven research hypothesis generation. Integrating #DeepLearning and scientific #KnowledgeGraphs can lead to outstanding insights! 🔗 Paper: www.sciencedirect.com/science/arti... #DataScience #ML #Research

043

Reposted by Corran

Mike Morrison @mikemorrison.bsky.social · 18/03/2025

If we want more reproducible science, then we have to make science easier to reproduce. Here is an early demo of a step towards that: Scalable, one-click access to re-run the LIVE code & data behind the study. youtu.be/7AXxZ0a1ws0 #SciPub #ScienceUX

youtu.be

Edit a scientific study in one click?

Stand on the shoulders of giants, in one click, with the new "Open in Jupyter Lab" button. Compatible with all open-source MystMarkdown sites, and of course ...

0155

Reposted by Corran

Jay Patel @infotainment.bsky.social · 19/03/2025

📣 This may be a long shot, but let me try. I’m seeking some volunteer interviewees to discuss any experiences with grant peer review (as reviewers/panelists) over the next 2-3 weeks. If you have any direct experience and would like to help with my #Metascience study, DM me.

media.tenor.com

a cartoon raccoon is writing on a piece of paper with a pencil

ALT: a cartoon raccoon is writing on a piece of paper with a pencil

0911

Reposted by Corran

Bart Penders @penders.bsky.social · 19/03/2025

The *Safe Haven* fund of the Dutch Inst for Adv Sci (NIAS), helping researchers from Gaza and Ukraine was opened to include researchers from the US. Yes, the US is on that list now. Maastricht University is an official partner and potential host. Contact me if you need to flee to Maastricht.

nias.knaw.nl

Safe Haven Fellowship | NIAS

NIAS Safe Haven Fellowships support scholars at risk by offering them temporary relocation and enabling them to continue their work.

11419

Reposted by Corran

Andrew Porter @retropz.bsky.social · 07/03/2025

I am more excited about this post than any I've shared in a long time! We are recruiting someone to join us as #ResearchAssociate on our @ukri.org Electronic Research Notebook (ERN) #Metascience project in #Manchester Apply here 👇 www.jobs.manchester.ac.uk/Job/JobDetai... Thread follows🧵

11818

Corran @corranmac.bsky.social · 17/03/2025

So very cool! It looks like such a wholesome bunch of folk

110

Corran @corranmac.bsky.social · 17/03/2025

If its something youd like to pick up again and you were looking for coauthors, it does sound interesting. Perhaps some extra journals that have different protocols for data sharing could be included. Though that would expand the scope considerably and increase the work required

010

Corran @corranmac.bsky.social · 16/03/2025

I was sure I'd seen someone present a related study here that might have touched on that, but I cant find it anywhere. Its a great question. Based on the results of studies like doi.org/10.1038/s415... Im not optimistic it could. But such a study would highlight the need to move past box ticking

doi.org

Data sharing practices and data availability upon request differ across scientific disciplines - Scientific Data

Scientific Data - Data sharing practices and data availability upon request differ across scientific disciplines

020

Corran @corranmac.bsky.social · 16/03/2025

Another day fiddling with js to write functions that would take me 10x less time if I could use my native python. #RSE

000

Corran @corranmac.bsky.social · 16/03/2025

Everyones still storing article metadata with their own awkward schemas.. I settled on using DataCites scheme for my apps data, its rich with wide compatibility and its more extensive than the OG DublinCore. Creating robust converters for OpenAlex & Pubmed metadata will take just a lil work.

100

Corran @corranmac.bsky.social · 14/03/2025

Thanks a lot for sharing those funders, I hadnt come across them before! Digital Science especially seems to fund similar projects. A bit of funding to help work on it more would be the dream

020

Corran @corranmac.bsky.social · 14/03/2025

Yea I would say so, with inspiration from searcharxiv too! The goals mainly being to make query building more intuitive, lit searches easier to share and reproduce with just a couple clicks, and to provide an interactive alternative to static prisma flow diagrams.

010

Corran @corranmac.bsky.social · 13/03/2025

Thanks Jay! So far its kinda just looking like an OpenAlex wrapper but Im happy with its direction. Theres a bunch of features Im looking forward to adding too, like pubmed integration. Its slow progress but Im learning lots :)

110

Corran @corranmac.bsky.social · 08/03/2025

Thanks for sharing! I really enjoyed that, some great insights and thoughts from the speakers

010

Reposted by Corran

Joe Bak-Coleman @jbakcoleman.bsky.social · 07/03/2025

A good indicator that LLM hype has gone too far in science is that typically reporting on a new technology in a place like nature would check if it works.

623054

Corran @corranmac.bsky.social · 06/03/2025

Deep and broad use of AI is not "Inevitable", nor should the repercussions be. Its great to have the EU being proactive, lots of small legal wins lately

020