Blog post: GRIOS at Metascience 2025: building bridges between evidence and policy
Our reflections from #Metascience25 are now up in our blog, incl. what we learned about #OpenScience from national frameworks, the Metascience Alliance, an IPCC-like proposal and what comes next for evidence-based Open Science policy.
Read more ๐ www.grios.org/grios-at-met...
#Metascience #Research
Metascience 2025 recordings are now available! #Metascience25 gathered researchers, funders, policymakers & more to explore how the research process can improve. The program included preconference symposia & plenary sessions on topics from policy change to AI in research.
๐ฅ cassyni.com/s/metasc...
What if together we created an IPCC for Open Science?
The next generation of #OpenScience policies needs more evidence and @grios-openscience.bsky.social is dedicated to supporting this mission, said @marind.bsky.social, @ouvrirlascience.bsky.social, at #Metascience25.
๐ Learn more: www.grios.org
I knew next to nothing about Ioannidis before #metascience25 but his talk was a classic example of how people argue against the "unrigorous" nature of science while being guilty of that same of lack of rigour. I mean, he had literally no citations for any of his numbers or definitions.
@cihrrc.bsky.social Alison Bourgon on "Inclusive Research Excellence & Opps for #metascience25" - Canada does not have a national research assessment system. Work collaboratively with govt funders across provinces and territories. Speaking DIRECTLY to their #TruthReconciliation #SocialJustice work.
Our Co-Chair Kelly Cobey presents our new Practical Guide for Research Performing Organizations in the session "The many dimensions of rethinking research assessment" at #Metascience25 - have you already explored it? You can download it at bit.ly/DORA-Guide
Pragmatism in #openscience is important; #GlobalSouth requires publishing in journals; #preprints and other #OS behaviors won't get compensation/recognition. @metaror.bsky.social partners work to ensure curation process yields objs research gets recognized.
#metascience2025 #metascience25
Itโs the final day of #MetaScience25 at UCL, London. ๐ง
At #ScienceOpen, we build publishing solutions grounded in metadata best practices to support FAIR and open research. ๐
Explore how we do it ๐ about.scienceopen.com/publishing-s...
#MetaScience2025
#OpenScience #MetaScience #FAIRdata
RoRI is delighted to be part of the founding group of organisations joining forces to build the Metascience Alliance! @cos.io #Metascience2025 #Metascience25
While I understand that the host addressed the tonality of the comment and question and wanted to open the door for panelists not answering it, I can't see what is wrong in asking a #Google representative about the firm's business model. #metascience25
When moderators shut down discussion -- even if it's ๐ถ๏ธ -- they should be conscious of their positionally v. ?-er. When a posh, white British member of the establishment shuts down a POC -- it's a bad look. You care about equity? Walk the walk. Very disappointed #metascience2025 #metascience25
What questions are "AI-able?"
1. Do you have data? Richness? Unbiased?
2. What is the problem you're trying to solve?
3. Why can't another method solve it?
Anna Koivuniemi, Google #DeepMind
#metascience2025 #metascience25
Genuine question: Why are we using #s of Nobel Prizes as a metric of "good science?" And calling it "objective?"
I find this very strange.
#metascience2025 #metascience25
@sabinaleonelli.bsky.social about the lure of convenience of AI, the risk of AI for AI's sake, and the underestimated cost of validation and maintenance. #metascience25
Challenge 3: Underestimating validation and maintenance - very expensive. "Ethics is part of science at EVERY step. Science is a system of decision making. We have to build responsible decision-making AI tools for science." #EnvironmentalIntelligence
#metascience2025 #metascience25
Challenge 2: AI for AI's sake - higher resource bias amplified - requires expensive hardware and expertise; "good for AI but not necessarily for science." Getting AI itself better doesn't help science. Secrecy around developing these models and poor incentives = risky #metascience2025 #metascience25
Dangers - increased opacity of models, reification of assumptions into dogmas (built in bias), loss of key skills
#metascience2025 #metascience25
Can AI free us up to think of larger/more innovative problems?Can it help reclaim #curiosityDrivenScience #science
#MoumitaKoley #metascience2025 #metascience25
For non-native ๐ด๓ ง๓ ข๓ ฅ๓ ฎ๓ ง๓ ฟ speakers, real concern about #equity #openaccess - See #AI to help get published; worried about further widening accessibility gap (These are NOT affordable tools now); reinforce article economy based on volume; #researchintegrity deterioration
#metascience2025 #metascience25
It's still in early days re: findings, #MoumitaKoley. ML to extract patterns from big data sets. Best oppose are in drug design/new material synthesis; Tools are exclusionary; dominated by LLMs - used most commonly for writing
#metascience2025 #metascience25
Questions #MoumitaKoley asked in her #AI study: when do you use AI/ML in research? Opportunities? Challenges to using it? Do you see #AI #machineLearning accelerating discovery? TLDR: Using AI/ML in communicating findings.
#metascience25 #metascience2025
#MoumitaKoley up next! AI/ML in India - will it be #equitable? Is #globalsouth sufficiently integrated in this work?
#metascience2025 #metascience25
Can #AI unblock bottlenecks in research? How can we better understand how it can affect #science? What would be you #HilbertProblem for the 21st? Real opportunity to accelerate the benefits for humanity by using AI responsibly."
#metascience2025 #metascience25
"There is a risk that #AI will make #science less creative. Could it reduce understanding? There absolutely are risks." Anna Koivuniemi (Head, Google #DeepMind Impact Accelerator)
#metascience2025 #metascience25
The AV issues at #metascience2025 have been a goddamn delight. The lights turning into disco balls. The mics screeching. Presenters just winging it bc slides won't show. Images/emojis just disappearing. Everyone handling with grace and patience. #metascience25 #openscience
Expensive/time-consuming to evaluate #reproducibility - can #AI help? Aiding researchers in reproducing the study or could AI do it entirely itself? Generally the AI-led teams are NOT reliable in reproducing studies. Humans and assisted were better.
#openscience #metascience25 #metascience2025
Very interesting to see researchers engage w/ each other on #metascience research -- seeing the pushback, questions, and engagement live -- including putting ppl on the spot! -- is very interesting. #ScienceInAction
"I don't know, to be honest." Very refreshing.
#metascience2025 #metascience25
What leads to irreproducible research? Did we have access to code? Partially shared?
Even if the code was missing or had problems, it was ok. If the data was wrong/missing, we COULD NOT reproduce the findings.
#opendata #openscience #opencode #metascience2025 #metascience25
nomadit.co.uk/conference/m... team defines #reproducibility as seen in the slide: Reproducibility = indep ๐ฉโ๐ฌdo draw same conclusions from am experiment by following documentation shared by investigators on the original paper
#replicationcrisis #academicsky #sciencesky #metascience25 #metascience2025
"We tried to replicate data that shared data or code. If they shared both, we were able reproduce 86% of the work. If you just share code alone, it's not enough. Much harder to repro without #opendata #computerscience"
Details here: nomadit.co.uk/conference/m...
#metascience25 #metascience2025
