I’m attending this, and will be arguing strongly for the values we use for generative AI use in metacheck:
1. AI should be optional and opt-in
2. AI should only be used to assist in categorising/identifying, never evaluating
3. All tools should provide transparent validation data
Anybody with a passing interest in Labour politics remembers the stench emanating from Mandelson and New Labour 15 years ago. His return to frontline UK politics was utterly ridiculous and indicative of rotten cronyism.
Three cool scientists introduce their work "Our new theory has been submitted to a peer-reviewed journal, but we feel its depths and intricacies are best appreciated on the limited-edition vinyl release."
My latest cartoon for @newscientist.com
New IfG report recommends devolving employment support. I think the case for this is very strong: if you look at a map of “where sickness related economic inactivity happens”, it is obvious that what you are seeing is primarily a problem of how local economies operate:
15 minutes; no questions; government got the blame; my department to have 60% of us sacked. Apparently sacking us will lead to a ‘new, improved’ curriculum. No voluntary scheme, straight to selection and legal minimum redundancy payment.
The future of science looks very similar to the history of science...
Those of us who live in the UK have paid Palantir more than a billion pounds to analyse our medical records and defence data
techcrunch.com/2026/04/19/p...
The problem with reforming research funding is that almost all senior researchers who could influence it have spent their entire careers conforming to the system.
There are viable other models but they would requiring adopting different values and painful self-reflection for senior researchers.
You allow people to write exceptions to the policy in cover letter.
That is true of any badly run hierarchical organisation.
If there is only a policy level effect then you just put a variable for exposure to the policy and leave age out of the model.
In that case you may need to account for multiple factors with your age variable including
a) Policy differences between <18 and >18.
b) Continuous non-linear biological effects of age.
Dichotomising would assume b is ignorable. Other methods such (e.g. spline?) would account for b.
If the author has got a very good reason to use dichotomisation in their model (or anything that breaches journal guidelines) they could state why their paper is an exception in their cover letter.
When countries have been really serious about funding defence they have made taxation more progressive. academic.oup.com/ereh/article...
When we talk about 'slashing welfare' to pay for defence, let's be very clear what this really means. It's one of these things:
1. Cutting working age social security support: 5 in 6 people on UC can't afford essentials. 75% of people coming to food banks are disabled...
1/3
It's not just an issue with Reform. BBC politics is entirely viewed through the lens of the first past the post and the winner takes all.
The idea that they should ask what would a party's priorities be in the event of negotiating a coalition in a hung parliament is alien to them.
This is the lowest figure since records began in 2013 and, for men, the first time the figure has dropped below 50.
Creating NHS health care centres into former retail areas may be a way of solving multiple problems at the same time.
www.theguardian.com/society/2026...
The UK has already had a couple of PMs who have put power before principles, and would have embraced facism if they thought it would enhance their careers. Parliament is supposed to reign them in. With Lizz Truss it was only the markets that forced parliament to act.
Are people engaging with the science on Linkedin though? An example start of a post.
"I'm delighted to share our long-awaited paper in the International Journal of Blah ....more"
Followed by a picture.
I suspect a lot of people clicking like but have no idea what the research is.
This is the biggest sign of the absolute guttering of US soft power. Whole countries are actively choosing Linux. Decades of Linux meming risks being overturned.
We're #hiring! 🔬 We have two exciting new opportunities to join our team, with a particular focus on supporting and delivering work investigating geographical inequalities in access to care, and the impact of musculoskeletal conditions on work participation.
"Emerging automated tools that run several plausible analyses of the same data — a technique called multiverse analysis "
Sounds a lot like sensitivity analysis except with a new buzz word and without critical thinking.
'instances of published resmearch have rocketed since the introduction of large language models such as ChatGPT. The problem is particularly acute in public health disciplines...as LLMs can scan vast datasets...until they find an association that the group or company wants to see'. 3/3
AI propagandists should watch more Father Ted. Mrs Doyle summed it up best 30 years go.
"Maybe I like the misery"
www.youtube.com/watch?v=dwz6...
Researchers should be a defence against populism and misinformation. Research cuts in anglosphere countries are partly a failure of research leadership. Are those leaving the anglosphere to carry on as they were going to provide appropriate leadership if they move elsewhere?
I am seeing defunding of numerous research areas across anglosphere countries. What countries are providing sufficient research funding and academic freedom to be able to employ the displaced scientists?
I have seen a lot of comments about researchers wanting to leave Anglosphere countries due to lack of funding and political interference. I see a couple of problems with this.
'It is uncomfortable and sometimes controversial to talk about a male ‘crisis’ in the face of entrenched and often worsening discrimination against girls and women. No country in the world has achieved gender equality — and one in four reported a backlash against women’s rights in 2024.'
Almost any area of research that requires causal inference from observational data is going to be produce inconsistent findings. What tends to be highlighted will vary a lot depending on the social and political context at the time of publication.