Mary Loftus

Lecturer, PhD/Researcher #LearningAnalytics #MachineLearning, Photographer, Mother. Did I say learner?
Former MSc Facilitator (NUIG), IT head (GSK), CoderDojo

[Mary Loftus @marloft]

Looking forward to reading work from co-authors @14prinsp @JulianaR71 @bonstewart @HayesAneta @Michael_Kwet @ConnectEdProf @Neil_Selwyn @dgasevic @BenPatrickWill & continuing the conversation...

My paper with @michaelgmadden is here: https://t.co/nxG5xKT849
Contact me for a copy

[François Chollet @fchollet]

Buy the book from Amazon: https://t.co/6g4BtE7vf6

Buy the book from the publisher: https://t.co/BIIbAZ2iBT

Read the free website: https://t.co/3CbcQ7hmjp

The new edition has a ton more content around generative AI. It also covers JAX & PyTorch fundamentals, and everything new in Keras 3.

[Mushtaq Bilal, PhD @MushtaqBilalPhD]

Zotero is the best citation management software. It's free and doesn't monetize your data.

But many folks have a hard time starting out on it.

Here's a step-by-step tutorial to get you started on Zotero:

[Rohan Paul @rohanpaul_ai]

This paper warns that using large language models for labeling text can often lead to wrong research conclusions.

Here, LLM hacking means the final statistical claim flips depending on model, prompts, or settings, not the underlying data.

The problem comes from what the authors call LLM hacking, where results flip depending on which model, prompt, or setup is chosen, not on the actual data.

They tested 37 real research tasks with 18 different models and found that incorrect results happened in about 31% to 50% of cases.

These errors include missing real effects, inventing effects that are not there, reporting the wrong direction, or exaggerating the size of an effect.

The risk is especially high when results are near the usual significance cutoff, which is where many social science studies operate.

They also show that 100 human labels can be more reliable than 100K LLM labels, especially for avoiding false discoveries.

Correction methods that adjust results after the fact do not really solve the issue, since they reduce one type of error but increase another.

Finally, they show it is very easy for someone to deliberately game results by trying different models and prompts until they get the answer they want.

----

Paper – arxiv. org/abs/2509.08825

Paper Title: "LLM Hacking: Quantifying the Hidden Risks of Using LLMs for Text Annotation"

[Steven Strogatz @stevenstrogatz]

https://t.co/ReWae73oXW

[Sligo PPN @SligoPPN]

Everybody involved is so proud of this amazing book & the outcome is testament to the schools involved.

For many, the @SDGoals are still unfamiliar. This book brings them to life for kids & families to all see their part in helping Ireland achieve it's goals
@SocialJusticeI

[STEM Passport for Inclusion @STEMPassport]

🚀 Calling all STEM professionals! 🚀 Make a real impact by mentoring a student from a disadvantaged background. Just 15 hours (May-Dec) can change a life!

✅ 6 hrs training ✅ 9 hrs mentoring & support

Sign up today! https://t.co/GAhJK2Q92u

#STEMMentors #MentorshipMatters

[jack @jack]

https://t.co/2GijDhlTNG

[SXSW @sxsw]

Privacy isn’t about hiding—it’s about trust and control over our own information. In her #SXSW 2025 Keynote, @signalapp President, @mer__edith breaks down why our privacy matters.

[lemn sissay OBE FRSL @lemnsissay]

Ireland, I know this is on RTÉ now. I would love to read your thoughts in the comments. I can't watch. It was intriguing and liberating to be interviewed by Tommy Tiernan cause seconds before I stepped out he didn't know who he was going to be speaking to. So... how did it go?

[Mariana Mazzucato @MazzucatoM]

AI technology benefits from decades of public investment. The challenge isn’t about regulation - we must ensure AI serves public purpose rather than becoming another machine of rent extraction for ‘big tech’.

My latest for @ProSyn with @TomValletti ➡️ https://t.co/qr2Uqf8YWK

[Kate Crawford @katecrawford]

The AI Summit ends in rupture. AI accelerationists want pure expansion—more capital, energy, private infrastructure, no guard rails. Public interest camp supports labor, sustainability, shared data. safety, and oversight. The gap never looked wider. AI is in its empire era.

[Grady Booch @Grady_Booch]

The story of computing is written in its artifacts, in the organizations in which they were forged, and expecially in the people who shaped them.

To inform my understanding of the stories of computing, I have studied several thousand books, here organized according the places where computing and the human experience intersect.

[Mary Loftus @marloft]

Thriving at @atu_ie - a warm welcome from @OFlynnATU to a conversation about growing compassion, respect, inclusion and belonging at our University... looking forward to this!

[Tony Lawlor @AnthonyLawlor]

Time to share these words by MaryOliver:
"Whoever you are, no matter how lonely,
the world offers itself to your imagination,
calls to you like the wild geese, harsh and exciting–
over and over announcing your place
in the family of things."
Be well, friends, on this solstice day

[Mary Loftus @marloft]

@simonmlewis @simonmontefiore @Channel4 Looks like a beautiful piece of film, Simon!

[Mary Loftus @marloft]

@EithneHarley So hard, Eithne 💔

[Nick Tiller, Ph.D. @NBTiller]

Two papers.

On the left, a non-randomized, non-crossover study showing that a one-off sauna exposure significantly decreased arterial stiffness, with implications on cardiovascular health.

On the right, a randomized, crossover (more robust) study, showing no benefits on arterial stiffness or blood flow. Note the colossal difference in Altmetric score and media attention.

The media favors sensational but flawed studies over rigorous ones, distorting public understanding. #science #media

[Mary Loftus @marloft]

@CameronKeighron @Europarl_EN @Diabetes_ie @Thriveabetes @uniofgalway Go Cameron!

[Mary Loftus @marloft]

@sumingkhoo https://t.co/E6cZlJQ0rN

[Mary Loftus @marloft]

@sumingkhoo Also, have you come across Bill Barres?

https://t.co/gD5GfvEZ0O?