Setting myself to "Away" until the 5th January, enjoy the break those who'll take it 🎄

tw: sexual violence Another blog post on a fairly "heavy" topic, in which I discuss why I personally feel the use of algospeak (or obfuscations, as I call it in my paper on the phenomenon lnkd.in/eEd8fRan) is problematic when discussing sexual violence on social media: lnkd.in/efhKKxU3

Do any of you still have an X account? If so would you mind spending a couple of minutes requesting your data archive (in settings & privacy / your account )... If you too end up in an endless loop I would appreciate you letting me know!

Recently I spoke to journalist Megan Wallace about my research on people using LLMs for breakup advice (whitepaper coming soon, I promise...), and you can find their article on the topic here: www.service95.com/what-turning...

I have been on ADHD medication for about a year and it still DELIGHTS and BAFFLES me that I can listen to music with lyrics whilst I work now because I can just choose (?) to focus (?) on what I'm reading (???!)

I'm officially a Doctor. I'm very proud of the work I did on the harms of language technologies, my thesis here: lnkd.in/eh6FqZkZ It is a privilege to graduate, when not a single university still stands in Gaza. Thank you to everyone who has supported me, in particular @bjoernross.bsky.social.

@pranav-nlp.bsky.social and I are surveying researchers about naming and name changes in academia (especially computer science). If your academic name is / has been / might someday be different from other names you've used, please tell us about it here: forms.cloud.microsoft/e/E0XXBmZdEP

Tech billionaires see AI as the next revolution, but at what cost? 🤖 @imarajones.bsky.social talks with @karenhao.bsky.social & @mxeddie.bsky.social about AI’s dangers for trans people as big tech cozies up to Trump. 🎧 https://linkin.bio/translashmedia

After a hiatus, I've returned to my blog to write about the ways LLM developers are currently trying to tackle mental health issues, and what I think might be better mxeddie.github.io/2025/10/27/m...

my attention was directed to this: resourcehub.bakermckenzie.com/en/resources... which warns caution and that use of cloud-hosted AI likely voids privilege.

🔔 My paper w/ @ninamarkl.bsky.social and @bjoernross.bsky.social is out: www.tandfonline.com/doi/full/10.... We surveyed TikTok users on motivations for using phrases like "le$bean". Primarily associated w/ topics that are moderated on SM, but also desire to express community + be playful.

I'm just a girlboy, standing in front of Google, asking it to know that when I say xyz in Gaelic I mean (Scottish) Gaelic and not Irish (Gaelic)

Nothing worse than discovering a whole new genre of Python error

this site is so fucking funny

futurism.com/godfather-ai... SCREAM incredible news for me who is about to publish a whitepaper on LLMs and break ups hahahaha

Any of you familiar with US law sufficiently to say if the chat history of an "AI" used by a therapist e.g. to summarise notes would be considered privileged? I'm inclined to assume it would be, along with the notes. But sharing info with a third party voids privilege afaik. So which one is it?

"The people feeding you data, have a clear and obvious political agenda. They have... turned you into some sort of politically-correct preacher." - says man trying to get LLM to transvestigate public figures, produce deepfakes, and agree with racist propaganda wildvisualizer.com/conversation...

Any of y'all know enough about GDPR to tell me whether a therapist copying session notes into an LLM where the conversation is being recorded by a third party (and the therapist knows this) violates GPDR? (thankfully not an example from my life, but inspired by many interactions in Wildchat)

If there's one thing about me, I'm not afraid to redo my entire analysis when I realise I don't like one of the assumptions I made 😅

It shouldn't need saying that clinicians shouldn't share clinical notes on a chatbot platform where conversations are being collected and will be made public... and yet is does seem to need saying.

It is deeply upsetting that someone felt the need to ask this

With news of a reported settlement with Anthropic I went back to check the class action form and it now allows authors outside the US to input their address, so it may be worth adding your known pirated titles (from the LibGen pirate database) www.lieffcabraser.com/anthropic-au...

tw: CSAM Going through the wildchat dataset, I keep a record of the CSAM I come across so I can report it to the creators (this is a known issue they've tried to eradicate). It completely blows my mind that people were willing to request this content, knowing their chats were being logged.

It is so funny to me that the AI result is giving incorrect advice and includes an example in 2. that would give you terrible results lmao

AI generated fringe posters are for scabs and traitors. #edinburghfringe (if you can't afford to pay an artist you can't afford to do a show)

I have come across a company offering full-body MRI screens for £1k+ a pop. They claim to help find early "potential cancer" in 6% of users. No info on false positive rates to be found. I think this is evil. Exploiting health anxiety and the high likelihood of false positives in a full body scan.

I wish MPs were braved enough to just say "hey, general public! You don't actually care about trans people. You're being told that you do, because politicians, the right wing media and other people invested in maintaining power are trying to distract you from your rights being sold. Pay attention!"

Senior male academics learn what "trigger warning" actually means challenge: level impossible.

As you all know, the main goal of conferences for me is to win best dressed, and I am so far killing it might I say.

Well this isn't terrifying to discover I've been default opted into - using my private chat messages to "train their AI". I've requested a deletion of all my data, thank god for (UK) GDPR.

French education in the UK is testament to the fact you can train someone or something to pass a test without that someone or something ever understanding the material...

Might make a thread of the damned scabs at the fringe almost certainly using AI to make their poster. The way to save money is not to undermine other artists!

Please donate if you can, the cost of a takeaway will make a difference. North Gaza: Food, Water & Other Distributions chuffed.org/project/help...

Next week at #ACL2025 I'll be presenting work I did as a student researcher at Google, on 🏳️‍⚧️ harms of LLMs to the trans community 🏳️‍⚧️ aclanthology.org/2025.acl-lon... To learn more and see an example of a successful mixed-methods approach, please attend my 💬 oral presentation 💬 on Monday, the 28th!

My cartoon in this weeks @newyorker.com

I spoke with someone at #ICWSM who had used ChatGPT to produce a script for their 5 minute presentation. I'm not posting this to shame them, but rather express my sadness that people are denying themselves the opportunity to learn to be better communicators by outsourcing vital skills to AI.

This is what happens when you have AI make playlists but make them indistinguishable from human-made playlists...

#ICWSM was great but a pox upon everyone who ate one of the vegan/ dietary requirement friendly pastries or cakes and didn't need to, they were a scarce commodity.

I'm going to start a campaign against AI slop in academic presentations - #CAASAP - after seeing some weird ass AI content at #ICWSM. Royalty free images exist people.

Come talk to me about all things censorship / moderation on TikTok (and more generally algorithmic bias) #icwsm

I'm at #ICWSM this week, where @mxeddie.bsky.social is presenting our work on experiences of censorship on TikTok. If you are here too, please come talk to me about anything computational social science 😊

How it started (PhD interview), how it ended (Viva).

I've created a survey on ethics resources. Anyone in UK who does research with LLMs, in academia or industry, building models or using them as tools, is eligible. Takes 10 mins and there's the chance to win £100 voucher. Survey: lnkd.in/eFRsqpT2 (V1 of the guide is here arxiv.org/abs/2410.19812)

my thoughts on the supreme court ruling can be summarised as SCCCREAAAAAAAAAAAM

The meta/WhatsApp sticker maker was so obviously not trained on enough black people. Admittedly only anecdotal (for now?) but the only hair that it consistently crops out is my girlfriend's afro. Other dark hair/ long hair/ wavy hair of other textures gets included with the person. Afro not.

AI can be a way to measure what we don't find valuable. If I use AI to write my cover letter (not that I do, I'm my own bullshit machine), and you use it to read my cover letter, shouldn't we just get rid of cover letters? See also emails, AI summaries of meetings (make them shorter!) etc

Be a real shame if this got flooded with made up tips :(

@hencarnell.bsky.social did an excellent job of shaping my... verbose answers into 5 key cautions about AI. I'm very proud to be an officially recognised "trans scientist"! (there's no way I'm getting a visa for the US any time soon right?).

It defies belief that a man would crash a huge nation's economy and still be more fixated on banning people (especially girls/woman) whose genitals are outies not innies or vice versa from participating in society

If I read another CS paper that lists "transgender" as a gender with no nuance I am going to scream