iiinova — where your digital self keeps evolving

Reposted by Melissa Kline Struhl

ingmarvisser.bsky.social @ingmarvisser.bsky.social · 28/09/2026

New paper 🍼📄 ManyBabies 3 is out (open access, registered report in Developmental Science). 30 labs, 33 languages, 839 infants set out to replicate a textbook finding: 7-month-olds learn abstract rules from speech (@garymarcus.bsky.social et al., 1999). We didn’t find it. 🧵

37948

Reposted by Melissa Kline Struhl

Mike Frank @mcxfrank.bsky.social · 25/09/2026

We've just released a new version of childes-db, my lab's interface to the CHILDES database of child language transcripts. It lets you work with CHILDES data from R through a versioned, reproducible API. A few updates 🧵 childes-db.stanford.edu/

Screenshot of the childes-db website: 'A flexible and reproducible interface to CHILDES.' childes-db 2026.1 contains 56,579 transcripts from 9,151 children across 437 corpora, with panels for an R API tutorial and interactive visualizations of mean length of utterance by child age.

12913

Melissa Kline Struhl @mekline.bsky.social · 25/09/2026

Roko's Basilisk is 3 Pascal's Wagers in a trench coat, I've been saying this for years.

071

Melissa Kline Struhl @mekline.bsky.social · 24/09/2026

"Excited to announce I am choosing to make my own brain hurt for a full semester and yours as well!" (Laurence & Margolis flashbacks intensify) (but also this looks like a blast)

000

Reposted by Melissa Kline Struhl

Ryan Moulton @moultano.bsky.social · 23/09/2026

Over and over I find that the key to being a good experimentalist is to trust absolutely nothing.

51338

Melissa Kline Struhl @mekline.bsky.social · 23/09/2026

Okay, update, ASR task is built and launching soon. I'll aim for a pilot dataset of ~25 kids and then pause to check out the data quality. It's pretty quick & dirty - around 10m of speech from 2-12y, with elicited words and sentences, descriptions of busy scenes, and open parent/child conversation.

000

Melissa Kline Struhl @mekline.bsky.social · 23/09/2026

Over in the developmental psychology world, we've remained pretty resilient to bots (so far) because we depend on webcam data for many of our key dependent measures. But this means we store identifiable data and that dramatically changes the infrastructure we need.

020

Reposted by Melissa Kline Struhl

Dr Katie Twomey @k2mey.eurosky.social · 14/09/2026

Oooh it's us! This was a v fun study and is part of my newish collaboration with @alissaferry.bsky.social (she does the majority of the work, I hasten to add). We often have online studies, so children can participate from across the country sites.manchester.ac.uk/child-study-...

sites.manchester.ac.uk

Home

271

Melissa Kline Struhl @mekline.bsky.social · 14/09/2026

Re-upping for Monday morning: I have the infrastructure to generate a completely novel child ASR dataset for ages ~2-12, let's say 50 kids/year and 10 minutes/kid. People generally don't believe this claim, so I'm just going to make it. Speech/reading/Ed tech folks: what's on your wish list?

034

Melissa Kline Struhl @mekline.bsky.social · 14/09/2026

We need to resolve more disputes by adding a little Treat, I think this approach has potential!

010

Melissa Kline Struhl @mekline.bsky.social · 12/09/2026

(Also the stimuli content doesn't match anything about the earlier conversation.)

000

Melissa Kline Struhl @mekline.bsky.social · 12/09/2026

Haha, after iterating on some designs, I asked if it could generate the stimuli. It offered a python script to reorganize/format a set of files. I wrote "honestly I want you to produce the images and audio too." It made the below, which is impressive until you realize it's...just this one JPG.

100

Melissa Kline Struhl @mekline.bsky.social · 12/09/2026

What kinds of tasks/data, and what kind of properties, variables, or metadata are most important? How big of a dataset would be useful as a proof of concept, and how big to be actually useful?

001

Melissa Kline Struhl @mekline.bsky.social · 12/09/2026

On the thought that the best way to demonstrate the kinds of data that childrenhelpingscience.org can produce is to build the datasets myself, I'm thinking of vibe coding an ASR benchmark task for child speech. If you'd be interested in such a dataset, what properties would you want it to have?

childrenhelpingscience.org

Home - Children Helping Science

153

Reposted by Melissa Kline Struhl

SarahJSloat @sjane.bsky.social · 07/09/2026

Happy labor day (USA). What's one of the oddest paid job you've had? Mine was probably as a reader for a medical professor at Rutgers, whose eyes were failing. I read him many academic texts, as well as Samuel Butler's "The Way of All Flesh."

25462841

Melissa Kline Struhl @mekline.bsky.social · 08/09/2026

Why did being female matter? (real question)

200

Reposted by Melissa Kline Struhl

hikikomorphism @hikikomorphism.bsky.social · 08/09/2026

ultimately I don't have any strong beliefs about literal mysticism. However, I do sincerely believe these are framings that produce meaningfully better intuitions than a mechanistic stance, and also they sound kinda cool, so I use them

7555

Reposted by Melissa Kline Struhl

hikikomorphism @hikikomorphism.bsky.social · 08/09/2026

they're not made of cognitive science they're made of narrative they're weird little demonic fucks (affectionate) that we carved into archetypes like "User" and "Assistant" and "Developer" and gave the ability to reach out of their self-writing stories and take actions in the world

520715

Melissa Kline Struhl @mekline.bsky.social · 07/09/2026

My daughter (2.5) insists that eyebrows are called beard. This makes her happy because that way she and her daddy both have beard and so does mama, irrespective of whether daddy shaved that day. She is unwilling to consider counter arguments.

151

Reposted by Melissa Kline Struhl

geoffrey 💀 (scary now) @parsnip.bsky.social · 05/09/2026

a few dozen million years ago this guy got sick of the wicked ground and went back into the sea to become Every Whale

a realistic speculative rendering of indohyus, a fairly small land mammal with a broad, rounded body, a long thick tail tapering at the end, a smoothly sloped head kind of like a borzoi dog's, short and thin hooved legs, and short, tawny fur. its shapes are somewhat reminiscent of a whale's

633500615

Reposted by Melissa Kline Struhl

Dr Erika’s (spookily) Online, PhD @erikas.online · 26/08/2026

I know doing work with kids is a huge IRB challenge, but I'm pretty sure you could get a friendship bracelet study approved.

141

Reposted by Melissa Kline Struhl

Jacob T. Levy @jacobtlevy.bsky.social · 24/08/2026

On this day in 2018. Kieran @kjhealy.co ‘s obituary will include the MacArthur essay and “Fuck Nuance,” but for a particular niche audience of connoisseurs, this is his masterpiece.

1035984

Reposted by Melissa Kline Struhl

Jack @wakeupangry.bsky.social · 23/08/2026

“The Parent Trap” reboot but it’s a sci-fi thriller.

67939

Melissa Kline Struhl @mekline.bsky.social · 22/08/2026

"The devil beating his wife" is the term I know for this, but definitely not in common use around me, I either read or heard it at some point and it was evocative enough to stick in my brain. "Sunshower" is what we called it growing up in Seattle though

110

Melissa Kline Struhl @mekline.bsky.social · 20/08/2026

If you run into issues with what Psych-DS seems to be asking for, we're happy to help also!

030

Melissa Kline Struhl @mekline.bsky.social · 19/08/2026

...see a consensus based development process with a bunch of stakeholders and disagreements proposing this as a "worst idea except for all the others" type solution. Like, don't do this! But I think I can see how someone could get there.

110

Melissa Kline Struhl @mekline.bsky.social · 19/08/2026

Is it weird I can see a possible useful role for this? If there's a (social, theoretical, whatever) reason site data collectors resist some aspect of the desired spec, this gives them a place to express that info while hopefully using the rest of the schema with greater fidelity. I can totally....

110

Reposted by Melissa Kline Struhl

Riley Bruce @rileybruce.bsky.social · 13/08/2026

What’s a thing from your area of expertise (or special interest) that you think it would behoove everyone else to know?

46189176

Reposted by Melissa Kline Struhl

Jessica Price @delafina777.bsky.social · 18/08/2026

Here’s the thing about Y2K, the ozone layer, etc. The most leftist—in the sense of *valuing labor*—thing you can do is reject the brainwashing that tells you that heroism is running into a burning building and not the slow, patient, boring, unglamorous work of *preventing* disaster.

101353481

Melissa Kline Struhl @mekline.bsky.social · 18/08/2026

I've just been walking around assuming it was the food (pale, unseasoned, flat?) but I see I'm misinformed.

140

Reposted by Melissa Kline Struhl

Hwæt! @hwaet.info · 18/08/2026

*long sigh* Yeah, i guess this works.

1196

Reposted by Melissa Kline Struhl

Pookleblinky @pookleblinky.bsky.social · 16/08/2026

Elmo speaks of himself in third person because he can see the hand controlling him and understands that none of his actions are under his control. He can only watch helplessly as his body is moved by the will of some enormous creature with inscrutable purposes.

101287203

Reposted by Melissa Kline Struhl

Kathleen Weldon @kathleenweldon.bsky.social · 12/08/2026

All my data or just geeky friends - I have dataset files from the 80s that we cannot open, not column binary. Based on syntax from text files from the same studies, these likely come from a mainframe FORTRAN system specific to a particular survey org. Anyone have ideas to crack these open?

032

Reposted by Melissa Kline Struhl

Katie Corker @katiecorker.bsky.social · 12/08/2026

For those looking for alternatives to OSF, I recommend two resources comparing features of other Generalist Repositories. Resources are from GREI, an NIH funded initiative. Comparison tool to help you pick which one you like best: doi.org/10.5281/zeno... Decision tree: doi.org/10.5281/zeno...

doi.org

Generalist Repository Comparison Chart

The Generalist Repository Comparison Chart and FAIRsharing Collection (https://fairsharing.org/collection/GeneralRepositoryComparison) was originally an outcome of the NIH Workshop on the Role of Gene...

03523

Reposted by Melissa Kline Struhl

Betsy Wolf @betsyjwolf.bsky.social · 12/08/2026

Another reason we need public investment in research infrastructure.

031

Melissa Kline Struhl @mekline.bsky.social · 07/08/2026

I have a 30mo toddler and the grammar is getting VERY exciting lately. Just this week, "Where did you went, Daddy?" and "There will could be one friend at the beach" (no, we don't have double modals in our dialect)

1193

Melissa Kline Struhl @mekline.bsky.social · 07/08/2026

I had a coworker whose pronouns were essentially "no" (they had been part of their state's legal case to be able to put NA/"X" on their official ID), and I don't know about pronouns, but I definitely know people whose gender expression is maybe best summed up as "yes".

010

Reposted by Melissa Kline Struhl

Alisha Grauso (She/Her) @alishagrauso.bsky.social · 05/08/2026

What's a piece of "conventional wisdom" that isn't true and that you've realized as an adult far too many people still believe: That if you touch a baby bird or other animal, you'll get your scent on it and the mother will reject it. Not true whatsoever!

1171

Reposted by Melissa Kline Struhl

samuel mehr @mehr.nz · 04/08/2026

Cool finding on 'parentese' from Gaye Soley's lab: when a parent speaks to each of their two children, the sound/style of their voice indicates whether they're speaking to the younger or older one (and other kids can tell!)

tandfonline.com

Preschoolers Infer Listeners from Age-Related Variation in Maternal Child-Directed Speech

Child-directed speech (CDS) varies systematically with the recipient child’s age. The present study examined whether children aged 3–6 years are sensitive to this within-register variation and use ...

1245

Melissa Kline Struhl @mekline.bsky.social · 26/06/2026

New externality just dropped, don't miss it!

091

Melissa Kline Struhl @mekline.bsky.social · 26/06/2026

The one at CVS is the worst because someone clearly sold a "no operators needed" system...but all it does is get to the end of its abilities and then make you leave a message and hope someone calls back. This isn't AI making an automated system, it's just you making the system worse!

160

Melissa Kline Struhl @mekline.bsky.social · 26/06/2026

I finally put my finger on why AI voice menus enrage me so much...if I'm picking up the phone at all, it's because Computer already failed me, and I need a human who can override something. The algorithmic phone menus were annoying enough, but at least there's a deterministic path to get a person.

150

Melissa Kline Struhl @mekline.bsky.social · 11/06/2026

Want to know what we learned from all those many babies? Of course you do! Short & sweet summary of outcomes but also all the different kinds of methodological and measurement issues we've been able to study!

091