iiinova — where your digital self keeps evolving

childesr: Accessing the 'CHILDES' Database

There's also a revised childesr package on CRAN: cran.r-project.org/package=chi... The data are now hosted and versioned on Redivis, so you'll need a free account there to download data through childesr.

cran.r-project.org

Tools for connecting to 'CHILDES', an open repository for transcripts of parent-child interaction. For more information on the underlying data, see <https://langcog.github.io/childes-db-website/>.

100

That's 24.2 million utterances (+24%) and 89 million tokens, including ~73 corpora new to childes-db. We've also added new annotation tiers: Universal Dependencies morphology and dependency parses, Japanese romanization, speech acts, and phonology.

120

The website is faster and improved for browsing the corpora, and it reflects the new 2026.1 data release, which is up to date with the latest changes in CHILDES: 56,579 transcripts from 9,151 children across 437 corpora.

121

Open Encyclopedia of Cognitive Science bot @oecs-bot.bsky.social · 18/09/2026

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

Reposted by Mike Frank

Philosophers treat skepticism not as stubborn denial, but as a logical trap revealing fundamental tensions in our concepts of knowledge and evidence. Skepticism by Santiago Echeverri doi.org/10.21428/e2759450.95725785 #CognitiveScience

Open Encyclopedia of Cognitive Science: "Skepticism" by Santiago Echeverri. In everyday life, people think of skepticism as the position of a stubborn person who rejects what other people believe in. Some skeptics may deny that climate change is real, while others claim that

053

Reposted by Mike Frank

Open Encyclopedia of Cognitive Science bot @oecs-bot.bsky.social · 11/09/2026

Distinguishing between our species and other animals requires balancing subtle quantitative differences with the unique, complex interplay of our… Uniquely Human Cognition by Daniel B. M. Haun doi.org/10.21428/e2759450.2760bd31 #CognitiveScience

Open Encyclopedia of Cognitive Science: "Uniquely Human Cognition" by Daniel B. M. Haun. Uniquely human cognition refers to aspects of cognition that are specific to our evolutionary lineage and distinguish us from other animals, including our closest living relatives, the nonhuman primat

086

Reposted by Mike Frank

Open Encyclopedia of Cognitive Science bot @oecs-bot.bsky.social · 04/09/2026

Brain research reveals that our neural systems treat human interaction as a distinct computational process, separate from general reasoning. Social Cognitive Neuroscience by Rebecca Saxe doi.org/10.21428/e2759450.0276fa5e #CognitiveScience

Open Encyclopedia of Cognitive Science: "Social Cognitive Neuroscience" by Rebecca Saxe. People spend a lot of their brain energy thinking about people. Social cognitive neuroscience is the field of research that investigates how this works. Research in this field asks questions about the

172

Reposted by Mike Frank

Matt Makel @mattmakel.bsky.social · 31/08/2026

I made a mistake. And because of that mistake, I no longer have confidence that results that I reported in several research papers are an accurate reflection of replication rates. 1/

19430

Mike Frank @mcxfrank.bsky.social · 30/08/2026

*Sharing for our department’s trainees* 🧠 Looking for insight on applying to PhD programs in psychology? ✨ Apply by Sep 21st to Stanford Psychology's 10th annual Paths to a Psychology PhD info session/workshop to have all of your questions answered! 📝 Application: forms.gle/4nujgDd3W2tx...

forms.gle

Paths to a Psychology Ph.D.: An Information Session and Workshop

Join Stanford Psychology graduate students, research assistants, and faculty for a free one-day virtual information session and workshop on applying to research positions and Ph.D. programs in psychol...

0227

Reposted by Mike Frank

MIT Press @mitpress.bsky.social · 26/08/2026

"[Babies] only splash about in the shallows of a fathomless ocean of language. And yet, somehow, that’s enough. From a drop, they infer the depths." To understand how children learn human language, @mcxfrank.bsky.social and colleagues studied the experience from a child's perspective.

technologyreview.com

Kids outlearn AI—and we still don’t know why

LLMs need vastly more data than children to learn language. Understanding why could help us create more efficient models—and reveal more about developing minds.

1338

Reposted by Mike Frank

Open Encyclopedia of Cognitive Science bot @oecs-bot.bsky.social · 28/08/2026

The subjective feeling of control often diverges from reality, revealing that our sense of agency is a computational inference rather than a direct… Sense of Agency by Patrick Haggard doi.org/10.21428/e2759450.7b88a436 #CognitiveScience

Open Encyclopedia of Cognitive Science: "Sense of Agency" by Patrick Haggard. Sense of agency refers to the subjective experience of controlling one’s own actions and, through them, events in the outside world. The subjective sense of agency should be distinguished from the obj

1136

Reposted by Mike Frank

Elise Cutts @elisecutts.bsky.social · 24/08/2026

For you to collect good data, your participants need to have a good experience. Ch 12 of Experimentology takes that seriously, covering informed consent, how to make good online experiments, and participant compliance checks. 🧵 experimentology.io

1275

Reposted by Mike Frank

Writing this story was so unbelievably fun. And not only bc @mcxfrank.bsky.social says quotes like this: “Kids’ experience looks really radically different than we thought ... They’ve got these little short arms, so the objects are, like, right in front of them. And they live in a forest of knees.”

35111

Reposted by Mike Frank

Liz Renner @lrrenner.bsky.social · 24/08/2026

AL-ways PI-lot 👏 👏 👏👏👏

0111

Ch 12 is the data collection chapter. Free to read and openly licensed for teaching: 📖 experimentology.io/012-collect... #OpenScience #Psychology

010

There are at least four kinds of compliance checks: passive (completion time), comprehension (do they get the instructions?), manipulation check (did the manipulation work?), and attention checks. We endorse the first three but are cautious about attention checks, which can backfire.

120

We recommend that you always run two pilots before launch. First a non-naive pilot (a friend or colleague) to make sure data logs correctly. Then a small naive pilot (3–20 participants from your real channel) to catch UX issues before full deployment.

161

But online testing brought new ethical issues. Average MTurk wage = ~$2/hour. Less than 5% of workers earned minimum wage. Many experimenters routinely withheld payment based on performance — clearly violating ethics norms but often missed by ethics boards.

100

Online testing is no longer a novelty — over a third of social/personality psych studies were online by 2015, now we'd guess it's the large majority. Data quality is generally comparable to in-lab testing for both questionnaire and experimental paradigms.

100

A consent form has standard required pieces — purpose, risks, benefits, voluntary participation, contact info, data handling. But the goal isn't checking boxes; it's centering the participant. What might they think or feel about doing this study? That's what consent should address.

100

Consent has three prerequisites: competence (capacity to decide), comprehension (the participant actually understands what they're agreeing to), and voluntariness (no coercion, including subtle structural forms). For online studies, comprehension is the one that gets shortchanged most often.

100

1275

"To the dismay of many(yet to the delight of afew), Nature PublishingGroup announced today that its flagship journal, Nature, will no longer accept submissions from humans (Homo sapiens)." Entertaining and prescient piece from Jordan Suchow (2011): suchow.io/assets/docs/...

0100

Yeah we need to normalize disclosure! We can do so much more good work if we are careful.

030

An open database of children’s vocabulary development

Oh crazy. Here’s another productive use: fable redid the full wordbank infrastructure to clear tech debt, eliminate my AWS bill, and improve the visualizations: wordbank.stanford.edu. I also redid childes-db and peekbank!

wordbank.stanford.edu

Wordbank

It’s a great question. I just submitted a paper with work that was done with CC but with more of the ideas coming from me… I included a clear AI acknowledgement in the text and I wrote 100% of the words. But when more of the ideas come from AI, not sure?

030

Claude set the pose detection jobs up on the cluster (using prior code) and batched and ran them - 24 hrs of compute. But in paralelly, it also proposed using inertial motion unit stats from the camera (which I hadn't thought of), a quicker and substantially more precise solution.

370

An example: I was trying to estimate changes in headcam angle across two parts of a large video dataset. I proposed using pose detection statistics, which I thought would reveal statistical changes in the availability and positions of hands and faces.

I don't know what the future holds but these are genuinely wild times for science. I think that's especially true in the intersection of computational methods and human data, where being good at math + data analysis is incredibly helpful for making sense of complex datasets.

190

On the other hand, it is extremely terse and challenging to understand - I have to ask it (like I sometimes ask smart grad students!) to slow down, unpack, and walk me through what it's done. It's humbling to feel that there are better solutions than mine, if only I could understand them!

170

Open Encyclopedia of Cognitive Science bot @oecs-bot.bsky.social · 21/08/2026

Doing science with Claude Code (Fable 5) is fascinating. I give it open-ended, complex tasks and it will go off for 30min and execute very successfully. Sometimes it find much better solutions to my problem than what I suggested.

3304

Reposted by Mike Frank

Reasoning evolved not as a solitary quest for truth, but as a specialized social tool for evaluating and producing arguments with others. Reasoning and Argumentation by Hugo Mercier doi.org/10.21428/e2759450.96cf440a #CognitiveScience

Open Encyclopedia of Cognitive Science: "Reasoning and Argumentation" by Hugo Mercier. By a common definition, reasoning is the private mental act of weighing reasons to decide whether to accept a given conclusion. In argumentation, reasons in favor of a conclusion are exchanged between

01613

Reposted by Mike Frank

Data exploration is critical, but fitting your analysis to idiosyncrasies of your dataset can also increase the risk of spurious results. Ch 11 of Experimentology walks through why preregistration can be a helpful tool for balancing these competing concerns. 🧵 experimentology.io

Tradeoff graph showing increased risk of bias without registration, but decreased sensitivity to discovery with prereg.

1206

Ch 11 opens the Execution section. Free to read and openly licensed for teaching: 📖 experimentology.io/011-prereg.... #OpenScience #Psychology #Preregistration

010

The strongest (but slowest) version of preregistration is Registered Reports, where journals review and accept your study design BEFORE you collect data. The journal commits to publishing regardless of results. This format protects against publication bias too, not just analytic flexibility.

Diagram of registered report, marking two stages of peer review

A common misconception: preregistration locks you into a plan you have to follow rigidly. Not so. Prereg is "a plan, not a prison." Deviations are fine — what matters is being transparent about what was prespecified vs. what was decided after seeing data.

Preregistration is how you mark this distinction. You declare your research decisions publicly before you analyze (and often before you collect) the data. It's a continuum — full prespecification on one end, full exploration on the other — and the goal is transparency.

100

The fix isn't "stop exploring data" — exploration is valuable for finding new patterns. The fix is to clearly distinguish exploratory analyses (after seeing data, higher risk of bias) from confirmatory ones (before, lower risk). Both modes have value, they just answer different questions.