iiinova — where your digital self keeps evolving

Nitzan Shahar @shaharnitzan.bsky.social · 09/09/2026

Fantastic work by @idoba.bsky.social showing humans can't stop learning automatically, even when it leads to losses. Now out ar @natcomms.nature.com !!

051

Reposted by Nitzan Shahar

Paul Sharp @paulbsharp.bsky.social · 03/05/2026

🚨 Our lab's first paper is published! The claim: anxiety distorts not just what looks threatening, but the task models people use to plan. 🧵 This tiny study provides preliminary support that world model variation is critical to explaining anxiety.

onlinelibrary.wiley.com

2124

Nitzan Shahar @shaharnitzan.bsky.social · 03/05/2026

We should keep stats assessments handwritten. But we should also encourage LLM use when teaching. The idea that LLMs prevent learning is too conservative and narrow. It's more about how you use them. LLMs can foster active and critical thinking. Instead of banning their use, we need to design them.

000

Nitzan Shahar @shaharnitzan.bsky.social · 03/05/2026

The fact that the exams should be handwritten doesn't mean teaching should avoid LLMs. You can teach stats with heavy use of custom LLMs, and have a handwritten exam. We got excellent results this way this year. Banning LLMs completly for students during training, is a very conservative reaction.

010

Reposted by Nitzan Shahar

Rob Sica @robsica.bsky.social · 02/05/2026

"no matter how you check, it turns out that under certain conditions modern children and educated adults believe that magical powers operate not only in the domain of the imagination but also in the domain of everyday reality"

sciencedirect.com

Magical thinking: atavism of evolution or a fundamental component of the mind?

Based on a review of empirical evidence, this paper seeks to determine whether belief in magic among contemporary individuals constitutes an ‘evolutio…

2108

Reposted by Nitzan Shahar

Daniel Lakens @lakens.bsky.social · 30/04/2026

New (free) workshop by the Paul Meehl Graduate School on agent-based modeling: paulmeehlschool.github.io/2026-04-24-a... The workshop will be exclusively in-person on May 27, 2026, in Eindhoven. Register before May 18th!

paulmeehlschool.github.io

New workshop announced: Theoretical Agent-Based Models | Paul Meehl Graduate School

We are excited to announce that registration is now open for the workshop, “Theoretical Agent-Based Models”. In this session, Matteo...

01411

Nitzan Shahar @shaharnitzan.bsky.social · 02/05/2026

You get to chose the priors. Promise.

010

Nitzan Shahar @shaharnitzan.bsky.social · 02/05/2026

Let's do a double blind. I think you are in for a surprise. 😀

100

Nitzan Shahar @shaharnitzan.bsky.social · 02/05/2026

Yes, it takes a lot of resources... ☹️ No doubt. I think in this case, they learned better how to read-and-write on their own.

020

Nitzan Shahar @shaharnitzan.bsky.social · 02/05/2026

I really enjoyed the stephen kosslyn's approach. He has a book about active learning with ai, give it try? Or we can agree to disagree.

010

Nitzan Shahar @shaharnitzan.bsky.social · 02/05/2026

So, in my view, we should not debate whether to allow LLM use for beginners or not. It is more about teaching them how to use it, making new LLM components that give responsible results for beginners, and facilitating active rather than passive learning.

150

Nitzan Shahar @shaharnitzan.bsky.social · 02/05/2026

The main issue was to get LLM use in the R course to allow them to face challenges and help with active rather than passive learning. We gave them specific agents that we made, which were designed to ask rather than answer. This made everything much more engaging than the classic way of teaching.

160

Nitzan Shahar @shaharnitzan.bsky.social · 02/05/2026

Saying "this is only for adults" ignores the problem and leaves it without your teaching hand. We don't ban all movies for all ages, we explain what and where is appropriate for which ages and make the rules so everyone can use them. 🙂

100

Nitzan Shahar @shaharnitzan.bsky.social · 02/05/2026

And did they stop using LLMs? We are not going to get far with policing LLM use. I can share from my own experience that making agents for my statistics and R courses only improved learning, by much. LLMs can improve active learning, and active coding, but we have to make the roles.

320

Nitzan Shahar @shaharnitzan.bsky.social · 02/05/2026

Saying "prompting code is useless" is ignoring the problem, in my view. We just finished a beginners R course, based on learning with LLMs. The final exam was closed-book, and they had to read and write code with pen and paper only. They had amazing results. They learned a lot. Yes, by using LLMs.

140

Nitzan Shahar @shaharnitzan.bsky.social · 01/05/2026

I feel AI is not going anywhere. But the wild prompting that students are doing now also can't stay. We have to make up AI tools, elements, agents, that the expert can direct how, when and why to use.

100

Nitzan Shahar @shaharnitzan.bsky.social · 01/05/2026

AI is complete nuts since it has no accountability. This is where stats experts have to step in. Make routines, skills, specific agents, that we can trust more. Create task specific AI elements that teach, code and that their creator can be held accountable if they misbehave.

120

Nitzan Shahar @shaharnitzan.bsky.social · 01/05/2026

When a stat newbie use spss, they also can't validate that it did what it said it did. You have to trust the guys who made spss to begin with. The *reputation* of who ever built spss parts is what's keeping it inline. If it was broken, the community would have noticed.

120

Nitzan Shahar @shaharnitzan.bsky.social · 01/05/2026

We all need to adapt, and fast. No question about that. I think we actually need to *encourage* using AI for coding, but we need to pave the way for correct use. Here what I see at the moment:

120

Reposted by Nitzan Shahar

Liron Rozenkrantz @lironrozenkrantz.bsky.social · 07/04/2026

This is the first computational account of "enhanced rationality" in autism (see our TICS from 2021: www.cell.com/trends/cogni..., allowing us to show that the difference lies specifically in reduced learning from irrelevant information, whereas learning from relevant info was equal between groups.

cell.com

Enhanced rationality in autism spectrum disorder

Challenges in social cognition and communication are core characteristics of autism spectrum disorder (ASD), but in some domains, individuals with ASD may display typical abilities and even outperform...

031

Reposted by Nitzan Shahar

Liron Rozenkrantz @lironrozenkrantz.bsky.social · 07/04/2026

This wonderful collaboration, led by @idoba.bsky.social and co-supervised by me and @shaharnitzan.bsky.social, is now published in Translational Psychiatry (www.nature.com/articles/s41...). We find that autistic individuals show more accurate and less biased credit assignment during an RL paradigm.

nature.com

Autism-associated learning patterns show reduced credit assignment to outcome-irrelevant features - Translational Psychiatry

Translational Psychiatry - Autism-associated learning patterns show reduced credit assignment to outcome-irrelevant features

162

Nitzan Shahar @shaharnitzan.bsky.social · 06/04/2026

Check out our new work, led by the fantastic Ido Ben-Artzi! Individuals with ASD show less outcome-irrelevant learning and are more accurate in following task instructions than their non-ASD peers. @idoba.bsky.social

030

Nitzan Shahar @shaharnitzan.bsky.social · 12/12/2025

It was a pleasure to present at the AI Week at TLV. We presented a systematic review by Gili Katabi looking at more than 200 papers examining ADHD classification using neuroimaging. It seems like the field is getting much better at it, now nearing perfect classification!

010

Nitzan Shahar @shaharnitzan.bsky.social · 05/09/2025

I'm thinking of Gauss, Bayes, Fisher, Pearson, Neyman. Who else should be on the list? Looking for those whose work shaped the way we analyze data today. Would also love to make this more gender diverse!

200

Nitzan Shahar @shaharnitzan.bsky.social · 05/09/2025

I’m teaching an MA state course in psychology & neuroscience, Bayesian stats flavored. Tempted to add some history on the side. Who would you say are the must-include figures in stat history? #stats

251

Nitzan Shahar @shaharnitzan.bsky.social · 05/09/2025

New preprint, showing that human value learning has an automatic aspect! Congrats @idoba.bsky.social for a fantastic, in-depth and rigorous work 🏆

030

Reposted by Nitzan Shahar

Micah G. Allen @micahgallen.com · 05/09/2025

Check out our shiny app for Bayesian power analysis of interoceptive psychophysics!

1315

Reposted by Nitzan Shahar

Tobias Hauser @tobiasuhauser.bsky.social · 19/01/2024

🚨 Job Alert 🚨 Developmental Computational Psychiatry lab is hiring! We are looking for postdocs in CBT, neuroimaging, developmental psychology, and a lab manager. We offer: brand new MRI, great coffee, wonderful place w nature Interested? Email me! Details: devcompsy.org/join-the-lab/ #neuroscience

11820

Nitzan Shahar @shaharnitzan.bsky.social · 24/11/2023

In our papers we always plot the posterior and report median, 89%CI and probability of direction. A nice reviewer said "you should really use 95%CI". How would you reply? Would you make the change? Here's the whole thing discourse.mc-stan.org/t/help-in-re... #stats @rmcelreath.bsky.social

discourse.mc-stan.org

Help in reply to reviewer regarding CI

Hi Everyone, A question on how to handle reviewers when reporting CI. We are doing RL and in our papers when reporting results I always try to put a figure with the whole posterior. We also tend to ...

5101

Nitzan Shahar @shaharnitzan.bsky.social · 24/11/2023

In our papers we always plot the whole posterior and report median, 89%CI and probability of direction. A reviewer said "you should really use 95%CI". How would you reply? Would you make the change? Here's the whole thing discourse.mc-stan.org/t/help-in-re... #stats @rmcelreath.bsky.social

000

Reposted by Nitzan Shahar

Krista Byers-Heinlein @kbyers.bsky.social · 27/09/2023

New preprint with an exciting and unexpected result! The ADHD vocabulary size advantage: Monolingual and bilingual young adults with ADHD have larger vocabularies than controls. Co-authored PhD student extrodinaire Melanie Brouillard (seriously, it was all her idea). osf.io/h2fcj/

osf.io

The ADHD vocabulary size advantage: Monolingual and bilingual young adults with ADHD have larger voc...

Individuals with ADHD and bilinguals have each widely been considered vulnerable to vocabulary deficits compared to their peers. However, the interaction of ADHD and bilingualism in affecting vocabula...

2296

Reposted by Nitzan Shahar

Nora Turoman @noraplethora.bsky.social · 27/09/2023

#workingmemory #cogpsych #devpsych #PsySciSky Please be mindful of what language your post is set to. It might not be the same as the language you write in, and it can mean that your post might not be seen by people whose apps are only set to see posts in given languages.

193

Nitzan Shahar @shaharnitzan.bsky.social · 26/09/2023

Thanks! This looks super flexible.

100

Nitzan Shahar @shaharnitzan.bsky.social · 26/09/2023

Yes, I agree. Also good for initial learning. The example is great - thanks.

000

Nitzan Shahar @shaharnitzan.bsky.social · 26/09/2023

All mighty #stats Do you know of any convenient way to stimulate hierarchical regression data in R? I want students to be able to define population parameters (including cor for group level), factors, sample size, the linear equation, and get some synthetic data.

592

Nitzan Shahar @shaharnitzan.bsky.social · 26/09/2023

Yes, interesting. It's a good idea to look into associations with mood - we will certainly do that. It might just be that hyp/imp, and associated feeling of energy goes with some euphoric and desired feelings. But as you said, all these reasons can co-occur!

010

Nitzan Shahar @shaharnitzan.bsky.social · 24/09/2023

Will send you an empirical result, hopefully before end of year! We are re-running to see if it replicates. Also, for young adults we will ask parents/partners to see if they agree and hold similar positive attitudes.

020

Nitzan Shahar @shaharnitzan.bsky.social · 24/09/2023

Nice. How about 🌱 for development if the graph icon is taken?

110

Nitzan Shahar @shaharnitzan.bsky.social · 24/09/2023

Great idea. Also social media might actually increase these attitudes... We haven't looked - but we can certainly do it for the next study - I will let you know :) Tell if you do get the feed running!

120

Nitzan Shahar @shaharnitzan.bsky.social · 23/09/2023

What would you think is going on? Would you have a different interpretation?

130

Nitzan Shahar @shaharnitzan.bsky.social · 23/09/2023

The other option is that hyperactive/impulsive symptoms simply co-occur with desired tendencies such as energy, creativeness, etc. If you have more sym, you "see" better the other, more desired parts.

120

Nitzan Shahar @shaharnitzan.bsky.social · 23/09/2023

Good question :) One option is that this might be a "coping mechanism". More symptoms might mean ppl need to have a stronger feeling that there are good sides as well. But - if that was true - then we should have seen a positive corr for both adhd subtypes, which we didnt...

120

Nitzan Shahar @shaharnitzan.bsky.social · 23/09/2023

Here is the preprint with the 12 item self-report my grad student Meriav (not here yet) came up with. psyarxiv.com/p3sw6/ @danidonovan.com @mgallen.bsky.social help me spread the word? Maybe put an adhd sci feed?

122

Nitzan Shahar @shaharnitzan.bsky.social · 23/09/2023

We got an effect in the opposite direction :) Turns out more hyperactivity/impulsivity, better attitudes toward adhd! This goes with the many claims by ppl with adhd saying they see many good aspects in this condition.

120