iiinova — where your digital self keeps evolving

Athanassios Protopapas @protopap.bsky.social · 01/10/2026

Lessons from students in a very highly selective private liberal arts college may not generalize very broadly to other types of institutions.

100

Athanassios Protopapas @protopap.bsky.social · 21/08/2026

There has been a lot of high-quality work on this topic in Norway, and as far as I know the findings have been unequivocally positive. Here are a couple of recent reports: doi.org/10.3386/w35109 doi.org/10.1080/1934...

doi.org

011

Athanassios Protopapas @protopap.bsky.social · 11/07/2026

I agree that this is not a downside to kagi. Good for them, and for us, that they make this service available for paying customers. I was merely reacting to "RIP Google".

011

Athanassios Protopapas @protopap.bsky.social · 10/07/2026

Is his book enshittification, Cory Doctorow writes that Kagi uses Google data for the search (paying Google for the database), just ranking them differently. His point is that Google has the right answers but shows suboptimal results on purpose (to increase ad revenue due to more searches needed).

131

Athanassios Protopapas @protopap.bsky.social · 17/05/2026

Beginner readers do not benefit from previewing upcoming words but are still negatively impacted by them. We found strong interference by adjacent words or unknown font strings in Grades 2 and 5. This can be a major factor in fluency development. doi.org/10.1016/j.co... @cognitionjournal.bsky.social

doi.org

Redirecting

041

Athanassios Protopapas @protopap.bsky.social · 11/05/2026

Good luck getting the message across. I've been wrestling with this in the context of dyslexia for a long time. doi.org/10.3390/brai...

doi.org

022

Athanassios Protopapas @protopap.bsky.social · 17/04/2026

Do you speak Norwegian? If yes, come join us as a doctoral student or postdoctoral researcher, to work on randomized controlled trials of math and writing interventions for upper elementary school children. www.jobbnorge.no/en/available... (predoc) www.jobbnorge.no/en/available... (postdoc)

001

Athanassios Protopapas @protopap.bsky.social · 17/04/2026

Maybe letting the UK back in was not such a great idea after all? With all the research cuts now the whole continent is looking to ERC for funding...

110

Athanassios Protopapas @protopap.bsky.social · 16/04/2026

Words are always surrounded by other words, forming sentences, in natural reading; and interference from adjacent stimuli has long been known in other visual tasks and domains. So it may sound strange that no model of visual word recognition or reading eye movements has considered this.

000

Athanassios Protopapas @protopap.bsky.social · 16/04/2026

Printed words are slower to recognize when surrounded by other words, especially at short distance. Interference is greater from the right side, where the adjacent word is also simultaneously lexically activated: doi.org/10.1016/j.jm...

doi.org

Redirecting

130

Athanassios Protopapas @protopap.bsky.social · 30/03/2026

...and aligns with the solution of a related paradox: that the rate of reading a sequence of words is better predicted by the rate of naming an array of digits than by the speed of reading the words individually doi.org/10.1037/edu0... (esp. "The Link to Reading Fluency" in the general discussion)

psycnet.apa.org

APA PsycNet

010

Athanassios Protopapas @protopap.bsky.social · 17/03/2026

What bothers me most are not the structural problems but the students' diminished engagement and gravitation toward AI rather than learning. It is especially disheartening in special education, where they will be expected to be able to help vulnerable human beings who need them to know what to do.

001

Athanassios Protopapas @protopap.bsky.social · 06/03/2026

...won't even link them to the final publication if the publisher has provided that directly! And while it does suggest some relevant and interesting studies, it also suggests many that are irrelevant, already seen many times, or even my own. They're clearly not improving it; what are they doing?

000

Athanassios Protopapas @protopap.bsky.social · 06/03/2026

Are we witnessing the enshittification of ResearchGate? Started promising and useful, but hasn't really gotten over toddler-stage yet. Won't read pdfs correctly, won't parse metadata from doi correctly, won't identify known authors correctly, won't join endlessly proliferating preprints,... 1/2

100

Athanassios Protopapas @protopap.bsky.social · 07/01/2026

I got it too, and I played because I was curious. Unsurprisingly, the ideas were vague nonsense. Perhaps some AI people think research is word play? Maybe *their* research is. In my experience, ideas have never been the bottleneck in research. Again, people are trying to use AI for the wrong thing.

010

Athanassios Protopapas @protopap.bsky.social · 19/12/2025

You can enter a nonbreaking space (through insert symbol) somewhere, then copy it, and paste it into the search field to replace globally with a regular space. Whoa, they're all gone!

010

Athanassios Protopapas @protopap.bsky.social · 15/12/2025

(just an irrelevant momentary distraction from doomscrolling)

020

Athanassios Protopapas @protopap.bsky.social · 15/12/2025

My first publication ever, 30 years ago, was in Nature (cognitive research on Mt. Everest is cool!), with serious limitations, severely underpowered and relentlessly p-hacked, edited from several thousand words down to 500. Still, a “not” was removed after proofs had been corrected and approved.

Scanned text reading "Fatigue also does explain the differential deficits we found."

260

Athanassios Protopapas @protopap.bsky.social · 09/12/2025

Perhaps this means that people do recognize 2 words simultaneously after all, because if they didn't how would they know they form a compound?

200

Athanassios Protopapas @protopap.bsky.social · 15/10/2025

In Greece, 6 pm is afternoon coffee time, not dinner. But anyway if you're invited to someone's house at 6 for anything, it would be very impolite to show up before 6:30, as nothing would be ready and the host would probably not even be dressed yet. Some time between 6:30 and 7:00 would be normal.

110

Athanassios Protopapas @protopap.bsky.social · 15/10/2025

In Norway, in my experience when you're invited at six you're expected to ring the bell within a few minutes of 6:00, but I guess you're sort of OK within 10 minutes or so. People text to say sorry they're late even if it's 10 minutes late.

010

Athanassios Protopapas @protopap.bsky.social · 20/06/2025

When naming visual sequences, efficiency calls for parallel processing of successive items but this leads to interference that threatens performance. To manage the load, constant monitoring allows dynamic adjustment of parallelism, seen in eye-voice span and articulation. dx.doi.org/10.1037/xlm0...

dx.doi.org

APA PsycNet

031

Reposted by Athanassios Protopapas

Jonathan Tonkin @jdtonkin.bsky.social · 29/04/2025

What I wish I knew: 33 thoughts for early career researchers Reflections I wish someone had shared with me early in my research career. predirections.substack.com/p/what-i-wis...

predirections.substack.com

What I wish I knew: 33 thoughts for early career researchers

Thirty three reflections I wish someone had shared with me early in my research career.

913569

Athanassios Protopapas @protopap.bsky.social · 23/04/2025

In two experiments we used a flanked masked (75 ms) visual word version of the visual world paradigm with five flanking conditions: none, repeated symbols, unknown font strings, pseudowords, and words. Flankers interfered with activation of the target word, more so for frequent words.

000

Athanassios Protopapas @protopap.bsky.social · 23/04/2025

Skilled readers use information from upcoming words to boost efficiency, but at what cost? We discovered that nearby words interfere with lexical activation of the fixated word, calling for a more nuanced approach to the role of preview in fluent reading. dx.doi.org/10.1016/j.jm...

163

Athanassios Protopapas @protopap.bsky.social · 29/01/2025

LQ is not only about the detailed content. It also requires strong and redundant connections among the three domains. As for improving them, I would say that reading is the established way.

110