You agreed to the rule. Did you agree to the experiment?

A review rule does not automatically authorize experimental use of an assignment. This paper examines that boundary through ICML 2026.

papers.ssrn.com/abstract=717...

#ICML #NeurIPS #ResearchEthics #AcademicAccountability #AIResearch

My paper proposal, “Performed Transparency: Media Visibility, Procedural Accountability, and Democratic Resilience in South Korea,” has been accepted for an oral presentation at the IPSA Regional Conference in Wrocław 2026, Track 2: Information Media and Technology.

buymeacoffee.com/jooyeolkim19...

No lab, no institutional funding, no supervisor — but the record keeps moving.

My sole-authored work has been accepted to 4S 2026:

“No Warranty, No Rollback: Critical AI Literacy After Black-Box Automation”

Public record:
jooyeolkim1990.github.io/cv.html

Trust-design problem:

When research platforms use face, voice, and live answerability as trust collateral, they may improve response quality in the short term.

But over time, they risk selecting not for “experts in general,” but for experts willing to monetize exposure at platform rates.

I have several accepted 2026 conference outputs, but no institutional funding.

So I made a public bridge-support cost table: what has been accepted, what costs remain, which ordinary support routes have already been checked, and where the bottlenecks are.

jooyeolkim1990.github.io/support/brid...

Medium sits awkwardly between academia, LinkedIn, and Substack.

It wants curated quality, social distribution, and monetization — but does not reliably deliver authority, reach, or audience ownership.

That is not just a Boost problem. It is a platform bargain problem.

medium.com/@jooyeolkim1...

Public record updated.

2026 conference work:
MCCT, EASST, ILERA Asia, EGOV-CeDEM-ePart.

Still doing independent research without institutional funding — on a 2017 LG Gram running by adapter life support.

Independent research is glamorous, clearly.

buymeacoffee.com/jooyeolkim19...

The USTR overreached on South Korea’s network usage fee debate.

But Korean telecom companies are not innocent either.

This is not simply CP vs ISP.

It is a fight over who gets the final bill for high-bandwidth digital life.

Users are already paying.

What disappears from AI answers doesn’t always feel censored.

It just feels like it was never there.

A short note on AI as an access layer:
jooyeolkim1990.github.io/work/when-in...

AI makes plausible, reviewable output cheap at scale.
The bottleneck shifts from production to selection.

Many institutions respond with gates and badges.
That is gate tightening, not better selection.

When overload is internal, stricter entry mostly manages appearances.

#AIinAcademia #arXiv

AI didn’t replace creators. It standardised them.

Markets don’t erase meaning — they erase friction.
Recent experimental evidence also suggests generative AI compresses the “average case” (NBER WP 34851, 2026).

Full piece here: medium.com/@jooyeolkim1...

Prolific denied verification access with no actionable reason or appeal.
No VPN used (SKB), passport checked.

No actionable “issue detected — retry” notice.
Instead: permanent denial + new accounts blocked.

Opaque exclusion → potential unreported selection bias in Prolific data.
#OpenScience

A small working paper revisiting Universe 25 —
not as pathology, but as a structural case of participation collapse under abundance.

Preprint here: doi.org/10.5281/zeno...

#economicsociology

Quick update:

CI v1.4 now has a public overview.
A short bridge paper revisiting Universe 25 (circulation failure, not abundance) is going up next.

Hardware is failing, but the work is still moving.

buymeacoffee.com/jooyeolkim19...

Potential sampling bias note:
Several peer-reviewed studies recruit Korean participants via Prolific.
However, Prolific’s VPN/IP filtering appears to block major Korean ISPs, even for non-VPN users.
This creates a risk of unreported sampling bias.
#MethodsMatter #SamplingBias #OpenScience