A New Take on the Categorical Imperative: Gatekeeping, Boundary Maintenance, and Evaluation Penalties in Science | Organization Science

Some of us didn't have the luxury of fitting one world growing up. That turned out to be the training we needed.

pubsonline.informs.org/doi/full/10....

A study of 55,497 academic evaluations found the penalty for being multidisciplinary INCREASES with performance. The best boundary-spanners get punished the most. Gatekeepers aren't confused. They're defending territory.

Solving the problem of inception: A cross-species perspective on strategies for a mechanistic refinement of intracortical microstimulation Solving the problem of inception: A cross-species perspective on strategies for a mechanistic refinement of intracortical microstimulation, Kozai, Takashi D Yoshida, Hooks, Bryan Mac, Vazquez, Alberto L, Gharbawie, Omar, Huang, Chengcheng, Graczyk, Emily L, Smith, Matthew, Taylor, Dawn M, Hernandez-Reynoso, Ana G, London, Alex, Ford, Paul, Anderson, Eileen, Chen, Xing, Flick, Amy Irene, Parker, Lisa S, Gaunt, Robert A

Bidirectional translation is not a buzzword.

iopscience.iop.org/article/10.1...

The field's biggest unsolved problems sit at the boundary. Why do implants fail biologically? Neuroscience. How do you design around biological constraints? Engineering. How do you close the loop between clinical observation & mechanism? Someone who thinks in both directions.

When the BRAIN Initiative launched, a group of scientists at an IEEE conference "came out" one-by-one as engineers. They'd been doing engineering all along but framing it as neuroscience because that's where funding and prestige lived.

Everyone was code-switching. Nobody was rewarded for being both

Many engineers believe basic science doesn't matter. Just make better devices. Optimize the signal. The biology is someone else's problem.

Except the biology is why every chronic neural interface eventually fails. You cannot engineer around a problem you refuse to study.

Some neuroscientists believe engineers exist to help "real scientists" do "real work." I've had a collaborator use my grad student's labor and my biosketch to get a grant, then redirect the funds and cut us from paper. Not an authorship dispute. That's disciplinary hierarchy as resource extraction

I am an engineering scientist. Not an engineer. Not a basic scientist. Both and neither. I've spent my career in the gap between neuroscience and neural engineering, where the most important questions live and the fewest people are rewarded for working.

Many of us have actually lost the struggle already. My lab is closing in June, and I'm not the only one. By far.

The Biology of Brain-Machine Interface Failure YouTube video by Dr. Kozai

🎥 The Biology of Brain-Machine Interface Failure — recorded
📺 youtu.be/9HE9ormLgOA

B.I.O.N.I.C. Lab - Request for Information NIH Subject: NIH Strategic Plan RFI (NOT-OD-26-047), deadline May 16. 5 minutes of your time matters. Colleagues, NIH is developing its Strategic Plan for FY 2027-2031 and has issued an RFI (NOT-OD-26-0...

For more context, information: www.bioniclab.org/news/request...

💰 Every result in this talk required multi-year R01 support. Proposed NIH cuts, NOFO collapse, and multi-year funding compression are squeezing pay lines from ~15-20% to ~3-5%.
Industry will not fund pre-competitive biological questions with no product roadmap.
NIH RFI deadline: May 16.

The inflammatory cascade is the upstream input. Circuit-level dysfunction is the output that determines device performance.

Qs asked whether we are ignoring the inflammatory response.
We are not. We study it extensively. This talk focused on what it drives downstream: how glial activation alters neurocomputation at the tissue-electrode interface.

⚡ The neurons are not dying. Many go quiet because the metabolic infrastructure that sustains them has collapsed.
That distinction changes everything about how we design, evaluate, and intervene.

🔬 Clemastine improves chronic recording without reducing glial scar.
FUS conditional KO increasing oligodendrocyte cholesterol biosynthesis improved recording again, but myelin sheath thickness did not predict it.
Cholesterol biosynthesis activity, not myelin wrapping, is the variable that matters.

🧠 Three biological axes nobody is designing around:

BBB breakdown + pericyte loss → chronic leaky BBB
Glial reactivity alters excitability and synaptic function beyond 300 µm, well past the scar boundary
Oligodendrocyte loss cuts off axonal metabolic supply, not just conduction velocity

⚠️ The dominant narrative: BCIs fail bc of materials and foreign body encapsulation.
But signal loss begins before scar maturation. Two shanks on the same Utah array, 400 µm apart, identical in every way, show completely different FBR.
Materials alone cannot solve a fundamentally biological problem.

🎥 A few people asked for slides after my CMU seminar. I re-recorded the full talk instead.
Thanks to Bin He for organizing.

For U.S. medical researchers, shrinking labs and bare budgets are the new reality The Trump administration has slashed the number of grants from the National Institutes of Health, with far fewer focused on women, cancer and mental health.

Bhattacharya, Letai & other politicals at NIH keep trying to to say all at NIH is fine, because they spent the full budget last year. This well researched article from @carolynyjohnson.bsky.social explains exactly why many scientists are still struggling. www.washingtonpost.com/science/2026...

Aberrant accumulation of age- and disease-associated factors following neural probe implantation in a mouse model of Alzheimer’s disease Aberrant accumulation of age- and disease-associated factors following neural probe implantation in a mouse model of Alzheimer’s disease, Wellman, Steven M, Coyne, Olivia A, Douglas, Madeline M, Kozai, Takashi D Y

🌟 BCI devices have strong safety records for the populations studied. I raise this because the time to ask the question is before problems emerge, not after.
Paper: iopscience.iop.org/article/10.1...
Follow the series: bioniclab.substack.com

🩺 The implication is not to stop enrolling participants.
A 35-year-old with early-onset familial Alzheimer's history and one without may respond to the same implant very differently. We do not have the data to quantify that difference in humans yet.

🚫 This does NOT mean BCI implants cause Alzheimer's in healthy humans.
It means that in a brain already on a genetic trajectory toward neurodegeneration, a focal injury event may accelerate that progression.

🔬 We implanted microelectrodes into 2-month-old APP/PS1 mice, a familial Alzheimer's model. These animals do not normally develop MX04-labeled amyloid plaques until 4-5 months.
Following implantation, we observed plaque accumulation 2-3 months ahead of the typical timeline.

⚠️ An uncomfortable finding from our lab that I want to state carefully.
BCI trials are enrolling participants without routine genetic screening for neurodegeneration risk. We have mouse data that makes this worth examining systematically. 🧵

Depth vs. speed. That is the competitive axis. One side is being defunded.

NIH published 756 funding opportunities in 2024. 120 in 2025. 14 by mid-March 2026. The one mechanism funding neural interface biology expired in 2022.

We are not losing papers. We are losing the people who would have written them.

“Publishing gives it to China for free.” The U.S. won genomics with open data. China read the same papers. We won because the moat was not the knowledge but the ecosystem: trained people, startup culture, FDA expertise. Papers are open. Ecosystems are not.

China’s 7-ministry BCI strategy funds devices, manufacturing, regulatory speed. None targets the neurobiology of chronic interfaces. For a device that needs to work at year 5, that gap is not optional. The U.S. built this knowledge. China did not.

U.S. labs have since found failure modes no device addresses. Oligodendrocyte loss starves neurons of metabolic capacity. Functional connectivity degrades even when neurons survive. Stimulation fades within 1 min in human cortex. Safety models are wrong at a biophysical level.