Alex Shackman

Alex Shackman

@ajshackman.bsky.social

Laboratory for Affective & Translational Neuroscience, University of Maryland | shackmanlab.org | affective neuroscience | "pessimism of the intellect, optimism of the will.”

2.3K followers711 following2K posts

Alex Shackman
@ajshackman.bsky.social

#RedTeamEverything AI outsourcing is a mistake, whether you are a trainee or full prof. Not using AI to learn, grow, self-critique, check, test, and generally facilitate rigor at scale is also dumb.

Tim Verstynen@tdverstynen.bsky.social

One thing that has genuinely shocked me over the past year is how many faculty are now using LLMs for nearly every part of their jobs. Not as tools to assist their thinking, but to offload the entire job itself. 1/7

Alex Shackman reposted

Robin Aupperle
@raupperle.bsky.social

Help spread the word about the Northeastern Oklahoma Psychology Internship Program (NOPIP) consortium, APA-accredited for >25 years! At the Laureate Institute for Brain Research, interns have a protected day of research in our clinical and neuroscience research labs. Deadline 11/15! nopip.org

nopip.orgNortheastern Oklahoma Psychology Internship ProgramNortheastern Oklahoma Psychology Internship Program (NOPIP) is an APA-accredited internship consortium and a member of APPIC. We are comprised of 5 agencies located in and around Tulsa, OK. Our commit...

Alex Shackman reposted

Phil Lewis
@phillewis.bsky.social

Students from North Carolina A&T can be seen marching back to their HBCU after voting on the first day of early voting in the state. The Guilford County Board of Elections voted to remove the school’s on-campus polling site. (video via Todd Simmons) greensboro.com/news/local/g...

Alex Shackman
@ajshackman.bsky.social

Agree. We're redteaming our own work too.

👊 Ulrich Schimmack 👊💥 ❤️ 🇨🇦 🇪🇺 🇺🇦 👎🇺🇲@ulrichschimmack.bsky.social

AI finds more mistakes in the reviews than in my manuscript. What does that tell us about academic publishing? replicationindex.com/2026/10/15/p...

replicationindex.comPeer Review Uncovers Errors in Human Information Processing: But Who Is Making Them? - Replicability-IndexPeer review is supposed to uncover errors in manuscripts. Sometimes it does. Sometimes it provides a rather nice demonstration of the errors humans make while trying to find errors in other people's w...

Alex Shackman reposted

Steven Meisler
@stevenmeisler.com

Diffusion MRI (dMRI) is a powerful tool to study white matter maturation. In our new preprint, we process and distribute a new resource of >24,000 ABCD dMRI scans using open source tools! We then evaluate how methods shape inferences about development. 🔗 www.biorxiv.org/content/10.6...

Alex Shackman reposted

Matteo Carandini
@carandinilab.net

If you liked this piece on Adrian's discoveries, consider reading the book that inspired me to write it: "Galvani's Spark", by Alan McComas. Strong recommend. It's expensive if you buy it new, but often available used (e.g. on AbeBooks) for a reasonable price. global.oup.com/academic/pro...

The Transmitter @thetransmitter.bsky.social

In two 1926 articles, Edgar Douglas Adrian revealed that neurons communicate through trains of spikes. @carandinilab.net describes the remarkable discovery and how it shifted focus from the medium (the nerve) to the message (the neural code). #neuroskyence www.thetransmitter.org/history-of-n...

thetransmitter.orgThe neural code at 100In two 1926 articles, Edgar Douglas Adrian revealed that neurons communicate through trains of spikes.

Alex Shackman reposted

Nicole Rust
@nicolecrust.bsky.social

Good AI use case: code check! We’ve long had a “second pair of eyes” on our data analyses pipelines before a paper leaves the lab. Even for those who are ”AI rather not”, Claude Opus is both good at that job and fast; it finds things others do not.

Alex Shackman
@ajshackman.bsky.social

1/2. Dr. Rust is correct and you can audit so much more than just code/pipelines. You can audit practically everything...

Nicole Rust@nicolecrust.bsky.social

Good AI use case: code check! We’ve long had a “second pair of eyes” on our data analyses pipelines before a paper leaves the lab. Even for those who are ”AI rather not”, Claude Opus is both good at that job and fast; it finds things others do not.

View full thread
Alex Shackman
@ajshackman.bsky.social

also - darn it. mis-pasted the punchline slide. all of these project were QB'ed by passionate, committed, talented scientists. all of them had at least one issue worth a conversation, some deeper sleuthing, or a fix. all of them were caught before we shipped to you, the larger scientific community.

Alex Shackman
@ajshackman.bsky.social

and once final thought - you can also audit papers and preprints (but not confidential ms's) against shared github/osf/zenodo code and against pre-registration.

Alex Shackman
@ajshackman.bsky.social

1/2. Dr. Rust is correct and you can audit so much more than just code/pipelines. You can audit practically everything...

Nicole Rust@nicolecrust.bsky.social

Good AI use case: code check! We’ve long had a “second pair of eyes” on our data analyses pipelines before a paper leaves the lab. Even for those who are ”AI rather not”, Claude Opus is both good at that job and fast; it finds things others do not.

Alex Shackman
@ajshackman.bsky.social

Not shown in the slide-deck - Auditing papers against the underlying code and output. Do the methods verbiage and results in the human-crafted paper actually match what was done or is there slippage? Was anything omitted? Any errors?

Alex Shackman reposted

tor
@torwager.bsky.social

A new paper just out on negative affect and emotion regulation in the amygdala (Bo et al., 2026, eLife). FMRI in two large community samples (N = 358), based on Pete Gianaros' P01 data...a wonderful collaboration! elifesciences.org/reviewed-pr... Thread below 👇

View full thread
Alex Shackman
@ajshackman.bsky.social

Ugly Bonus 2! Red = Threat > Safe, n=220, no smoothing. Blue = VENAT (any vein; not just BVR). Ce/CeA = Green. Yellow = Th>Sf in Ce ROI. Purple = Activation in VENAT. White = All 3 (not evident).

View full thread
Alex Shackman
@ajshackman.bsky.social

5/6 Oversimple! “Smoothing…merges…BVR... [venous] signal w/ amyg”. Results seem to reflect signal mixing/artifact at the edge and noise reduction further from BVR. Decisive claims not licensed by results. But smoothing is clearly A Bad Idea

Alex Shackman
@ajshackman.bsky.social

6/6 Bottom line. #1. Claims from our group and others re Ce/CeA recruitment are sound. #2. Use unsmoothed data for assessing amygdala engagement, esp for nuclei claims. Be mindful of “artifact”/mixing at the dorsomedial edge. Axial slices informative. A careful, valuable preprint! Kudos to authors.

View full thread
Alex Shackman
@ajshackman.bsky.social

3/6 Misleading! Amygdala is “substantially confounded.” The “confound” appears to be a mix of drainage from amyg (i.e. source not sink) + other regions. BVR lags the amygdala in ~95% of subs. Tongue movement activates BVR, but NOT amyg, weakening over-simple “confound” statements

Alex Shackman
@ajshackman.bsky.social

4/6 Overstated! “prevents reliable attribution” to amyg nuclei (Fig 4). Localization issues appear to mainly be for nuclei within ~4 mm of BVR (e.g. AAA, ATA, CMN). BLA and Ce appear comparatively safe w/ unsmoothed data

Alex Shackman
@ajshackman.bsky.social

things i've learned working w/ claude over the past 6 months - better ways to implement robust regression, better techniques for comparing dependent correlations, better ways to systematically test the robustness/fragility of brain-beh/-symptom associations, stronger exp critique (biases/confou)

Alex Shackman
@ajshackman.bsky.social

...and the Jeffreys–Lindley paradox (just signif frequentist p's + big'ish n's can have stronger bayesian evidence for the null) and the interpretive tricksiness of ordinal interactions (sensitivity to arbitrary scale/measure choices; e.g. time vs speed; can create, nullify, or reverse sign)...wild!