Lightnews — Scholar-powered news

Terry McGlynn

@hormiga.bsky.social

9.1K followers 960 following 4.8K posts

Ecologist, entomologist, writer. Chair of Academic Senate and Professor at CSU Dominguez Hills. ScienceForEveryone.science and I'm the Small Pond Science guy. he/him Black Lives Matter. In favor of DEI, justice, access, opportunity. Abolish ICE.

Posts Media Videos Starter Packs

Pinned

Terry McGlynn @hormiga.bsky.social · Aug 26

This is the account that auto-posts my blog posts. 👇🏻👇🏻👇🏻👇🏻

Science For Everyone @hormiga.scienceforeveryone.science.ap.brid.gy · Aug 26

On accepting and acknowledging your authority

If you don't own it, you're failing the people who need you.

scienceforeveryone.science

3 9

Terry McGlynn @hormiga.bsky.social · 1h

this is explained in the piece

Terry McGlynn @hormiga.bsky.social · 4h

Tell me you didn’t read the piece without telling me you didn’t read the piece

Terry McGlynn @hormiga.bsky.social · 4h

Where did you get that I’m uncritically celebrating?? Did you read the piece?

Terry McGlynn @hormiga.bsky.social · 4h

If you take a look my the previous two reposts, would you believe they are about the same piece of writing. My replies have been very Rashomon.

1 1

Reposted by Terry McGlynn

Guy Leonard @guyleonard.bsky.social · 9h

We should be training people NOT to do this.

You have the experience and knowledge to assess the output and correct it and notice mistakes.

Your students and mentees do not.

1 1 10

Reposted by Terry McGlynn

Dr Dr Oona Snoeyenbos-West @oonawest.bsky.social · 5h

Understanding of experimental design and hypothesis testing is fundamental & key💡 Otherwise: garbage in, garbage out! An insightful and thought provoking read here! 👇🏻

Terry McGlynn @hormiga.bsky.social · 3d

Y'all. I just got ChatGPT to do everything in R for this manuscript. I mean EVERYTHING. And it's all legit and reproducible. I'm shook.

How are we mentoring our trainees in statistics now? Who needs to learn coding in R line by line, and who doesn't?

scienceforeveryone.science/statistics-i...

Statistics in the era of AI

How do we mentor, teach, and do stats when AI can do so much of the work?

scienceforeveryone.science

Reposted by Terry McGlynn

Don Moynihan @donmoyn.bsky.social · 1d

Writing something about ICE and sweet Jesus it is so incredibly dark what they are doing to these cities. We have all seen the videos, but when you wade into the details of their actions, it is extraordinarily bleak.

80 800 3.6K

Reposted by Terry McGlynn

Barred and Boujee aka Madiba Dennie @audrelawdamercy.bsky.social · 1d

aaand we are officially in the hide-your-neighbors stage

Sophie Sherry @sophiesherry.bsky.social · 2d

During a military-style raid on a building in Chicago’s South Shore, one resident heard a knock on his door.   It wasn’t the feds — it was a mom and her 7-year-old daughter, pleading for help.

He let them hide in his unit for the next 3 days.

chicago.suntimes.com/immigration/...

Neighbor shielded 7-year-old during South Shore federal raid: ‘I didn’t want them to take her’

During the Sept. 30 raid one tenant protected a terrified girl and her mom. Remnants at the complex, including a detailed map of all the units, offer clues to what authorities may have known before th...

chicago.suntimes.com

300 7.7K 19K

Terry McGlynn @hormiga.bsky.social · 1d

hmmm. yeah. I've only started down this road but I think two ways to approach this is to actually hand over the dataset and ask gpt to conduct the analyses and then, or ask gpt to give you the code (to run in rstudio). I wonder which one is more or less error prone or if it doesn't matter

Terry McGlynn @hormiga.bsky.social · 1d

(just asking because I've had replies (online and otherwise) all over the map about capabilities, and I'm wondering how of much of this is about the version folks are using. Everybody using Claude Code seem to get it to do precisely what they ask it to, as long as the prompts are highly specific)

Terry McGlynn @hormiga.bsky.social · 1d

have you been using ChatGPT5 (the new one?)

Terry McGlynn @hormiga.bsky.social · 1d

That is crux of where we are, I think.

I think it's down to how our communities build consensus and how this winds up in peer review. In my field I know I'll be pushing for emphasizing the human element as much as possible (insofar as the coding aspect is akin to point-and-click software)

Terry McGlynn @hormiga.bsky.social · 1d

and that's the problem. Which is where I was leaving us off with something along the lines of 'what do we need to teach, and how we need to teach differently'

Terry McGlynn @hormiga.bsky.social · 1d

(I didn't read it as "ignoring" ethics.) It didn't get around to mentioning it and I think most of us agree that scientific ethics should be foundational for all of our teaching and training. And yeah, we need to be a lot more explicit about novel ethical challenges associated with the novel tech.

1 2

Terry McGlynn @hormiga.bsky.social · 2d

that's a good point. In that case, I wonder if the people who created point-and-click stats software like GraphPad are also considered to be people debasing the art of coding when they run stats and build figures.

Reposted by Terry McGlynn

Jessica T. @jesstlibrarian.bsky.social · 10d

In non-fiction I'm interested in WHEN TREES TESTIFY: SCIENCE, WISDOM, HISTORY, AND AMERICA’S BLACK BOTANICAL LEGACY by Beronda L. Montgomery. She was recently a Michigan State prof, so I'm partial. #ewgc #booksky

Book cover: Green foliage including fruit, leaves, and flowers, frame the title on a black background.

2 6 22

Terry McGlynn @hormiga.bsky.social · 2d

Also considering how much the community who have grown the R stats into what it is have been all about transparency, public archival, etc., then we shouldn't have any worries about copyright or privacy when it comes to training AIs with previously existing code.

1 2

Terry McGlynn @hormiga.bsky.social · 2d

For those who see the coding process to do statistics as an inherently human endeavor and introducing a robot into the process degrades the art and humanity, I can see why some folks feel that way. But that's not how I feel about it. I see it like driving a car to get from one place to another.

Terry McGlynn @hormiga.bsky.social · 2d

I also think that data management using code created by a robot is not a creative endeavor. But it works. And after you decide on your experimental design and statistical approach, getting a robot to develop that code to make that happen is not art, and that it works.

Terry McGlynn @hormiga.bsky.social · 2d

I can understand that developing efficient and elegant pipelines for data management and the process of making scientific decisions and then codifying them in code to see the results of those decisions is a creative endeavor.

AND....

1 1

Terry McGlynn @hormiga.bsky.social · 2d

I suppose for people who have invested a huge amount of time to develop the technical skills to code to be able to manage complex datasets and run statistical tests, it's weird to see that it is now possible for other people to have that stuff done by a robot.

Terry McGlynn @hormiga.bsky.social · 2d

It’s easy to imagine because people say and do this all the time.

I don’t because I think art is a human endeavor. I think choosing and interpreting statistics is, too. But getting numbers to line up and generate a test statistic (from a test devised by people), I don’t think it needs to be.

1 1

Terry McGlynn @hormiga.bsky.social · 2d

Huh

Terry McGlynn @hormiga.bsky.social · 2d

Have you worked with Chatgpt5 or Claude code?

Terry McGlynn @hormiga.bsky.social · 2d

I’m so sorry. And I promise that everybody in my shop comes first, and anyhow nobody around me cares about grants and pubs as much as the success of our students.