seanpmackinnon.bsky.social
Instructor at Dalhousie University. Personality, statistics, mixed methods
1,642 posts
1,472 followers
531 following
Regular Contributor
Active Commenter
comment in response to
post
At my university, students seeking accomodations for disability have also doubled to around 20-25% of all students since COVID, a big chunk of which is mental illness.
Attempts to access services are way up, even as funding is cut. I doubt social media is the cause, but something has likely changed
comment in response to
post
Ignoring the social media piece for a moment, I think there is reasonable evidence of increased mental health issues overall, post-COVID at least. The stats on disability are more well-tracked, of ehich mental illness is a subset
www.statcan.gc.ca/o1/en/plus/5...
comment in response to
post
I assume it was some time before that post, not in reaction to it.
Try not to take it personally, they block a LOT of people. If you talked a lot about politics in any capacity in the past that is probably why.
comment in response to
post
Always the way
Though I deleted it for other people's bad ttrpg opinions lol
comment in response to
post
I deleted my reddit account a few years ago after one too many annoying replies lmao
comment in response to
post
My perspective these days is that the benefits of grades are solely administrative and logistical, which is why educators continue to be stuck employing them despite their problems.
To eliminate grades means solving problems like transfer credit, scholarships, and admission prerequisites
comment in response to
post
The dillema I find myself in is that grading is a net negative for learning.
Yet, it is the financial engine of the university in terms of credentialing and an administrative tool to apportion too few resources so it is not going anywhere soon.
So grading fairly is the best I can think to do
comment in response to
post
I had tried out perplexity AI literature reviews to try and understand why people liked them.
It was fine, but was mostly just directly copying or paraphasing the abstracts in the same way undergrads have done in papers since forever
comment in response to
post
Dunno, but I'm sure they make more money than me lmao
comment in response to
post
I had a colleague look at me incredulously and with genuine surprise when I said that I don't use it for any of my writing. Like genuinely shocked that I wrote out all my own artisanally crafted words for papers and other academic tasks.
Academia is so often driven by outcome over process
comment in response to
post
Yeah, for sure I am like that too. Gotta have an in progress space that doesn't need archiving!
comment in response to
post
The only way to know is to say it to young people and watch their reaction
comment in response to
post
Update:
"Why is this imputation process taking so long to converge, I shouldn't have to leave it overnight"
*squints at tiny typo buried in the R code*
Right. The problem is me. I'm the problem.
comment in response to
post
Thank you! Exactly what I couldn't remember
comment in response to
post
If you move to frequent in class assessment, you'll need to have something like "drop two lowest grades, missing counts as zero" or (more equitably and less inflationary, imo) "can miss two, and if missed roll the weight into the other assessments"
comment in response to
post
For exams, the sad reality is that a lot of make up exams are unavoidable now.
I try to run only 1 or 2 group make-up sessions, and don't tell them the day/time until after the main exam happens.
If they miss those, move the weight to the final exam (the final exam is cumulative for that resason)
comment in response to
post
The hardest thing when stats consulting is like when it's like:
X is easy, no problem
Y is also easy, no problem
Doing both AT THE SAME TIME? Madness, impossible. Time to quit my job.
comment in response to
post
Online classes are great for accessibility and students can definitely learn from them.
But ... assessment/certification straight up doesn't work. Valid grading isn't really possible, imo.
This was true before LLMs (e.g., people just bought papers) but now it's cheaper, faster, and more noticeable
comment in response to
post
I went in blind knowing nothing and was blown away, honestly. That time travel music scene in the middle had me entranced
comment in response to
post
We prefer "elder millennial" thank you
comment in response to
post
The trick is differentiating between:
a) X is always way worse than Y, such that X is a mistake
b) X is better than Y, but the difference is so slight the results will be virtually identical
I find it hard to be motivated by (b) when the amount of extra effort it takes is intense (lazy, I know)
comment in response to
post
Yeah, this seems about right to me too.
One of the major issues is that statisticians themselves also often are trying to push certain approaches for fame and glory just like all other academics, so may prefer (or disprefer) things based on their own research and writing program.
comment in response to
post
We watched a few episodes, and there was big "I have no memory of this place" energy.
Like, I remember the characters I liked and the broad strokes, but uh, having a really tough time remembering what is happening lol
comment in response to
post
Congrats!!! The real reward is no more applications to fill out for promotion for life
comment in response to
post
My one private joy is that our local MP Andy Philmore (derogatory) jumped ship and became mayor when the libs started tanking b/c he has no loyalty, and now libs are totally going to steamroll this district and another lib is gonna win instead
I say this as an NDP voter who has no chance of victory
comment in response to
post
Being a repository of positively arcane, obscure knowledge is the truest joy of being a prof, ngl
comment in response to
post
What software do you use for reference management? And is it like, searching within the articles for facts/summaries or just APA formatting kinds of stuff?
comment in response to
post
What kind of "cleaning" were you doing on the transcripts? Did you have any worries about confidentiality / ethics (we talk about this at the REB)?
Way back in the day I had to transcribe my own recorded interviews, and it is a really long, soul-sucking job, I know!
comment in response to
post
Also, could you elaborate on the assignment use case? Did you just kind of get an LLM to assign a grade based on your rubric as a point of comparison?
Sorry for all the questions, I just find the current discourse very polarized so trying to understand what others actually like about LLMs
comment in response to
post
Thank you, I really appreciate it.
A lot of the examples seem like what formerly would have been using a search engine (e.g., stackoverflow), so presumably you find it better than a search engine?
The image to .csv seems actually novel and useful; code stuff is well-known.
comment in response to
post
Would you be willing to share more specifically what you used them for most recently and how it saved you time? Most commentary on LLMs is overly general
I haven't found a use case for much of my day-to-day drudgery (grading, journal reviews, ethics reviews, analyzing confidential data, teaching)
comment in response to
post
I swear to god, all I want is for pp to lose his seat and get bullied out of politics forever.
I don't ask for much out of life
comment in response to
post
NYT journalists:
*Open Python IDE*
print("I'm sad")
> I'm sad
"The computer science experts were adamant about this just being a so-called "string", but I was... intrigued"
comment in response to
post
Oh, that does seem useful!! I've been looking for something of the sort for my students, so I'll check it out
comment in response to
post
Snark aside, if I can ever get off my lazy ass, I want to write a paper that is just about the interpretation of coefficients in loglinear models. Like, how to interpret the numbers in words with various transformations. I feel like there needs to be more "how to say stats in words" tutorials
comment in response to
post
Why even have coefficients, when you can have
* = significant
** = even more significant
*** = the most significant
comment in response to
post
Outside of a power simulation (which might not hit anyway if they don't understand) maybe also list the number of rows/trials.
If there is 100 trials, then 400*100 = 40,000 observations.
I've found listing the number of rows helps non-stats people understand.
comment in response to
post
"Calculating p-values and confidence intervals from the output is left as an exercise for the reader"