AI chatbots promise clearer health info. While they can sound caring, they can't improve your care. They can be accurate in general terms but fail in complex cases. They can be more complex than existing communications. More details in my latest newsletter below
Feeds: 🛟 😷 🧪 #academicsky
Big fan of @skypeascientist.bsky.social
I always recommend them as a good place to start if you want to get more public engagement experience as a scientist.
And right now they're funding in-person science communication projects.
For Feeds: 🛟 😷 🧪 #academicsky #phdsky
Sharing for relevant Feeds:
#publichealth #episky #academicsky #phdsky
And lastly, the Public Health Connections Lounge group on LinkedIn can also be a good resource. People ask for feedback on resumes and projects, and share job postings on Wednesdays. (Even just searching the #WorkItWednesday hashtag on LinkedIn can be helpful)
Noorain Khan runs a monthly newsletter focused mostly on higher-up roles in the nonprofit sector: noorainkhan.substack.com
Public Health Careers shares postings mostly from state and local health departments in the US:
www.publichealthcareers.org/jobs/
Refreshing my thread of public health (and adjacent) job postings:
First, a spreadsheet by Public Health Hiring Help that is still very much actively updated: docs.google.com/spreadsheets...
oo interesting
Taylor percussion hammer, with a triangular rubber head on a metal handle
Are there any medical professionals out there with any strong opinions on the Taylor reflex hammer? Or is it a totally mundane object in the profession?
Working on a research project and casting a wide net to catch some details I might otherwise miss
#medsky #nursesky #pediatrics #publichealth
Folks in health communication are getting pressure to use AI tools. But how do you know if they work?
I made a practical testing guide: run 25 tests, track failures, calculate error rates, document everything.
No data science background needed.
Feeds: #commsky 🛟 🧪 #PhDsky
P.S. You should still tailor these messages to your audience and context. (i.e. I do not recommend just sending people to the PHCC website to answer their questions).
I've enjoyed learning about the Public Health Communication Collaborative through their ambassador program. They're committed to science-backed resources about vaccines, infectious disease, etc. Their materials are a great professional resource: publichealthcollaborative.org/trending/
🛟🩺💉😷
reposting for feeds: tenure-track assistant professor position in epidemiology at U Minnesota.
#academicsky #phdsky #episky #publichealth
via LinkedIn, job posting list targeting recent MPH grads: docs.google.com/spreadsheets...
Mix of biotech, research, state public health
this short hand-drawn koi loop has been sitting in my youtube drafts so long i can't remember why
so I figured I'd finally share it www.youtube.com/shorts/dtpeZ...
#animation #art
Spritesheet of 8 frames of a black and white pixel art mascara wand entering the tube.
Trying my hand at monochrome pixel art for a new hobby project and it's an interesting mix of tedious, meditative, and fun
youtu.be/gBDPk5euxwI?...
www.youtube.com/watch?v=QEok...
I tend to to just post things related to my health research and health communication work. but I have wanted a change of pace lately so here's a thread of my lo-fi background videos I use as an outlet for the psychedelic animation loops I like to make
#art #animation
I still like to find time to play around with visual art and animation loops... just very very slowly and spradically. Sometimes it takes a year to get around to finishing one. This has aligned well with having fun with lo-fi/bg study music videos, so here's a dumpster fire
youtu.be/2WxXAyb61O4
Our recommendation: Use word length, syllable counts, etc. instead of readability formulas. They still work as goalposts for streamlined writing.
Redo your analysis with different preprocessing decisions.
Test your content with your actual audience. Numbers alone won't tell you if it works.
Results: every preprocessing choice gave significantly different median estimates. Linear regression showed baseline estimates weren’t able to predict this variation.
Methods: We manually counted words, sentences, and syllables in 100 tweets from state health agencies (2012-2022). Then we calculated SMOG and Flesch-Kincaid reading grade levels 8 different ways - sometimes counting URLs, hashtags, and numbers, sometimes not. Then we compared the results.
Goal: to show people in public health how much text preprocessing matters, even if you are using readability formulas by hand. Do you count hashtags as words? URLs? If so, how? These are text preprocessing decisions that change word counts and syllable counts that readability formulas rely on.
Problem: Readability formulas promise simple scores to tell us if our writing is clear enough. But using them requires lots of choices that cause results to vary. In NLP, this is called text "preprocessing". In public health, this is ignored in favor of a focus on "errors" in computer calculations.
Readability scores don't just vary due to "error", they vary due to context-specific decisions. The example text reads: "The deluge in new cases is fueling a sharp rise in #COVID19 hospitalizations. as of this morning, there were 276 virus-stricken patients in our state's hospitals. Thats a 39% increase over the past week and a 109% increase since 3/17" The hashtag is highlighted an annotated with the question, "Count as a 4-syllable word? Count at all?" The date is highlighted and annotated with the question, "Count as a 1-syllable word and 3-syllable word? Count as "March 17"? Count at all?"
New paper out! We studied how different ways to count words can changes reading grade level estimates. Bottom line: Don't take reading grade estimates at face value. AND show your work when you do them! More info in thread below.
Feeds: #commsky 🛟 🧪 #PhDsky
Link: www.tandfonline.com/eprint/MGCTE...
This aligns with my usual refrain: generative AI tools might address 1 problem while causing others. This ripple effect is not part of the conversation with these tools already incorporated into electronic health records, emerging tech tools for sci comm, and in smaller public health departments
