Sharing the second post in my newsletter series on AI and Health Literacy: https://www.linkedin.com/pulse/3-data-quality-problems-deflate-ai-hype-health-samuel-mendez-phd-yi8pc
After analyzing 48,000+ health messages, I discovered 3 critical data quality problems that challenge prominent AI narratives:
Problem #1: Data redundancy killed 1/3 of our training set.
Duplicate content shrunk our dataset from 48K to 32K samples. Public health's repetitive messaging helps humans remember-but hurts AI training.
Problem #2: Even experts can disagree on "clear" standards Our trained raters struggled with consistency on certain questions for a national audience. If humans can't agree on active voice in tweets, how can AI? Deeper community engagement is key.
Problem #3: Simple tools can beat complex AI.
A basic keyword search for command verbs like "wear" or "wash" can identify calls to action, at a fraction of the cost. It won't capture every possible call to action, but it can capture those you want to promote.
The bottom line: Evidence, not industry hype, should drive our tech choices. Sometimes that means AI. Sometimes it means a simple keyword list.
------------------------------
Samuel Mendez
PhD Student
Harvard T.H. Chan School of Public HealthBoston, MA United States
smendez@g.harvard.edu------------------------------
Original Message:
Sent: 05-19-2025 02:42 PM
From: Samuel Mendez
Subject: AI & Health Literacy: Insights from Recent PhD Research
I recently defended my dissertation, "Computational Health Literacy Assessments: Risks, Opportunities & Future Directions" and wanted to contribute to the broader conversation around AI and public health more quickly than the pace of an academic journal article.
One key takeaway: for one communication problem you might address through AI, you will be creating a few more.
Would love to hear if you find this useful (or not)! I am definitely trying to reach more folks in health literacy and health communication roles.
------------------------------
Samuel Mendez, PhD
Harvard T.H. Chan School of Public Health
------------------------------