Medical Education Flamingo, MD, PhD
@MedEdFlamingo
Followers
1K
Following
4K
Media
486
Statuses
3K
A curious researcher presents easily digestible #MedEd knowledge for educators & students. Join my free newsletter to get the useful #AI tips for you!
Joined December 2019
Our study published in @JSurgEduc today! 1st-year med students using the AI tool, I vibe-coded through @Replit, outperformed 2nd-years in surgical diagnosis. No human-written content. No human in the loop. Scalable, low-cost diagnostic training is no longer hypothetical.
4
12
41
Our new study has been published today! Large language models are able to generate not only multiple-choice questions but also more complex item types like Key-Feature Questions (KFQs). #MedEd We developed a detailed prompt and quality criteria for KFQs. We found that 15% of
0
1
3
AI isn’t missing from our lives because it lacks capability. It’s missing because our workflows, tools, and organizations are still built for humans. We won’t feel AI’s full impact until we redesign around it.
0
0
0
Non-medical audiences tend to overestimate what current AIs can do in medicine. Not out of hype, but because they miss the nuance. The reality is more modest; powerful tools, yes, but still deeply dependent on clinical context. It requires reasoning under uncertainty. #MedEd
1
4
10
AI has turned the doctor–patient dyad into a triad. Patients and clinicians now both consult chatbots, sometimes with life-or-death consequences. The new skill in medicine is learning to treat AI as a third participant in care. #MedEd
0
5
15
A new AMEE Guide has been published yesterday in @MedTeachJournal. It's on a crucial topic: Formative assessment and feedback in #MedEd. Definitely worths reading.
1
11
25
A further approach would be putting it in a role that we have never thought that humans can do. This could be the most valuable.
0
0
2
If you want to find innovative research ideas: Treat AI as a new species, not a tool. Put it in a role you think only humans can do. Then compare it to humans. #MedEd
1
1
6
Introducing vibe 'vibe coding' Don't know how to code? No problem, there's vibe coding. Don't know how to vibe code? No problem, just ask your browser agent to do it for you!
Neurology Trivia Game Code
3
1
6
However, I cannot deny that giving access to the data makes it work better and eventually life gets easier.
0
0
2
Moreover, it has some other serious risks that you should be aware of.
Do NOT install any agentic browsers like OpenAI Atlas that just launched. Prompt injection attacks (malicious hidden prompts on websites) can easily hijack your computer, all your files and even log into your brokerage or banking using your credentials. Don’t be a guinea pig.
1
2
5
OpenAI’s new AI browser, Atlas, will know your browsing history to "help you better". It's scary to give all data to them. But the truth is if you use Chrome, Google already knows everything. The tradeoff isn’t new, only who you’re giving data to. I’ll stick with Firefox.
1
1
6
In addition to that, our country as a whole ranks third. It shows the excitement of Turkish researchers in #MedEd for using AI.
0
1
3
According to a recent bibliometric analysis, our @Gazi_Universite is among the top 10 universities in terms of research productivity on ChatGPT in medical education. It contributed 2% (8 out of 407) of all publications analyzed between March 2023 and June 2025.
3
2
7
We used GPT-4o-mini to generate both clinical questions and personalized feedback entirely without human involvement, and no hallucinations were detected in expert review. Although the total AI cost was really low (only $0.51 for 5 days and 40 students), we still had to pay to
0
0
15
Published yesterday in @MedTeachJournal: AI wrote student performance summaries in half the time, and just as well. Faculty showed no clear preference between AI and human drafts. When they thought a paragraph was AI-made, they liked it less; even if it was human-written:
0
7
9
"Without a post-event reflective process, what the participants have learned is largely left to chance, leading to a missed opportunity for further learning, and making the simulation encounter less effective." #MedEd Ref: 10.3109/0142159X.2013.818632
0
3
9
"... debriefing, a specific form of feedback, was the most important part of training using simulation, and a respondent called it the ‘‘heart and soul’’ of simulator-based training." #MedEd Ref: 10.3109/0142159X.2013.818632
0
1
7
A meta-analysis published today... 3-option MCQs outperform tradition: * Similar reliability and discrimination as 4–5 option items * Faster to complete * Fewer non-functioning distractors Quality or quantity? #MedEd Stop writing fake distractors and test what matters.
0
3
15
This is what I would do if I had financial support. Our brilliant colleagues in Singapore, @chowminyang and Olivia Ng, ran a 1-day Vibe Coding workshop for medical educators, and participants built an app for medical education. With the help of AI, anyone who can clearly
0
2
18
"Effective learners are able to generate internal feedback and cognitive routines while engaged in a task." Ref: 10.1080/01421590701775453
1
2
16