Youki Terada
@youkiterada.bsky.social
Senior Editor, Research @edutopia.org ~ Former STEM and edtech researcher at UC Berkeley and Lawrence Hall of Science ~ Berkeley B.A./M.Ed. grad
Check out my latest articles: https://www.edutopia.org/profile/youki-terada/
Check out my latest articles: https://www.edutopia.org/profile/youki-terada/
yes, absolutely! Check out this article from @edutopia.org: www.edutopia.org/article/fina...
November 6, 2025 at 5:58 PM
yes, absolutely! Check out this article from @edutopia.org: www.edutopia.org/article/fina...
I asked Claude to generate 20 multiple-choice questions and exactly zero had "A" as an answer, and "C" made up more than half.
August 20, 2025 at 10:29 PM
I asked Claude to generate 20 multiple-choice questions and exactly zero had "A" as an answer, and "C" made up more than half.
New study finds that AI chatbots overgeneralize when summarizing scientific studies:
doi.org/10.1098/rsos...
What made me do a spit-take is that asking them to be accurate—"do not introduce any inaccuracies when summarizing"—actually doubled the overgeneralizations.
h/t @tatertutt.bsky.social
doi.org/10.1098/rsos...
What made me do a spit-take is that asking them to be accurate—"do not introduce any inaccuracies when summarizing"—actually doubled the overgeneralizations.
h/t @tatertutt.bsky.social
![3.2. Asking LLMs for faithful science summaries increased algorithmic overgeneralizations
Explicitly requesting accurate responses from LLMs seems intuitive to retrieve summaries that capture all relevant details of input texts. However, we found that this backfired. Compared to a simple summarization request, asking for responses faithful to the original text produced a twofold increase in the likelihood of generalized conclusions, in some models, increasing overall algorithmic overgeneralizations by up to 15% (e.g. ChatGPT-4o (UI), table 2). This finding extends previous research that suggests adding information intended to improve LLM accuracy in LLM prompts can be counterproductive [46].
One potential explanation of this backfire effect may be related to the content of the accuracy prompt, which stated ‘do not introduce any inaccuracies when summarizing’ (figure 1). Psychological research on humans found ‘ironic rebound’ effects, which can occur when people attempting to free their mind of a target thought experience higher levels of occurrence and accessibility of the thought compared with people intentionally focusing on the thought (e.g. ‘Don’t think of a pink elephant now’ [47]). While this remains to be tested and caution about LLM anthropomorphism is warranted [34], our prompt may have triggered a similar tendency in LLMs. By calling for attempts to free processing of inaccuracy, it may have made the occurrence of inaccuracy more likely, suggesting an algorithmic version of the ‘ironic rebound’ effect.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:swzi6armiswuyjjs6tc4r6ur/bafkreicutu3scl4cfajoaxfkxkijljg2eyja3bva2aaqhenbt7dtumschq@jpeg)
May 20, 2025 at 9:38 PM
New study finds that AI chatbots overgeneralize when summarizing scientific studies:
doi.org/10.1098/rsos...
What made me do a spit-take is that asking them to be accurate—"do not introduce any inaccuracies when summarizing"—actually doubled the overgeneralizations.
h/t @tatertutt.bsky.social
doi.org/10.1098/rsos...
What made me do a spit-take is that asking them to be accurate—"do not introduce any inaccuracies when summarizing"—actually doubled the overgeneralizations.
h/t @tatertutt.bsky.social
If ChatGPT thinks a student is Black, it adopts a more authoritarian tone when giving feedback.
doi.org/10.1080/1539...
doi.org/10.1080/1539...
April 8, 2025 at 5:13 PM
If ChatGPT thinks a student is Black, it adopts a more authoritarian tone when giving feedback.
doi.org/10.1080/1539...
doi.org/10.1080/1539...
An extensive meta-analysis finds that teachers' self-reflection has a bigger impact on student achievement than experience, professional development, or pedagogical knowledge.
It's not just what you know, but knowing what works for each student.
doi.org/10.1016/j.ij...
It's not just what you know, but knowing what works for each student.
doi.org/10.1016/j.ij...
April 7, 2025 at 11:27 PM
An extensive meta-analysis finds that teachers' self-reflection has a bigger impact on student achievement than experience, professional development, or pedagogical knowledge.
It's not just what you know, but knowing what works for each student.
doi.org/10.1016/j.ij...
It's not just what you know, but knowing what works for each student.
doi.org/10.1016/j.ij...
Mastering "overlapping" is key to good classroom management.
Instead of stopping a lesson to address misbehavior, seasoned teachers use subtle cues like "the look" or a name call to keep students on task without breaking the lesson’s rhythm.
doi.org/10.1016/j.ce...
Instead of stopping a lesson to address misbehavior, seasoned teachers use subtle cues like "the look" or a name call to keep students on task without breaking the lesson’s rhythm.
doi.org/10.1016/j.ce...
March 24, 2025 at 11:57 PM
Mastering "overlapping" is key to good classroom management.
Instead of stopping a lesson to address misbehavior, seasoned teachers use subtle cues like "the look" or a name call to keep students on task without breaking the lesson’s rhythm.
doi.org/10.1016/j.ce...
Instead of stopping a lesson to address misbehavior, seasoned teachers use subtle cues like "the look" or a name call to keep students on task without breaking the lesson’s rhythm.
doi.org/10.1016/j.ce...
Interesting little detail I found tucked away in a study. A subtle tweak to grading—going from a 100-point scale to a simple 0-3 scale—made assignments feel less intimidating.
psycnet.apa.org/doi/10.1037/...
psycnet.apa.org/doi/10.1037/...
February 5, 2025 at 7:53 PM
Interesting little detail I found tucked away in a study. A subtle tweak to grading—going from a 100-point scale to a simple 0-3 scale—made assignments feel less intimidating.
psycnet.apa.org/doi/10.1037/...
psycnet.apa.org/doi/10.1037/...
Teachers overwhelmingly feel underappreciated, but public surveys indicate they're highly respected and valued. This is a huge disconnect that's causing too many teachers to quit, a new study suggests.
doi.org/10.1007/s409...
doi.org/10.1007/s409...
February 4, 2025 at 9:42 PM
Teachers overwhelmingly feel underappreciated, but public surveys indicate they're highly respected and valued. This is a huge disconnect that's causing too many teachers to quit, a new study suggests.
doi.org/10.1007/s409...
doi.org/10.1007/s409...
Why are brain breaks so effective? In a recent @edutopia.org article, I unpack the research and share 6 tips to keep students focused during a lesson. #EduSky #EdResearch
www.edutopia.org/article/6-re...
www.edutopia.org/article/6-re...
January 29, 2025 at 5:52 PM
Why are brain breaks so effective? In a recent @edutopia.org article, I unpack the research and share 6 tips to keep students focused during a lesson. #EduSky #EdResearch
www.edutopia.org/article/6-re...
www.edutopia.org/article/6-re...
Teachers make a difference.
In a recent study, nearly 1 in 3 new teachers say they were inspired to teach by the lasting impact of their previous teachers.
doi.org/10.3102/0013...
#EduSky
In a recent study, nearly 1 in 3 new teachers say they were inspired to teach by the lasting impact of their previous teachers.
doi.org/10.3102/0013...
#EduSky
January 14, 2025 at 5:53 AM
Teachers make a difference.
In a recent study, nearly 1 in 3 new teachers say they were inspired to teach by the lasting impact of their previous teachers.
doi.org/10.3102/0013...
#EduSky
In a recent study, nearly 1 in 3 new teachers say they were inspired to teach by the lasting impact of their previous teachers.
doi.org/10.3102/0013...
#EduSky
This year, @edutopia.org put together a collection of our most widely-read, impactful research articles and videos—covering everything from classroom management and executive function to belonging, grading, and engagement. #EduSky
Go check it out: www.edutopia.org/package/a-co...
Go check it out: www.edutopia.org/package/a-co...
December 24, 2024 at 10:08 PM
This year, @edutopia.org put together a collection of our most widely-read, impactful research articles and videos—covering everything from classroom management and executive function to belonging, grading, and engagement. #EduSky
Go check it out: www.edutopia.org/package/a-co...
Go check it out: www.edutopia.org/package/a-co...
School shootings have skyrocketed in recent years.
The resulting vicarious trauma "contributed to declines in interest, preparation, and satisfaction in the teaching profession in the past decade."
Learning can't happen if teachers and students don't feel safe in school. #EduSky
The resulting vicarious trauma "contributed to declines in interest, preparation, and satisfaction in the teaching profession in the past decade."
Learning can't happen if teachers and students don't feel safe in school. #EduSky
December 18, 2024 at 1:09 AM
School shootings have skyrocketed in recent years.
The resulting vicarious trauma "contributed to declines in interest, preparation, and satisfaction in the teaching profession in the past decade."
Learning can't happen if teachers and students don't feel safe in school. #EduSky
The resulting vicarious trauma "contributed to declines in interest, preparation, and satisfaction in the teaching profession in the past decade."
Learning can't happen if teachers and students don't feel safe in school. #EduSky
Teachers are so overworked that 84% say there aren't enough hours in the day to finish all of their work, new data reveals.
This forces teachers to take work home, encroaching on time traditionally set aside for family, friends, and unwinding.
Chart via www.edutopia.org/go/the-resea...
🧵 1/8
This forces teachers to take work home, encroaching on time traditionally set aside for family, friends, and unwinding.
Chart via www.edutopia.org/go/the-resea...
🧵 1/8
December 15, 2024 at 11:37 PM
Teachers are so overworked that 84% say there aren't enough hours in the day to finish all of their work, new data reveals.
This forces teachers to take work home, encroaching on time traditionally set aside for family, friends, and unwinding.
Chart via www.edutopia.org/go/the-resea...
🧵 1/8
This forces teachers to take work home, encroaching on time traditionally set aside for family, friends, and unwinding.
Chart via www.edutopia.org/go/the-resea...
🧵 1/8
Google's AI response to "how much of a lecture do students remember" is to throw pseudoscience at me. sigh.
November 26, 2024 at 10:29 PM
Google's AI response to "how much of a lecture do students remember" is to throw pseudoscience at me. sigh.