Learning Collider
@learningcollider.bsky.social
Building AI that Expands Human Opportunity
Research lab founded and led by UT-Austin economist Peter Bergman. peterbergman.bsky.social
LearningCollider.org
Research lab founded and led by UT-Austin economist Peter Bergman. peterbergman.bsky.social
LearningCollider.org
Incredible work, Noam, Moitshepi, Claire, and the entire Youth Impact squad!
December 31, 2024 at 9:21 PM
Incredible work, Noam, Moitshepi, Claire, and the entire Youth Impact squad!
Bottom line: LLMs can be powerful tools for research with the necessary guardrails. But if those guardrails can't be placed, LLMs should not be used.
Double bottom line: applying these same guardrails for benchmarking analysis and building AI tools support more responsible, accurate models.
Double bottom line: applying these same guardrails for benchmarking analysis and building AI tools support more responsible, accurate models.
December 17, 2024 at 5:14 AM
Bottom line: LLMs can be powerful tools for research with the necessary guardrails. But if those guardrails can't be placed, LLMs should not be used.
Double bottom line: applying these same guardrails for benchmarking analysis and building AI tools support more responsible, accurate models.
Double bottom line: applying these same guardrails for benchmarking analysis and building AI tools support more responsible, accurate models.
2️⃣ Researchers should only use open-source AI-models with transparent training data
3️⃣ Researchers must validate AI-generated outputs by comparing them to outputs from established methods on a small sample and appropriately correct for errors.
3️⃣ Researchers must validate AI-generated outputs by comparing them to outputs from established methods on a small sample and appropriately correct for errors.
December 17, 2024 at 5:14 AM
2️⃣ Researchers should only use open-source AI-models with transparent training data
3️⃣ Researchers must validate AI-generated outputs by comparing them to outputs from established methods on a small sample and appropriately correct for errors.
3️⃣ Researchers must validate AI-generated outputs by comparing them to outputs from established methods on a small sample and appropriately correct for errors.
And spillover effects on truancy rates, through student social network mapping, from Bergman & @maibennett.com
www.learningcollider.org/research-lib...
Hope these are useful!
www.learningcollider.org/research-lib...
Hope these are useful!
Better Together? Social Networks in Truancy and the Targeting of Treatment — Learning Collider: Building AI that Expands Human Opportunity
Because students skip school with friends, social network maps can help target truancy interventions and reduce class absences.
www.learningcollider.org
December 17, 2024 at 4:56 AM
And spillover effects on truancy rates, through student social network mapping, from Bergman & @maibennett.com
www.learningcollider.org/research-lib...
Hope these are useful!
www.learningcollider.org/research-lib...
Hope these are useful!
Related summary on text messaging enrollment plus Bergman, Lasky-Fink & Rogers:
www.learningcollider.org/research-lib...
www.learningcollider.org/research-lib...
Research Summary: Enrollment Methods in Text Messaging Interventions Affects Adoption Rates — Learning Collider: Building AI that Expands Human Opportunity
School districts committed to empowering families with actionable, accessible information in support of student success should consider not only what digital communications tools they use, but how the...
www.learningcollider.org
December 17, 2024 at 4:56 AM
Related summary on text messaging enrollment plus Bergman, Lasky-Fink & Rogers:
www.learningcollider.org/research-lib...
www.learningcollider.org/research-lib...
Thanks, Sarah!
@laurakgee.bsky.social - depending on your audience, we have a quick summary of Bergman & Chan's paper on low-cost text messaging and great resources from J-PAL linked here:
www.learningcollider.org/research-lib...
@laurakgee.bsky.social - depending on your audience, we have a quick summary of Bergman & Chan's paper on low-cost text messaging and great resources from J-PAL linked here:
www.learningcollider.org/research-lib...
Leveraging Parents through Low-Cost Technology: The Impact of High-Frequency Information on Student Achievement — Learning Collider: Building AI that Expands Human Opportunity
Automated text messages to parents show major promise for boosting school attendance and grades at a low cost to districts.
www.learningcollider.org
December 17, 2024 at 4:56 AM
Thanks, Sarah!
@laurakgee.bsky.social - depending on your audience, we have a quick summary of Bergman & Chan's paper on low-cost text messaging and great resources from J-PAL linked here:
www.learningcollider.org/research-lib...
@laurakgee.bsky.social - depending on your audience, we have a quick summary of Bergman & Chan's paper on low-cost text messaging and great resources from J-PAL linked here:
www.learningcollider.org/research-lib...