Mar 10, 2025
Morph: AI Safety Education Adaptable to (Almost) Anyone
Shafira Noh, Wan Aimran
🏆 Education Track Prize
One-liner: Morph is the ultimate operation stack for AI safety education—combining dynamic localization, policy simulations, and ecosystem tools to turn abstract risks into actionable, culturally relevant solutions for learners worldwide.
AI safety education struggles with cultural homogeneity, abstract technical content, and unclear learning and post-learning pathways, alienating global audiences. We address these gaps with an integrated platform combining culturally adaptive content (e.g. policy simulations), learning + career pathway mapper, and tools ecosystem to democratize AI safety education.
Our MVP features a dynamic localization that tailors case studies, risk scenarios, and policy examples to users’ cultural and regional contexts (e.g., healthcare AI governance in Southeast Asia vs. the EU). This engine adjusts references, and frameworks to align with local values. We integrate transformer-based localization, causal inference for policy outcomes, and graph-based matching, providing a scalable framework for inclusive AI safety education. This approach bridges theory and practice, ensuring solutions reflect the diversity of societies they aim to protect. In future works, we map out the partnership we’re currently establishing to use Morph beyond this hackathon.
Mateusz Jurewicz
Very nice template project! I think it looks at an interesting research area (an LLM’s tendency to produce human-seeming responses) and makes reasonable effort to investigate it within the time constraints of a hackathon. I think some ways to further improve the paper itself could be to clearly definite the term “anthropomorphic” in the introduction (within the context of the presented work), draw more on the existing literature on the subject (e.g. “Anthropomorphic response: Understanding interactions between humans and artificial intelligence agents” by Kim, 2023) and expanding the discussion section. Love that the project includes a nice, freely available github repo and would be excited to see the dataset expanded, perhaps by providing further sub-categories of different aspects of the generated text's anthropomorphism or a rating on an ordinal scale.
Esben Kran
Really cool project and excited to see continued work in this direction! It's a clear improvement over Park et al. (2024) in terms of getting uniform elicitation across dark pattern categories compared to the ShareGPT90k dataset. Interestingly, we seem to replicate the anthropomorphism results with Claude from our short experiments in https://www.apartresearch.com/project/anthroprobe. It would have been interesting to see the correlation with human expert coders (read; the researchers) and spot any annotation mistakes given the difference in annotation adherence. I can see a bunch of great directions to take this benchmark. Great work.
Cite this work
@misc {
title={
@misc {
},
author={
Shafira Noh, Wan Aimran
},
date={
3/10/25
},
organization={Apart Research},
note={Research submission to the research sprint hosted by Apart.},
howpublished={https://apartresearch.com}
}
