Elamur Research⌕
☰
home / papers / 2025 / Bridging the Socioeconomic Gap in Education: A Hybrid AI and Human Annotation Approach

Bridging the Socioeconomic Gap in Education: A Hybrid AI and Human Annotation Approach

Summary

The paper addresses the shortage of high-quality annotated texts about students with low socioeconomic status (low-SES). NLP has barely studied this topic: only about 20 papers in the ACL Anthology explicitly mention SES. The authors took the Kelbessa et al. (2024) dataset of 74 Reddit narratives and checked it. 64 narratives met the criteria. To these they added 64 contrastive posts that were the least semantically similar to them. They then expanded the dataset with a hybrid pipeline over three iterations. First, traditional ML models and LLMs classified the posts. Next, clustering filtered out outliers, and finally humans verified the result manually. The negative examples for each new round were taken from the false positives of the previous one, so the task became harder each time. From a pool of nearly 800,000 posts, they collected 323 new narratives, so the corpus became roughly five times larger than the initial one. The LLaMA model extracted information from the texts about the authors' backgrounds, their difficulties and the ways they cope with them. Topic analysis showed that the new data cover a wider range of topics. The work shows how to grow a corpus for understudied social groups at low cost by combining automatic models with human annotation.

Key points

  • Candidate pool: 799,032 posts from the Reddit archive for 2005–2013 (over 1 TB), selected from the subreddits college, broke, financialaid, fafsa and others. On average, a post has 134 words.
  • Negative examples were found using a Sentence Transformer from the top of the MTEB leaderboard. The authors compared cosine similarity, negative Euclidean distance and negative Manhattan distance. The 64 least similar posts were checked manually.
  • 22 models from scikit-learn and Hugging Face Transformers were compared: RF, MNB, SVM, LR, XGBoost, RoBERTa, DeBERTa, ELECTRA, ALBERT, XLNet, as well as LLMs in few-shot mode. Evaluation used 5-fold cross-validation with a 70/15/15 split.
  • 323 new narratives were added to the 64 initial ones. PCA and K-means identified five topic clusters: work, family, mental health, societal challenges, and criticism of the system (cost of education, student debt).
  • Sentiment in the original dataset: 10 positive, 29 neutral and 25 negative texts; in the collected one, 37, 138 and 148. The authors promise to release the code and data. Among the limitations, they name the focus on US higher education and incomplete information extraction by the LLM.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2025.conll-1.23

@inproceedings{abdelgaber2025bridging,
  title = {{Bridging the Socioeconomic Gap in Education: A Hybrid AI and Human Annotation Approach}},
  author = {Abdelgaber, Nahed and Jahan, Labiba and Doshi, Arham Vinit and Suri, Rishi and Pavel, Hamza Reza and Zhang, Jia},
  booktitle = {Proceedings of the 29th Conference on Computational Natural Language Learning},
  year = {2025},
  doi = {10.18653/v1/2025.conll-1.23},
  url = {https://doi.org/10.18653/v1/2025.conll-1.23}
}

References

17 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Nancy E. Adler and Katherine Newman. 2017. The impact of socioeconomic status on access to healthcare: A review of the literature. Social Science Medicine, 181:25–33. find
  2. Jacob H. Brown and Michael L. Carr. 2013. The impact of financial aid on college enrollment and completion: Evidence from a randomized study. find
  3. Jose Camacho-collados, Kiamehr Rezaee, Talayeh Ri-ahi, Asahi Ushio, Daniel Loureiro, Dimosthenis Anty-pas, Joanne Boisson, Luis Espinosa Anke, Fangyu Liu, and Eugenio Martínez Cámara. TweetNLP: Cutting-edge natural language processing for social media. find
  4. Amanda Cercas Curry, Zeerak Talat, and Dirk Hovy. 2024. Impoverished language technology: The lack of Motti Kelbessa, Estephanos Jebessa, and Labiba Jahan. 2024. Addressing educational inequalities of low ses students: Leveraging natural language processing for impact. In Proceedings of the 17th Int… find
  5. Vasileios Lampos, Nikolaos Aletras, Daniel Preoţiuc-Pietro, and Trevor Cohn. 2014. Predicting and characterising user impact on twitter. In Proceedings of the 14th Conference of the European Chapter of the Association for Computational Linguistics, pages 405–413. find
  6. Sang Min Lee, M Harry Daniels, Ana Puig, Rebecca A Newgent, and Suk Kyung Nam. 2008. A data-based model to predict postsecondary educational attainment of low-socioeconomic-status students. Professional School Counseling, 11(5):2156759X0801100504. find
  7. Fons Mentink. 2016. Machine driven predictions of the socio-economic status of twitter users. Master’s thesis, University of Twente. find
  8. Niklas Muennighoff, Nouamane Tazi, Loïc Magne, and Nils Reimers. 2022. Mteb: Massive text embedding benchmark. arXiv preprint arXiv:2210.07316. find our card
  9. F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay. 2011. Scikit-learn: Machine learning in python. Journal of Machine Learning Research, 12:282… find
  10. Gerard Salton and Christopher Buckley. 1988. Term-weighting approaches in automatic text retrieval. Information Processing & Management, 24(5):513–523. find
  11. Judith Scott-Clayton. 2015. The role of financial aid in promoting college access and success: Research evidence and proposals for reform. Journal of Student Financial Aid, 45(3):3. find
  12. Southern Methodist University (SMU). 2019. Think-play-hack: World views. find
  13. Stefan Stieglitz, Milad Mirbabaie, Björn Ross, and Christoph Neuberger. 2018. Social media analytics– challenges in topic discovery, data collection, and data preparation. International journal of information management, 39:156–168. find
  14. Marvin A Titus. 2006. Understanding college degree completion of students with low socioeconomic status: The influence of the institutional financial context. Research in Higher Education, 47:371–398. find
  15. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. 2023. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971. find
  16. Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pierric Cistac, Tim Rault, Remi Louf, Morgan Funtowicz, Joe Davison, Sam Shleifer, Patrick von Platen, Clara Ma, Yacine Jernite, Julien Plu, Canwen Xu, Teven Le Scao, Sylvain Gugger, Mariama Drame, Quentin Lhoe… find our card
  17. Wenpeng Yin, Jamaal Hay, and Dan Roth. 2019. Benchmarking zero-shot text classification: Datasets, evaluation and entailment approach. arXiv preprint arXiv:1909.00161. find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/abdelgaber-2025-bridging-socioeconomic-gap-education-hybrid/