Debarati Das

I am a part of the IDEAS Research group led by Scott Counts, where I work on measuring and steering the impact of Generative AI on Productivity.

During my PhD, I was a member of the Minnesota NLP group led by Prof. Dongyeop Kang and was co-supervised by Prof. Jaideep Srivastava, who leads the Data Mining and Research Group (DMR) at the University of Minnesota Twin Cities.

Previously, I have interned at Microsoft Research and 3M R&D. Before completing my PhD, I earned my Master’s degree at the University of Minnesota Twin Cities, where I worked on applying NLP to online social media analysis.

My research sits at the intersection of natural language processing and computational social science, with a focus on designing and evaluating large language models (LLMs) in human-centered, high-stakes contexts. I’m especially interested in how LLMs can support expert decision-making by participating in structured and interpretable workflows. Alongside that, I work on characterizing the generative behavior of LLMs, with an emphasis on understanding how their outputs align (or misalign) with users’ values, expectations, and goals. This includes applications like email personalization, task delegation, and other AI-augmented workplace tools where trust, intent matching, and social nuance are essential.

I’m excited about building modular, accountable NLP systems that not only generate fluent text but also support meaningful human–AI collaboration in real-world environments.

news

Aug 20, 2025	🥳 Our PATHs paper is accepted to EMNLP 2025! See you in Suzhou, China :)
Jul 07, 2025	🥳LawFlow is accepted to COLM 2025! See you in Montreal :)
Apr 30, 2025	✅ LawFlow, which captures lawyers’ end-to-end thought processes, is out on ArXiv!
Dec 15, 2024	🎤 Gave a talk titled “When AI learns to Listen: Generative AI in the Social Sciences” to high school students for the Sparkway Learning PythonAI Workshop 2024.
Nov 04, 2024	🎤 Honored to be a speaker at the Women in AI and DS Conference 2024 and discussed the impact of artifacts in LLM-generated data with an incredibly talented crowd!
Oct 08, 2024	🏆 I received the 2024 AnitaB.org Advancing Inclusion scholarship!
Jul 21, 2024	🥳 Our paper “Publics’ Perceptions of Legitimacy in Corporate Social Advocacy: A Computational Analysis of the Influence of Ideological Congruence” is published in Public Relations Review
Jul 18, 2024	🎤 Gave a talk titled “Beyond Words: Incorporating Graph Data into LLMs for Comprehensive Social System Analysis” at IC2S2 2024! P.S The cheesesteak is just fantastic at Philadelphia 🤤
Jun 16, 2024	✅ Presented my paper “Which Modality should I use – Text, Motif, or Image? : Understanding Graphs with Large Language Models” at NAACL 2024 in Mexico City! [Tweet]
Jun 06, 2024	🥳I am thrilled to be interning at Microsoft Research this summer!

selected publications

preprint

Prototypical Human-AI Collaboration Behaviors from LLM-Assisted Writing in the Wild

Sheshera Mysore, Debarati Das, Hancheng Cao, and 1 more author

May 2025

Bib PDF

@misc{mysore2025prototypicalhumanaicollaborationbehaviors,
  title = {Prototypical Human-AI Collaboration Behaviors from LLM-Assisted Writing in the Wild},
  author = {Mysore, Sheshera and Das, Debarati and Cao, Hancheng and Sarrafzadeh, Bahareh},
  year = {2025},
  month = may,
  eprint = {2505.16023},
  archiveprefix = {arXiv},
  primaryclass = {cs.CL},
}

preprint

LawFlow : Collecting and Simulating Lawyers’ Thought Processes

Debarati Das, Khanh Chi Le, Ritik Sachin Parkar, and 8 more authors

May 2025

Bib PDF

@misc{das2025lawflowcollectingsimulating,
  title = {LawFlow : Collecting and Simulating Lawyers' Thought Processes},
  author = {Das, Debarati and Le, Khanh Chi and Parkar, Ritik Sachin and Langis, Karin De and Madson, Brendan and Berryman, Chad M. and Willis, Robin M. and Moses, Daniel H. and McDonnell, Brett and Schwarcz, Daniel and Kang, Dongyeop},
  year = {2025},
  eprint = {2504.18942},
  archiveprefix = {arXiv},
  primaryclass = {cs.CL},
}

journal

Publics’ perceptions of legitimacy in corporate social advocacy: A computational analysis of the role of ideological congruence

Hao Xu, Debarati Das, Jisu Huh, and 2 more authors

Public Relations Review, May 2024

Bib PDF

@article{xu2024publics,
  title = {Publics’ perceptions of legitimacy in corporate social advocacy: A computational analysis of the role of ideological congruence},
  author = {Xu, Hao and Das, Debarati and Huh, Jisu and Rim, Hyejoon and Srivastava, Jaideep},
  journal = {Public Relations Review},
  volume = {50},
  number = {4},
  pages = {102486},
  year = {2024},
  publisher = {Elsevier},
}

preprint

Under the surface: Tracking the artifactuality of llm-generated data

Debarati Das, Karin De Langis, Anna Martin-Boyle, and 8 more authors

arXiv preprint arXiv:2401.14698, May 2024

Bib PDF Website

@article{das2024under,
  title = {Under the surface: Tracking the artifactuality of llm-generated data},
  author = {Das, Debarati and De Langis, Karin and Martin-Boyle, Anna and Kim, Jaehyung and Lee, Minhwa and Kim, Zae Myung and Hayati, Shirley Anugrah and Owan, Risako and Hu, Bin and Parkar, Ritik and others},
  journal = {arXiv preprint arXiv:2401.14698},
  year = {2024},
}

conference

Which Modality should I use-Text, Motif, or Image?: Understanding Graphs with Large Language Models

Debarati Das, Ishaan Gupta, Jaideep Srivastava, and 1 more author

In Findings of the Association for Computational Linguistics: NAACL 2024, May 2024

Bib PDF Website

@inproceedings{das2024modality,
  title = {Which Modality should I use-Text, Motif, or Image?: Understanding Graphs with Large Language Models},
  author = {Das, Debarati and Gupta, Ishaan and Srivastava, Jaideep and Kang, Dongyeop},
  booktitle = {Findings of the Association for Computational Linguistics: NAACL 2024},
  pages = {503--519},
  year = {2024},
}

conference

Balancing the Effect of Training Dataset Distribution of Multiple Styles for Multi-Style Text Transfer

Debarati Das, David Ma, and Dongyeop Kang

In Findings of the Association for Computational Linguistics: ACL 2023, May 2023

Bib PDF

@inproceedings{das2023balancing,
  title = {Balancing the Effect of Training Dataset Distribution of Multiple Styles for Multi-Style Text Transfer},
  author = {Das, Debarati and Ma, David and Kang, Dongyeop},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2023},
  pages = {3932--3943},
  year = {2023},
}

workshop

Adbert: An effective few shot learning framework for aligning tweets to superbowl advertisements

Debarati Das, Roopana Chenchu, Maral Abdollahi, and 2 more authors

In Proceedings of the Eighth Workshop on Noisy User-generated Text (W-NUT 2022), May 2022

Bib PDF

@inproceedings{das2022adbert,
  title = {Adbert: An effective few shot learning framework for aligning tweets to superbowl advertisements},
  author = {Das, Debarati and Chenchu, Roopana and Abdollahi, Maral and Huh, Jisu and Srivastava, Jaideep},
  booktitle = {Proceedings of the Eighth Workshop on Noisy User-generated Text (W-NUT 2022)},
  pages = {171--179},
  year = {2022},
}

journal

Influence of Consumers’ Temporary Affect on Ad Engagement: A Computational Research Approach

Xinyu Lu, Debarati Das, Jisu Huh, and 1 more author

Journal of Advertising, May 2022

Bib PDF

@article{lu2022influence,
  title = {Influence of Consumers’ Temporary Affect on Ad Engagement: A Computational Research Approach},
  author = {Lu, Xinyu and Das, Debarati and Huh, Jisu and Srivastava, Jaideep},
  journal = {Journal of Advertising},
  volume = {51},
  number = {3},
  pages = {352--368},
  year = {2022},
  publisher = {Taylor \& Francis},
}

conference

A computational analysis of Mahabharata

Debarati Das, Bhaskarjyoti Das, and Kavi Mahesh

In Proceedings of the 13th International Conference on Natural Language Processing, May 2016

Bib PDF

@inproceedings{das2016computational,
  title = {A computational analysis of Mahabharata},
  author = {Das, Debarati and Das, Bhaskarjyoti and Mahesh, Kavi},
  booktitle = {Proceedings of the 13th International Conference on Natural Language Processing},
  pages = {219--228},
  year = {2016},
}