My current research interests lie at the intersection of Artificial Intelligence, Education, and the Learning Sciences. I’m particularly interested in how LLMs can support learners in reasoning through complex ideas, not by giving answers, but by guiding reflection, encouraging reasoning, and deepening understanding.

My past work spans Requirements Engineering, Legal AI, and Human-Computer Interaction, where I explored how LLMs can detect and resolve ambiguity, predict legal judgments, and how thoughtful design choices can improve collaboration between humans and AI systems in decision-making contexts.

Ongoing Work

1. Detecting and Resolving Pragmatic Ambiguities in Natural Language Requirements Using Multi-Level Domain Knowledge Bases

Natural language requirements are essential for bridging communication gaps between diverse stakeholders in software development. However, some requirements may be misinterpreted due to varying contextual knowledge and domain-specific expectations of the stakeholders, a phenomenon known as pragmatic ambiguity. Our framework uses LLMs and multiple domain knowledge bases to simulate stakeholders of varying expertise and detect and resolve pragmatic ambiguities.
Authors: Pavithra PM Nair, Preethu Rose Anish
Keywords: Pragmatic Ambiguity, Requirements Engineering, LLMs, Retrieval-Augmented Generation
Status: Under review
We present Vichara, a novel framework tailored to the Indian judicial system that predicts and explains legal judgments. Vichara decomposes case proceeding documents into decision points. Decision points are discrete legal determinations that encapsulate the legal issue, deciding authority, outcome, reasoning, and temporal context. The structured representation isolates the core determinations and their context, enabling accurate predictions and interpretable explanations. Vichara's explanations follow a structured format inspired by the IRAC (Issue-Rule-Application-Conclusion) framework, and adapted for Indian legal reasoning. This enhances interpretability, allowing legal professionals to assess the soundness of predictions efficiently.
Authors: Pavithra PM Nair, Preethu Rose Anish
Keywords: Legal NLP, Explainability, Judicial Decision Modeling
Status: Under review

3. COMPASS: An Organization-centric Framework for Compliance and Penalty Assessment Using Large Language Models

Compliance with laws and regulations is increasingly complex for MNCs which must manage shifting legal obligations across jurisdictions, industries, and operational contexts. Manual interpretation of dense statutory documents is resource-intensive, error-prone and often disconnected from operational needs, creating risk of compliance failures and penalties. We introduce COMPASS, a framework for automating the extraction of deontic modalities from statutory documents tailored to an organization's operational profile, transforming them into compliance tasks, and deriving associated elements, including penalties.
Authors: Gokul Rejithkumar, Sachin Pawar, Pavithra PM Nair, Nitin Ramrakhiyani, Preethu Rose Anish
Keywords: Organization-centric compliance automation, Penalty assesment, LLMs
Status: Under review

4. Hey Sruthi!: A Co-Reflective Chatbot for Surfacing Latent Knowledge in Community-Development Fieldwork Group Chats

Community-development programs often struggle to capture and use the situated knowledge of the fieldworkers carrying them out. While fieldworkers regularly share updates, these are often superficial, overly sanitized, or lost in informal channels. Consequently, field knowledge rarely informs decision-making, and interventions risk overlooking on-the-ground realities. We introduce Sruthi, an LLM-powered, peer-mode co-reflector embedded in multi-party fieldworker group chats. Drawing on Murphy’s framework of collaborative learning and Mezirow’s theory of transformative learning, Sruthi encourages participants to expand updates, surface dilemmas, and build on each other’s insights.
Authors: Srividya Sheshadri, Manju Balakrishna Pillai, Apoorv Gupta, Balu Mohandas Menon, Unnikrishnan Radhakrishnan, Aswathi Padmavilochanan, Megha TM, Sooraj S Nair, Pavithra PM Nair, Bhavani Rao R
Keywords: Chatbot, LLM-powered, Community Development, Implementation Science, CSCW, ICT4D
Status: Under review

Published Work

For the most recent list, check out my Google Scholar.

1. To Solicit or Not to Solicit? Impact of AI Assistance Delivery Mechanisms on Decision-Making

Pavithra PM Nair, Gilad Gressel, Malavika Nambrath Anand, Krishnashree Achuthan
International Journal of Human–Computer Interaction. 2025
[Code]

2. An Integration of Indian Philosophy and Machine Learning for Human-like Cognition

Akhbar Sha, Anvita Reddy Inture, Jose Joseph, Pavithra PM Nair, MJ Sheeba, T Anjali
International Conference on Automation, Computing and Renewable Systems (ICACRS). 2023