Ankana Saha
Apple, Responsible AI and Human Evaluation
I’m an AIML Resident at Apple. I recently earned my PhD in Linguistics from Harvard University with a dissertation examining the division of labor between semantics and pragmatics in the referential domain, with a particular focus on demonstrative expressions, under the guidance of Gennaro Chierchia, Kate Davidson, and Veneeta Dayal. My research lies at the intersection of semantics and pragmatics, investigating how utterances acquire meaning, how meaning interacts with linguistic structure and discourse context, and how these processes can be modeled in both human and artificial systems. I’m particularly interested in using insights from linguistic and experimental data to better inform our understanding of language models’ capabilities and limitations, with a growing focus on AI ethics and bias. You can find out more about my research here.
🗞️ Recent News
🎤 Invited talk — Discourse-Pragmatic Sensitivity in Large Language Models at the Experimental Study of Meaning Lab, University of Pennsylvania, February 3, 2026.
🎓 Dissertation defense — I defended my PhD dissertation, titled Demonstratives and the Semantics–Pragmatics Interface on December 8, 2025.
🎤 Invited talk — From Humans to Language Models: Testing the Boundaries of Pragmatic Understanding at the Harvard Language & Cognition Seminar, November 11, 2025.
🎤 Invited talk — Bridging Modalities: Tracking Reference in Discourse Across Spoken and Sign Language at the Linguistic Meaning Lab, Cornell University, October 10, 2025.
📄 New paper — The or That? Evaluating Language Models’ Sensitivity to Discourse Structure in Anaphora (with Jennifer Hu and Kathryn Davidson) at the Conference on Language Modelling (COLM) 2025 PragLM Workshop. [pdf]
📄 New paper — Demonstrative Descriptions and Anti-Uniqueness at Sinn und Bedeutung (SuB) 29. [pdf]
📄 New paper — Anaphoric Demonstratives in Mandarin (with Yağmur Sağ, Jian Cui, and Kathryn Davidson) at Semantics and Linguistic Theory (SALT) 34. [pdf]