Rationalize: Shared Semantic Reasoning for Human-AI Alignment
Aritra Dasgupta, Naga Datha Saikiran Battula, Avina Nakarmi, Sohom Sen, Subhodeep Ghosh, Xun Song
Why It Matters
What makes this one worth your time
This framework could facilitate better collaboration between humans and AI, making AI systems more interpretable and aligned with human intentions.
Rationalize proposes a role-pair framework for improving human-AI alignment through shared reasoning.
Summary
The paper introduces a framework called Rationalize, which conceptualizes human-AI interaction through complementary role pairs to enhance shared semantic reasoning and alignment between humans and AI models.
Key contributions
- Introduction of the Rationalize framework for shared semantic reasoning.
- Identification of complementary role pairs to enhance human-AI interaction.
- Development of a collaborative research agenda for alignment design and assessment.
Notable insights
- The role-pair framework emphasizes the importance of explicit communication of purposes and assumptions in human-AI interactions.
- The distinction between aligning AI to humans and aligning humans to AI based on roles is a nuanced approach to understanding collaboration.
Possible limitations
- Not stated in the abstract.
Abstract
arXiv:2605.30632v1 Announce Type: cross Abstract: We introduce Rationalize, a role-pair framework for shared semantic reasoning between humans and AI models in data-driven sensemaking. Building on ideas in human-machine teaming and critical thinking, we conceptualize human-AI interaction as a series of complementary role pairs (Explorer-Guide, Investigator-Informant, Teacher-Student, Judge-Advocate) operating in a shared reasoning space. In this space, human analysts and AI models (such as LLMs) make purposes, questions, assumptions, evidence, inferences, and implications explicit, facilitating alignment not only at the output level but at the level of rationalization of intent and action by each side. We relate these role pairs to the bidirectional human-AI alignment framework, illustrating how "aligning AI to humans" and "aligning humans to AI" differ by role, and sketch a collaborative research agenda for alignment design and assessment using element-level and role-specific approaches.