Designing for Social Collaboration and Human-AI Alignment

Abstract

As digital communication increasingly shifts toward AI-mediated environments, maintaining emotional nuance and social alignment remains a critical challenge. While communication tools like emojis have traditionally bridged the gap left by absent face-to-face signals, the integration of Large Language Models (LLMs) introduces a new paradigm of interaction, changing how social cues are interpreted and expressed. The main objectives of this research are to address the gap between LLM understanding and human value-sensitivity through three key areas: (a) analyzing how existing tools like emojis shape emotional clarity, (b) designing a chatbot to facilitate social connection during initial interactions between strangers, and (c) developing a collaborative system for stakeholders to co-create evaluation criteria for "LLM-as-a-judge" workflows. By focusing on these interventions, my work seeks to ensure that the evolution of AI-mediated collaboration preserves the essential nuances of human values and intentional social alignment.