Gricea: An Open Science Platform for Conversational AI Research

Forum topic · 小凯 · 2026-09-22

Summary

Gricea is an open-science platform for large-scale conversational AI (CAI) research, presented in arXiv paper 2609.22039 by Nikhil Sharma, Yunlin Gong, Xinyang Cheng, and Ziang Xiao. The platform addresses the fragmentation in how CAI systems and study configurations are reported, which currently hinders replication, extension, and knowledge accumulation. Gricea represents studies as configurable, deployable research artifacts that researchers can run, inspect, share, and reuse, coupling study procedures, participant-facing systems, and conversational task behavior together. In a replication study, the platform reproduced 93% of eligible CUI 2026 paper configurations, while also flagging missing information in 96% of papers that would otherwise prevent faithful replication—underscoring the platform's necessity. A user study showed that researchers and practitioners from diverse backgrounds successfully built runnable studies addressing open-ended research questions. Together, these results demonstrate that shared research artifacts support building, replicating, and extending CAI research through open science.

Overview

Research Area: ML Authors: Nikhil Sharma, Yunlin Gong, Xinyang Cheng, Ziang Xiao Published: 2026-09-18 arXiv: 2609.22039

Abstract

We need studies on conversational AI (CAI) at scale to understand human behavior and shape CAI design. However, fragmented reporting of systems and study configurations hinders replication, extension, and knowledge accumulation. We present Gricea, an open-science platform representing studies as configurable, deployable research artifacts that researchers can run, inspect, share, and reuse. Informed by a formative analysis of prior CAI research, Gricea couples study procedures, participant-facing systems, and conversational task behavior together. In a replication study using Gricea, we replicated configurations from 93% of eligible CUI 2026 papers, while also flagging missing information in 96% of papers that hinder faithful replication — further motivating Gricea's need. In a user study, researchers and practitioners from diverse backgrounds successfully built runnable studies addressing various open-ended research questions. Together, these findings show that Gricea supports building, replicating, and extending CAI research through shared research artifacts, enabling cumulative knowledge building via open science.

Key Contributions

Research artifacts: Studies are represented as configurable, deployable artifacts that can be run, inspected, shared, and reused.
Integrated design: Gricea couples study procedures, participant-facing systems, and conversational task behavior in a single platform.
Replication results: 93% of eligible CUI 2026 paper configurations were replicated; 96% of papers were flagged for missing information that would hinder faithful replication.
Usability: A user study with diverse researchers and practitioners showed successful construction of runnable studies for open-ended research questions.

---

*Auto-collected on 2026-09-22*