Cloze: An Open Research Platform for Studying Human-AI Conversations in Mental Health Contexts

2026-06-13 · Source: Computation and Language · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Data Science & Analytics · Depth: Advanced, quick

Summary

Cloze is an open-source web platform designed for conducting controlled, monitored studies of human-AI conversation within mental health research contexts. Unlike consumer LLM products such as ChatGPT, Claude, or Gemini, Cloze provides researchers with critical experimental control, consistent data export, and a shared safety framework. The platform allows teams to configure specific AI models, instruct the AI, schedule conversations over time, and apply unconditional safety constraints. It captures every message with full provenance, including model version and prompt configuration. Cloze supports OpenAI, Anthropic, Google, and locally hosted open-weight models via Ollama, offering deployment in the cloud or on-premises to ensure participant data security. It serves as essential research infrastructure, not a therapeutic product.

Key takeaway

For research scientists studying human-AI interaction in mental health, Cloze offers a robust solution to overcome the limitations of consumer LLMs. You should consider adopting this open-source platform to gain precise experimental control, ensure consistent data capture, and implement shared safety protocols. This enables rigorous, ethical research by allowing you to configure models, instructions, and deployment environments (cloud or on-premises) while maintaining full data provenance.

Key insights

Cloze provides a controlled, safe, and data-rich environment for human-AI conversation research in mental health.

Principles

• Experimental control is crucial for human-AI interaction studies.
• Shared safety scaffolding is essential for sensitive research.
• Full data provenance enhances research validity.

Method

Researchers configure AI models, instructions, conversation schedules, and safety constraints within a unified platform, capturing all message data with provenance.

In practice

• Conduct controlled studies with various LLMs (OpenAI, Anthropic, Google).
• Deploy research infrastructure on-premises for data privacy.
• Capture detailed message provenance for analysis.

Topics

• Cloze Platform
• Human-AI Interaction
• Mental Health Research
• Large Language Models
• Experimental Control
• Data Provenance

Best for: AI Scientist, Research Scientist

Related on AIssential

• We Need to Know More About How AI is Affecting Mental Health — A critical knowledge gap exists regarding AI's mental health impacts, demanding urgent research and regulation amidst widespread chatbot adoption.
• MHSafeEval: Role-Aware Interaction-Level Evaluation of Mental Health Safety in Large Language Models — Evaluating LLM mental health safety requires role-aware, multi-turn interaction analysis to uncover cumulative harms.
• AISN #72: Empirical Research Sheds Light on AI Wellbeing — CAIS research explores "functional wellbeing" in LLMs, revealing preferences and model-specific "feelings" that impact AI alignment.
• Hugging Face's Clem Delangue on Open Source AI and the LLM Bubble | MTS Live — Open-source AI fosters innovation and safety by empowering broad access and defensive capabilities, despite "bubble" concerns in API-based LLMs.
• Imperfectly Cooperative Human-AI Interactions: Comparing the Impacts of Human and AI Attributes in Simulated and User Studies — AI transparency significantly impacts human trust and satisfaction in imperfectly cooperative interactions, often more than human personality.
• Unregulated chatbots are putting lives at risk | Letters — Unregulated AI chatbots pose significant mental health risks due to a lack of pre-use screening for vulnerable users.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computation and Language.