Multi-Perspective AI Research Agent System

A qualitative research group led by Sabrina Lin conducts field-based observational studies - specifically TCS (Therapeutic Communication Studies) - that require analyzing human behavior through multiple philosophical and analytical frameworks simultaneously.

The research is rigorous by nature. A single observation sample needs to be examined through sensory, structural, cultural, and poetic lenses to produce valid findings. With large datasets and tight timelines, doing this manually meant either slow output or compromised depth.

A qualitative research group (managed by Sabrina Lin) needed a way to analyze complex qualitative samples (TCS observations) without the inherent bias of a single human or AI perspective.

Qualitative research requires looking at the same data through multiple lenses-sensory, structural, cultural, and poetic. Manually coordinating these viewpoints across large datasets was slow and prone to 'tunnel vision'.

• Subjective Bias: Research analysis limited to one viewpoint
• High Complexity: Coordinating four distinct philosophical frameworks
• Scaling Friction: Analyzing thousands of field observations manually
• Data Inconsistency: Difficulty in quantifying 'tension' or variance between observations

Deep Loom developed a 'Parallel Curator' architecture using n8n and Claude 4.5. This system processes a single research input through four specialized AI agents simultaneously.

The workflow reconciles these diverse perspectives into a unified 'APR' (Average Perspective Rating) and automatically flags data points with high variance for human review.

• The Phenomenologist (B1): Analyzes lived experience and embodied sensation
• The Structural Objectivist (B2): Focuses on geometry, spatial logic, and verifiable structure
• The Cultural Contextualist (B3): Decodes social symbols and cultural performance
• The Poetic Synthesist (B4): Identifies metaphor, resonance, and thematic ambiguity
• Automated Synthesis: A centralized processing hub that calculates APR Tension and archives results to Notion

The Parallel Curator system processes a single observation through four philosophical lenses simultaneously - Phenomenological, Structural, Cultural, and Poetic - and synthesizes them into a single APR score in seconds. Analysis that would require hours of manual multi-perspective review now runs automatically.

The tension metric is the most operationally useful output. A high APR tension score means the four curators disagreed significantly - which is exactly the flag a researcher needs to know where to spend their limited human review time. Low-tension samples archive automatically. High-tension samples surface for deeper examination. The system decides what deserves human attention.

Manual data entry into Notion is fully eliminated. Every observation has a complete, structured audit trail: all four curator outputs, the calculated APR, the tension score, and the archival timestamp. The research team's time shifted entirely to synthesis and interpretation - the work that requires human judgment.