Photo by Beth Stevenson on Unsplash
Project Overview: Engineering Social Behavior Inside Strict Compliance Boundaries
At Meta, I designed the social engagement system that made a four-hour, fifty-person simulated mixer actually work. Capturing realistic audio data required a real social crowd: overlapping speech, movement, and genuine conversational behavior. To achieve this, our team built a simulated cocktail party at an external vendor site, complete with prototype hardware sprinkled throughout. The core challenge was that every standard social crutch was stripped away: participants were complete strangers, used code names, were strictly prohibited from sharing personal identifying details (like jobs or hometowns), had no shared background, and were bound by a rigid, IRB-approved prompt list. Left unchecked, a room of fifty restrained strangers would fall silent—jeopardizing the data collection entirely.
My Role: Lead Research Assistant & Experience Designer
While the protocol, hardware, and analysis belonged to other specialists, none of it produced data if fifty strangers stood around in silence. I owned the problem of how to generate natural social mixing inside strict compliance constraints. My responsibilities included:
- Designing the end-to-end participant experience and engagement strategy across multiple collections.
- Translating rigid IRB parameters into creative, compliant social mechanics.
- Running parallel technical execution, including hardware setup, device management, piloting, and live troubleshooting.
Methodology & Execution: Compliance as a Creative Constraint
Instead of fighting our regulatory boundaries or proposing elaborate formats like murder mystery parties (which would have triggered costly protocol amendments), I used compliance as a design brief:
- IRB-Compliant Human Bingo: I worked backwards from approved conversation prompts to curate 221 concrete, findable traits (e.g., “Can juggle successfully”). I designed 25 unique sheets so only two participants held identical cards, forcing independent circulation across the entire room.
- Themed Visual Design: Using Canva, I created seasonal, professional bingo sheets that immediately shifted participant mindsets from an intimidating technical trial to a welcoming event.
- Multi-Gear Pacing for Four-Hour Sessions: A single activity decays over four hours. I layered high-circulation bingo with seated origami stations and custom-printed Jenga blocks featuring approved prompts for self-facilitated group discussions.
- Technical & Operational Execution: Managed equipment setup, software prep, device management for a high volume of hardware, and live troubleshooting during sessions.
Key Findings: Lessons in Behavioral Design
This project yielded critical insights into managing human behavior within strict system limitations:
- Constraints are a design brief, not a blocker: Prohibiting personal information didn’t kill conversation; it just shifted it toward skills, preferences, and experiences—which happened to match our protocol.
- Engagement needs more than one gear: Alternating between high-circulation, seated, and group-discussion formats successfully held the room for four hours.
- Aesthetics affect data: The themed sheets were cheap to produce but had an outsized effect on participant willingness to engage.
- Low-tech beats high-tech: Paper, blocks, and origami outperformed anything requiring scripting, adding zero failure points to a hardware-heavy environment.
Outcome & Strategic Impact: Reusable Standards & Usable Data
The engagement strategy transformed a high-risk social environment into a successful, high-yield data collection:
- Usable Crowd Dataset: Kept fifty strangers engaged, talking, and circulating across the full four-hour session, turning potential silence into a rich, high-fidelity acoustic dataset.
- New Organizational Standard: The human bingo and multi-tier engagement framework became the default standard for subsequent large-group collections and behavioral studies.
- Efficiency and Compliance: Proved that creative, low-tech interventions could be stood up faster and more safely than elaborate scripted formats, establishing a reusable method rather than a one-off fix.
