AI Voice-based Age Verification
PDPC–IMDA Innovation Challenge · ParallelChain Lab
​
Role
Product Designer (Team of 2)
​
Focus
Interaction design for AI-mediated identity verification · Trust and legibility in probabilistic systems
​
Outcome
Winner of the Singapore Government PDPC–IMDA Innovation Challenge
Interaction model informed ParallelChain’s voice-based identity verification flows
.png)
PROBLEM
The PDPC–IMDA Innovation Challenge called for privacy-preserving alternatives to traditional age verification methods such as ID uploads and facial recognition. Our team developed a voice-based age estimation prototype as a privacy-preserving alternative to document and facial verification. This introduced a new interaction challenge: enabling users to confidently complete a verification process driven by probabilistic AI inference.
​
Voice-based age estimation introduced a fundamental interaction challenge. Unlike deterministic systems, AI inference is probabilistic, non-instantaneous, and invisible to users. Without clear feedback, users cannot understand what the system is doing, whether it is functioning correctly, or how to recover from failure.
​
The core design problem was defining an interaction model that made AI-driven verification understandable, predictable, and trustworthy in a regulated context.
MY CONTRIBUTION
I led research and interaction design, focusing on making system state legible throughout the verification process. I defined the interaction patterns for recording, processing, and recovery, ensuring users could understand system behaviour and maintain progress even when inference was uncertain.
​
This work established reusable interaction patterns for integrating voice-based AI verification into identity workflows.
PROBLEM
Legal and financial professionals manage large, structured datasets spanning dozens of interdependent entities.
Bulk updates such as jurisdictions, attributes, or ownership states required repetitive point-and-click actions across many items, often taking hours and increasing error risk.
​
While AI could compress this work into a single instruction (for example, “Move all US entities to the Cayman Islands”), speed without control undermines trust in regulated environments.
KEY DESIGN DECISIONS
Make system state legible
I introduced persistent recording indicators, guidance, and explicit state transitions so users could always understand when the system was listening, processing, or awaiting input. This reduced ambiguity during an unfamiliar AI-mediated interaction.
​
Align interaction feedback with AI processing constraints
Voice inference introduced unavoidable latency. I designed countdown timers and processing indicators to align user expectations with system timing, reducing abandonment caused by perceived system failure.
​
Design verification as a recoverable process
Because AI inference is probabilistic, failures are not always definitive. I replaced terminal error states with guided retry paths, allowing users to continue without restarting and reinforcing continuity and trust.
OUTCOME
The prototype demonstrated a viable interaction model for voice-based age estimation and won the Singapore Government PDPC–IMDA Innovation Challenge. The interaction patterns informed ParallelChain’s integration of voice as a verification modality within its identity platform.
TAKEAWAY
AI systems introduce uncertainty because their decision-making process is not directly observable. Designing explicit system feedback, predictable state transitions, and recovery paths makes probabilistic systems understandable and usable. Establishing legible interaction patterns is critical to enabling adoption of AI-mediated identity workflows.