BALVOI.
AI-powered news platform that scores articles for bias and teaches readers how to spot it
Frames across the design system
Canonical breakpoints (1440 / 1024 / 768 / 375)
“Bias isn't binary — show it sentence by sentence.”
— Core design principle
Product surfaces
Reader · DeBiasIt™ · Trust Score · SoNo Training · Settings
How I worked
⌗ WHAT I OWNED
Lead design across marketing, reader, analysis tools, and every edge case.
— Overview
BalVoi™ is a reading platform with a built-in AI bias analyst. Readers paste in a news article (or pull one by URL) and BalVoi's DeBiasIt™ engine returns a sentence-by-sentence visualization of the article's bias, paired with a Trust Score that breaks down source reputation, language load, sourcing depth, and disclaimer presence. Alongside the analysis sits the SoNo Micro-Training Series — short video lessons that teach readers how news bias actually works. I led the design across the entire surface area: marketing, registration, the reader, the analysis tools, settings, subscriptions, and the moderation handling for harassment, non-English, and unintelligible content.
Frames across the system
Canonical breakpoints
Product surfaces shipped
Edge-case states designed
— Reflection
BalVoi taught me that designing for AI-augmented reading is mostly trust design. The bias score and Trust Number could've been black-box outputs. Instead I spent the most time making them legible — every score breaks down into its inputs, every sentence highlight links to its rationale, every disclaimer is visible by default. The goal was a platform that earns reader trust precisely by being transparent about its own confidence.
The other lesson was that edge cases ARE the product. Most bias tools refuse to handle harassment, non-English text, or unintelligible source articles. Designing the failure states — the explanations, the next-step suggestions, the language-roadmap disclosure — turned out to be where BalVoi's editorial voice came through most clearly.