Engineering trust in product reviews with human-AI synergy
Amazon reviews are critical for buyer decisions and seller success, but their credibility is often lacking. In 2022, Amazon detected 200 million fake reviews, and even advanced AI systems misclassify them, creating distrust.
As the design lead for a 24-hour Industry Project with Amazon, I improved review authenticity while prioritizing user privacy. I guided engineers and data analysts to extend Amazon Vine, integrating human feedback with LLMs.
Achievements
Won 1st place for designing the best solution to ensure trust and transparency across the platform.
Role
UX Designer
Skills
Product Design
Problem Solving
Branding
Prototyping
Timeline
September 2024 (24 hours)
Reimagined Amazon Vine to go beyond products
We discovered Amazon’s existing program where verified users review products. Our goal was to expand it beyond collecting reviews—to also categorize feedback based on relevance and quality. To encourage participation, we proposed rewarding Vine Voices with Amazon credits, incentivizing them to evaluate the helpfulness of other reviews.
A familiar, two-step review process to flag helpful and…not so helpful reviews
Lacking access to Vine Voice, I designed the dashboard around Amazon’s review process, making it familiar and intuitive. This enabled users to strengthen LLMs by flagging helpful reviews and providing reasons, ensuring better categorization and filtering of malicious content.
No bots allowed, verified humans only
While Amazon allows anyone to browse without logging in, verified users bring trust and credibility to the platform. These users are authenticated through multi-factor authentication, ensuring they’re real humans. We added a layer of reliability and transparency to the review system by encouraging them to not only write comments but to evaluate reviews as 'helpful' or 'not helpful.'
With only 24 hours and no access to Amazon Vine, we had to start with a broad approach and move quickly. I relied on the existing Amazon review process to deduce user needs and made rapid design decisions, collaborating closely with developers to balance user goals with technical feasibility.
Design
Isanna Wong
Yasin Mete
Data
Somayyeh Gholizadeh
Crystal Gil Herrera
Nicole Hixon
Engineering
Sung Kim
Linh Pham
Angelo Yap
Cybersecurity
Isabelle Jaber