Project Info
001
Ongoing
Role
UX Designer
Team
Zhiyuan Chen
Wesley Deng ()
Jason Hong ()
Methods
Focus Group
Company Info
002
Who is WeAudit?
Overview
003
Background:
Bias Beneath the AI Hype
AI is part of our everyday lives, from virtual assistants to customer service. While it’s praised for solving problems and improving efficiency, its hidden flaws, like bias, are often overlooked. Bias in AI can damage trust, lead to bad decisions, and create legal risks for businesses. On a broader level, it can reinforce stereotypes or exclude marginalized groups, worsening existing inequalities.
78%
of executives believe data bias will become a bigger problem as AI/ML use increases
13%
of AI specialists rated their AI projects as very fair
36%
of organizations surveyed reported suffering from AI bias, with 62% experiencing lost revenue and 61% losing customers as a result
Problem:
AI teams struggle to scaffold end user bias audits
AI practitioners in the industry have their cultural blind spots when auditing biases in AI systems. While they are highly motivated to involve end users in the auditing process, they lack a platform for end users' auditing process and struggle with providing the necessary context, support, and scaffolding to empower end users to identify and analyze biases meaningfully and effectively.
Solution:
Empowering Industry AI Product Team with an auditing platform
We created WeAudit, a crowdsourced platform that empowers users to identify, discuss, and report bias in AI-generated content. By offering a centralized space, WeAudit educates users on bias, fosters meaningful discussions, and encourages actionable insights to mitigate the negative impacts of AI bias.
Impact:
Endorsements from Tech Giants
WeAudit demo is presented to many AI teams from Google, Apple, Meta, Microsoft and have received positive feedback. Many expressed an interest in adopting similar features to improve their own AI systems, validating WeAudit’s approach to addressing bias.
Challenge before I join
004
Siloed Team, Disconnected solutions
Our team has many ideas in supporting users to audit bias in AI, and each idea is prototyped into an individual product. Each team runs in silos and they are very heads down on developing the product and expanding capabilities and creating new products.
Product 1️⃣
Product 2️⃣
Product 3️⃣
There are many more products…
Frustrating user experience and overwhelmed AI teams
Having different products pose many challenges for AI practitioners. First of all, the disconnected experience can be frustrating and discourage end users from using them. Second, all the reports and data are also separate, which creates additional effort for AI practitioners to aggregate and analyze to obtain meaningful insights.
Define MVP Scope
007
Stepping up to create a clear and focused MVP
I took the initiative to refocus the team by stating, “We need to stop adding more features and define what truly matters for an MVP.” To make this happen, I first consolidated all standalone products into a single platform. Then, I prioritized key features and deprioritized others, ensuring the MVP was both streamlined and impactful.
Finding the overlap
Each product came with its own set of features, which made integration complex. To simplify, I created a union set of all features across the products and identified the overlapping ones, which became the foundation of our MVP.
Choosing what matters the most
There were many features outside the union set that I evaluated using an impact-value matrix. I assessed each feature by balancing its value, technical feasibility, and usability for users to determine whether it should be included in the MVP.
