Ana

AI News Aggregator

<-- Prev

Next -->

Ana is an AI-powered news aggregator that detects media bias and promotes critical thinking through transparent analysis and educational tools, designed for Gen Z users navigating information overload.

The image featured at the top of the about us page #1

ROLE

UX Designer, PM

SKILLS

UX Design
User Research
Information Architecture

TOOLS

Figma

Dovetail

Qualtrics

TEAM

2 Researchers
2 Designers

ROLE

UX Designer, Lead PM

SKILLS

UX Design
User Research
Information Architecture

Impact

Deliverable

82% of users found AI bias explanations credible

73% said Ana improved their critical thinking about news

80% found balanced perspectives

Built with a standardized iOS component library, the app presents bias detection results clearly while encouraging exploration of multiple perspectives. Over the 10-week project, I led product management and was the primary UI designer, ensuring alignment between research insights and design execution.

PROBLEM STATEMENT

How to rebuild GenZ's trust in the News?

SOLUTION

Building the AI Agent

We chose Perplexity AI as the core of our AI because it allows custom prompt instructions. To avoid hallucinations and bias reproduction, we developed a rule‑based prompt that combines two frameworks :

5Ws for source credibility

Adapted from University of Washington, this method asks Who wrote the article, What the claim is, When it was published, Where it comes from and Why it was written.

5Ws for source credibility

Adapted from University of Washington, this method asks Who wrote the article, What the claim is, When it was published, Where it comes from and Why it was written.

5Ws for source credibility

Adapted from University of Washington, this method asks Who wrote the article, What the claim is, When it was published, Where it comes from and Why it was written.

Media Bias / Fact-Check Score

This industry standard converts qualitative cues- economic & social values, editorial language and source selection into a 7 point system and 5 star factuality rating.

Media Bias / Fact-Check Score

This industry standard converts qualitative cues- economic & social values, editorial language and source selection into a 7 point system and 5 star factuality rating.

Media Bias / Fact-Check Score

This industry standard converts qualitative cues- economic & social values, editorial language and source selection into a 7 point system and 5 star factuality rating.

To test whether Ana could outperform humans, we ran a “Media Bias Detection Sprint” in class. Participants manually examined headlines and then compared their judgements with Ana’s analysis.

The exercise showed that people were swayed by emotional language, whereas Ana systematically categorised spin and statement bias. This validated our hypothesis that a rule‑based AI can augment human intuition.

RESEARCH METHODS AND INSIGHTS

User Interviews

9 in‑depth interviews and card‑sorting sessions helped us uncover pain points such as confusion about subtle bias types and mistrust of opaque algorithms.

User Interviews

9 in‑depth interviews and card‑sorting sessions helped us uncover pain points such as confusion about subtle bias types and mistrust of opaque algorithms.

User Interviews

9 in‑depth interviews and card‑sorting sessions helped us uncover pain points such as confusion about subtle bias types and mistrust of opaque algorithms.

Surveys

A survey of 43 U.S. respondents (average age 26.7) revealed that 82% had never used an AI news aggregator and that they highly valued seeing the full story.

Surveys

A survey of 43 U.S. respondents (average age 26.7) revealed that 82% had never used an AI news aggregator and that they highly valued seeing the full story.

Surveys

A survey of 43 U.S. respondents (average age 26.7) revealed that 82% had never used an AI news aggregator and that they highly valued seeing the full story.

Affinity diagrams, personas & journey maps

We organized interview notes in FigJam to identify themes

Created three personas to ensure our design addressed diverse goals

"AI can help but I need to know when to trust it"

Laura Lefty

"AI can help but I need to know when to trust it"

Laura Lefty

"AI can help but I need to know when to trust it"

Laura Lefty

"it's exhausting sorting through headlines to find unbiased facts"

Neutral Nial

"it's exhausting sorting through headlines to find unbiased facts"

Neutral Nial

"it's exhausting sorting through headlines to find unbiased facts"

Neutral Nial

"I want news thats relevant to me!"

Richard Righty

"I want news thats relevant to me!"

Richard Righty

"I want news thats relevant to me!"

Richard Righty

DESIGN PROCESS

We started with a user‑flow diagram to explore tasks - Article Bias Analysis, News For You, Onboarding and Discussions

Used information architecture techniques to determine the navigation. Used RITE Tests to iterate prototypes rapidly.

Tree Testing

To determine the priority of order of the information we used tree tests

Tree Testing

To determine the priority of order of the information we used tree tests

Tree Testing

To determine the priority of order of the information we used tree tests

RITE Tests

Rapid Iteration Testing and Evaluation based on real human behavior

RITE Tests

Rapid Iteration Testing and Evaluation based on real human behavior

RITE Tests

Rapid Iteration Testing and Evaluation based on real human behavior

The user first starts of by uploading an article for ANA to summarize

Our custom prompt instructs the model to summarize the article, answer the 5 Ws, calculate MBFC scores and produce a transparent report

The generated article has the headline, the reporter and their company for credibility. Users can read all the articles by that particular reported by simply clicking on their name.

The report is then sorted into sections for easy consumption.

Added Bias Slider for users
to understand how different sources
report the same news

Factuality Rating and details about
the reporter are easily visible
right under the headline

Users can share their thoughts on the Discussion board. A small indicator highlights the alignment of the reference article to reduce the friction in opinions.

The Journey map showed that “evaluation and trust building” was the most critical stage, prompting us to prioritise transparency and education in the design.

To enhance transparency, the methodology section explains how ANA was trained.
Micro-lessons are a glossary that users can refer to understand keywords.

Key Learnings

Design System Efficiency

By adopting a standardized iOS component library the team was able to resolve inconsistencies, enhance usability, and create a polished, mobile-first experience

Design System Efficiency

By adopting a standardized iOS component library the team was able to resolve inconsistencies, enhance usability, and create a polished, mobile-first experience

Design System Efficiency

By adopting a standardized iOS component library the team was able to resolve inconsistencies, enhance usability, and create a polished, mobile-first experience

Product Strategy

Final design emphasized transparency, on-boarding tutorials, bias reports, and interactive lessons to build trust and help users interpret the AI’s reasoning.

Product Strategy

Final design emphasized transparency, on-boarding tutorials, bias reports, and interactive lessons to build trust and help users interpret the AI’s reasoning.

Product Strategy

Final design emphasized transparency, on-boarding tutorials, bias reports, and interactive lessons to build trust and help users interpret the AI’s reasoning.

Iterative Design

By iterating through lo-fi and mid-fi prototypes, incorporating peer feedback, and conducting RITE, the team continuously refined the product’s usability and features.

Iterative Design

By iterating through lo-fi and mid-fi prototypes, incorporating peer feedback, and conducting RITE, the team continuously refined the product’s usability and features.

Iterative Design

By iterating through lo-fi and mid-fi prototypes, incorporating peer feedback, and conducting RITE, the team continuously refined the product’s usability and features.