top of page
Make your own decisions
Chrome extension to understand online content better
My Role in the startup
Product Designer (Remote)
Timeline
June 2019 - September 2020
Overview
Factmata finally released the most awaiting TrustedNews
extension long after it acquired from Adblock Plus's Eyeo.
One year of a journey gone into iterating this product both
from tech and design perspective. I was fortunate enough
to be part of this and was responsible for the final designs of the product.
Worked with a bunch of experienced and talented people
Dhruv Ghulati
CEO
Sameh Frihat
Data Scientist
Katia Stambolieva
Product Manager
Startup
in partnership with
Hasan Halacli
Web Developer
*NDA is signed and only the data that can be displayed in public is mentioned here.
Trusted News is the #4 Product of the Day on ProductHunt
Context
Context
TrustedNews uses AI to assist newsreaders in evaluating the quality of the online content they read. In its first release, it scores the objectivity for a selected article, testing whether it is written from a neutral perspective as opposed to a personal one.
Download and Install the TrustedNews chrome extension
Go to the article or webpage for which you want to check the objectivity score
Open the extension and it scores Objectivity Score
Context end
History
History of TrustedNews
Factmata acquired the TrustedNews chrome extension from Eyeo. Their product allowed users to check the trustworthiness of the news sources they were reading online through a Chrome extension. By acquiring the product, and Factmata aim at improving and extending the functionality of TrustedNews by potentially making it a valuable tool for building a community of users.
History ended
Eyeo's TrustedNews
Eyeo's TrustedNews interface
It used a tool from MetaCert to power the app, a database of information that’s used to provide a steer on the bias.
TrustedNews Extension by Eyeo
Eyeo's end
Plans and goals
Plans and goals made to improve TrustedNews
It’s hard to tell what’s real news and what’s fake these days, especially on the internet. TrustedNews aims to help users understand the news you are reading.​
-
The plan is to build extensions for other browsers.
-
To make it more interactive so that the users can get involved in the process of rating articles → debating how an article is written with others and engaging in rating systems, rather than just being handed a rating to trust.
-
One step closer to a safer, more transparent internet.
Plans end
Insights
Insights, Reviews found from Eyeo's TrustedNews
Google Chrome Store Web Stats
Users > 10,618
27 Reviews
Overall Rating
Main Issues in Trusted News
Lack of trust in company
Lack of transparency
Lack of user empowerement
Lack of context
Reviews Found on multiple sites
Negative
"Liberal sites are marked as trustworthy while conservatives are untrustworthy: made people wonder about the political opinions of the people running it"
Negative
"It would be better to educate people on how to recognize and judge the veracity of news rather than serving them the prepared product"
Negative
"I would prefer that a news aggregator would show all sides of a story and let me choose what to believe"
Negative
"Delegating to someone to decide what is right and what is wrong is essentially wrong. I wouldn’t use it."
Negative
"I don’t know the criteria that it uses in order to judge the content as trustworthy or not. And the rating depends a lot on definitions"
Negative
"Not Trustworthy enough. The main problems are that it does NOT rely on users’ feedback and that it doesn’t provide any specific information on why a certain piece of content might be biased: eg does not state political leanings on CNN. Rating system is not transparent."
Insights end
Survey
Survey was carried out to understand users
The survey has been carried out to understand how people behave around the online news and what kind of behaviour they exhibit while consuming their news and try to understand what can be done to ease the understanding for them.
Some Q/A from the survey
Conclusions are drawn out
from the survey
-
Under Factmata, TrustedNews needs to preserve its core functionality of advising users about the credibility/trustworthiness of the content they are reading.
-
Factmata needs to work towards extending the coverage of TrustedNews to the whole web while providing information about WHY content has been labelled in a certain manner (eg: what makes this content biased and how did Factmata evaluated it?).
-
People want to contribute to the spotting of misleading content online by tagging different sentences. This is strongly in favour of the Overlay format.
-
There is no urgent need for building an IOS/Android app- most people consume news on the web. However, it would be preferable to have Trustednews working on all formats (chrome extension + website + app). For now, this means focusing on the format website + chrome extension (Option B in mockups).
-
People are interested in the community feature of TrustedNews. They want it to be a place where to find trustworthy content while sharing their opinions with others and working towards the mission of creating a better internet.
Survey end
Interview
1:1 interviews are hosted with heavy news consumers
Several questions were questioned to the users who are mostly heavy news consumers/social justice warriors. All of these questions try to solve the three main questionnaire we had in mind. We took 8 interviews and we tried to answer these questions with their help.
-
How do they react to hateful content and misinformation online?
-
What do they do currently / what problems do they face?
-
What would motivate them to annotate, using our proposed solution?
Few Q/A from the interviews
Social / News: on what devices?
-
On my cellphone.
-
Usually on my phone, I’m browsing on the train.
-
Mostly on mobile: it’s for social. Computer is for work.
-
On my phone. Laptop is work, phone is for brain breaks.
-
On my phone.
-
Mostly on my phone.
-
On mobile.
-
On phone.
How does it make you feel?
-
It’s been happening so long, I’m desensitised. “Here we go again”
-
It’s upsetting. Kindness matters. There’s enough hate in this world.
-
I see people I know who are smarter than that, who subscribe to dumb views. These are influential people.
-
I’m disappointed. Platforms should control content, but their algorithms reward content reaction, not the quality or source.
-
It’s frustrating. Especially when it’s people I’ve grown up with.
-
Words have consequences. Views shape how we treat each other.
-
It’s particularly a problem if the country is unstable, like the UK is now.
-
I see so many right-wing views now – it’s very different to 5 years ago, more like 1930s Germany.
-
I don’t want to engage, I end up not feeling good.
-
I wish I never saw, or engaged with, the bad post.
What’s the biggest problem with that?
-
We’re not going to change other people’s minds.
-
Retaliating makes things worse.
-
Some people don’t think they are doing anything wrong.
-
Fear of the consequences, for me.
-
I’ve lost faith in Facebook’s ability to actually do anything about the problem.
-
It’s never worth getting into a Facebook argument, better to move on.
-
I don’t want to affect a friendship, even if their views are antithetical to my own.
-
I can’t change someone’s biased belief.
-
I’m Jewish, but I can’t convince someone to not be anti-semitic.
-
It’s too difficult, doing a Facebook Report.
-
There are trolls, willingly spreading misinformation.
Interview End
User Personas
User Personas are defined to uncover behaviors
Personas are made to build a habit-forming product around the people by understanding why they want to use the product and what are the main frustrations they are facing which in turn not allowing them to use the product at ease.
Persona1
Persona2
Persona4
Persona1
1/4
UP end
Lowfi
Low-fidelity mockups designed after brainstorming
Concept A - Gamification
Max is an online opinion AI agent and he is on a mission to make the internet a better place.
Max Searching
Max Talking
Max Cannot Read
Max Found Results
Concept B - No Gamification
There are no gamification elements like Max here. This is a simple, plain version of how TrustedNews scores the content.
Lofi end
User Testing Report on above two concepts
We undertook some initial user testing of designs for the TrustedNews extension. We showed 2 versions of the extension in either one of two conditions to 5 users. The versionswere; A – Gamified version including ‘Max’ character design, B – Official version. Theconditions were 1 – Prompting user ratings (Active), 2 – No prompting for user rating (Passive). 3 users viewed version A and B under condition 1 and 2 users viewed version Aand B under condition 2.
testin
User Profile
-
None of the users currently used a plug-in to review the news they were reading
-
All (n=5) were interested in using such a plug-in
-
All were highly concerned about the quality of the news they were reading (rated
4.4/5) -
They were quite likely to use a plug-in to review news articles and sources (rated
3.38/5) -
They were neutral about rating individual articles or websites themselves (both
rated 2.78/5)
Results - Comparison
-
3 users preferred the gamified version, compared to 2 who preferred the official
version -
The plug-in rated as best at achieving its stated aim was the gamified version with
active user rating (Version 2, Condition A; rated 4/5). -
Both versions with active user rating were seen as better at achieving their aim than
no user rating (3.5/5 vs. 2.5/5, n.b. rated between groups, so rater bias maybe
present) -
The official version was seen as more efficient than the gamified version (4.2/5 vs.
3.2/5) -
They were seen as equally fun (both 3.4/5)
-
The official version was slightly more understandable (4.2/5 vs. 4/5)
Design Issues
-
Most commonly cited issues were; a lack of clarity about what was happening in the system and, an unfamiliar design which users didn’t recognise
-
Other issues included; the inappropriate tone of voice (for gamified version), Lack of user freedom to do what they wished (for gamified version) and a busy design (for official version)
Conclusion
Whilst users showed a narrow preference for the gamified version of the plug-in, it is clearfrom its lower ratings for efficiency, fun and understandability and a wider range of designissues that this version needs some further design work to deliver a great all-roundexperience for users. This was a limited study and clear differences between either versionor condition were not found to suggest discarding any one direction at this stage. However,it has suggested paths to improve the design across any one of the 4 design options (2versions x 2 conditions).
testig end
final
Final designs after iteration and their features
After so many iterations of designs and user testing, we came up with a simple yet precise design that users are aware of and we are able to finish with good user testing report on this.
Scoring the article
With one click on Trusted-News icon, we analyze any text inside the current tab (if it exists), and provide you with a score in less than 5 seconds that helps you caveat what you are about to read.
​
Objectivity is defined as when content is written from a neutral rather than a personal perspective. Phrases like “in my opinion” or “I think” are used by authors to reflect their individual thoughts, beliefs, and attitudes.
View article with similiar score
Click on the link and view an other article with similar score.
Overall URL Objectivity
We also score the overall objectivity for the URL the article present in.
Feedback on the score
Feedback from users is collected to choose whether the article is objective or not which in turn helps our models to score better.
Phrases are highlighted
The sentences which are more likely to be objective are highlighted in yellow colour.
​
It uses a unique machine learning model trained by data science experts at DataRhine, which gives each sentence a score - 0 if the sentence is subjective, 1 if its objective. It is trained on sentence-level using a rich set of features including opinion bearing phrases (phrases authors use to reflect their individual thoughts, beliefs and attitudes) and other semantic markers.
TrustedNews helps with fighting Covid-19
In the current version of TrustedNews, you can navigate to a section about Covid-19. Once there, you can help us by answering some simple questions about the current article you are on, to increase the accuracy of this upcoming model. These are:
-
Does this text contain Covid-19 medical misinformation?
-
Does this text contain Covid-19 general misinformation?
​
Our next release - due shortly - will contain a unique artificial intelligence model that will alert you to articles that contain medical and general Covid-19 misinformation.
Covid-19 Medical Misinformation
Content suggesting a course of action that could lead to harmful health-related beliefs or behaviours (such as incorrect claims about cures and treatments)
Content contradicting or discrediting the official government guidelines on health and safety (such as ‘social distancing recommendations’)
Other types of misinformation around Covid-19
Content stating facts on Covid-19 that have not been proven by documentation, research, or been verified by reliable and publicly available resources (such as ‘conspiracy theories’)
Content that singles out the blame for the pandemic on individuals, institutions or racial groups
Error is there is no text or if the page is not accessible
Today, the browser extension only works best on long-form articles e.g. Boris Johnson’s US trade deal will make Britain a paradise for disaster capitalists
For now, it does not work on social media pages, tweets or profiles. (Although we’d like to do this). It works best when you are on a tab with a specific piece of content, which doesn’t require infinite scrolling.
final end
TrustedNews in action
We finally made it very simple for the users to just visualize what the score looks like. We also provided
extra features for those who seek more opinionated content on the web.
video
Some design elements are added later for the tech convenience
Video end
next
What is coming next for TrustedNews ?
Soon, TrustedNews will be providing more labels such as Political bias, Propaganda, Clickbait, Racism,
Sexism, Toxicity.
​
Want to know more how the TrustedNews work?
Visit
next end
closing
Closing Notes
We were happy that TrustedNews is declared as #4 on Product Hunt. Users are interested in this product
and they are sending feedback at regular intervals. I as an individual is very happy to contribute to this
and want to thank my CEO Dhruv for making this possible.
​
Stay tuned with this product, there will be many more upcoming features added to this.
closing end
Have a look at other recent works
Coming Soon
Be safe, avoid fake - news
A publicly available dashboard to regularly process, track, and analyze COVID-19 related disinformation and misinformation.
Coming Soon
Play , Learn and Win
One of its kind chess app where players can
find different ways to compete in the chess world
Coming Soon
bottom of page