Identifying Misinformation with AI


The way we consume news has changed significantly over the last 10 years, and that comes with some significant problems: thanks to smartphones and social media, we are overloaded with information and exposed to more news than ever, and, due to the emergence and escalation of misinformation & ‘fake news’, we don’t know if we can trust the news we read anymore.

Simmer is a news companion service that helps users quickly consume and understand online news, and identify & avoid misinformation, through on-demand breakdowns of news articles generated using machine learning technology. Users can generate these breakdowns through the mobile app or browser extension, containing analysis of the headline (scoring the headline based on how misleading it is), summary of the article content, a list of sources cited in the article, a set of similar stories from credible stories, and explanations of key topics in the article. Users can view a feed of pre-generated breakdowns of popular stories, and explore various trending news topics, via the mobile and web apps.


Angus Williams – Computer Science with Innovation MEng


I’d become increasingly aware of and concerned about misinformation, sensationalism & bias in news and digital media over the last year, and, as a computer scientist interested in potential of machine learning to solve human problems, I wanted to investigate how technology could help contribute to solving the problem of misinformation and create value for users around news consumption.


Having started with a focus on misleading statistics, my research led me towards the problem of misinformation in general, and issues in news consumption around social media and sensationalism. I used systems thinking to explore the ways in which misinformation can spread. As this project utilised a design thinking process, I also conducted user research to understand specific habits people have around news consumption, and would later test UI mockups of Simmer with potential users. One of the key insight I gained early on was that objectively determining the truth is incredibly difficult, hence the focus on breaking down the article and providing as much information as possible in a concise manner to help a user assess the veracity of a story. I also learned about the particular nuances of how younger generations consume news: how ‘assumed knowledge’ used, and how many young people use search engines as their main way of finding out more about news stories and topics, which formed the motivation for the news explainer aspect of Simmer.


In the age of malicious fake news and sensational headlines that exaggerate events for clicks, there is a need for people to be more critical of what they read in the news and online in general. The average 21-35yr old sees over 35 headlines every day, and global trust in news media is rapidly decreasing, from 51% in 2015 to 28% in 2020. No one has the time to read all the news they see, let alone research and fact check information for themselves. High quality journalism still exists, but young people consume around 60% of their news passively, by bumping into it, eg. on social media, where misleading and false information can spread rapidly.


This idea could help people become more rational in how they consume news, providing a “sanity check” to help users look past the emotional reactions that headlines are designed to trigger, and understand what’s actually going on in the news and why they should care. I’m currently building prototypes, with the potential for a basic version of Simmer to be released for free as a civic tool, or potentially pursuing investment as a startup, developing the system for B2C & B2B purposes.

For more information visit: