Mission: Teaching Media Literacy

From social media posts to podcasts, a public benefit corporation out of Colorado has made it its mission to help the public filter through misleading and inaccurate information.

Translated to “to the source,” Ad Fontes Media uses its team of around 60 analysts, as well as machine learning (ML) capabilities, to rate online content using the company’s content analysis methodology. “We take a nonpartisan approach in mitigating our own biases,” said Vanessa Otero, founder and CEO of Ad Fontes. “We have folks from center, right and left political perspectives, and they’re diverse across age, race, gender and personal characteristics.”

The methodology adheres to the Media Bias Chart, which follows a scoring system based on bias and value/reliability. The chart seen in analyses rates bias on the x-axis from left to right—the more conservative, the further to the right the content will appear. Whereas, the y-axis shows the news value and reliability portion, from the lowest point of inaccurate and fabricated information to the highest point of thorough factual reporting or fact-dense analysis.

Created in-house, the Content and Analysis Rating Tool software pulls in statistics from uploaded content, such as the number of nouns, verbs, adjectives and adverbs. However, the analysts input the actual score by looking at headline/graphic; expressions; underlying veracity; language; political position, etc.

All the content is input into the ML model, which also includes custom features in Ad Fontes’ algorithm, including conventional machine learning and language processing. “We’re working towards implementing large language model-based analyses of the texts as well,” Otero shared.

“Over the years, we’ve rated about 80,000 individual pieces of content manually,” she stated, explaining that the information gathered was resourceful in training their ML model.
 

Artificial intelligence (AI), however, will never completely replace human capability, Otero said. The question of reliability and trust is far too important, she explained. “Even now, our humans rate articles every day and then our model rates those same articles and then we check to see how accurate our model is over time.”

Continuous updates to article analyses will only make the model more accurate. To ensure the management of content at a larger scale, Otero stated that 99% of the content Ad Fontes rates goes through AI.

And the content is not limited to the United States, with over 100 international news sources rated by the company. Otero also plans to branch out to international locations within the next 12 to 18 months. “The biggest target countries and regions are areas in Europe, Brazil and India,” she stated. “[Brazil and India are] such large democracies with big populations; and Brazil has its own low-reliability information source problems, as does India.”

Otero’s mission is to be instrumental in helping humans navigate the learning curve of the new information age.

“Just like you can teach literacy, you can teach media literacy,” she said.