Tackling media bias with NLP

Alarmed by the rise of fake and misleading news on social media, Zain Muhammad Mujahid seized the opportunity to tackle the problem as part of his M.Sc.

“Fake news is a ubiquitous problem now and has a toxic effect on society,” said Mujahid, who is graduating from Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) with a master’s degree in natural language processing (NLP).

“A lot of people believe anything they read on social media and have their opinions swayed by content that is either false or heavily biased and manipulative. It has a corrosive effect on people and society, so I was excited to work with professors who are leaders on the subject.”

Joining MBZUAI in 2022, Mujahid has been tutored and supervised by one of the world’s foremost experts on fake news: Professor Preslav Nakov, department chair and professor of NLP.

Focusing his research on bias demonstrated by individual media outlets, rather than trying to monitor individual stories, Mujahid explained he is profiling new media bias using LLMs.

“Each news outlet exhibits bias on various topics, like political affiliations or stances on social issues,” he said. “Leveraging 16 identifiers, I utilized pre-existing knowledge from LLMs like ChatGPT to predict media bias from media outlets directly, making it more efficient than traditional methods, such as manually collecting and analyzing thousands of articles. The support of existing language models streamlined the process.”

Mujahid aims to develop a practical application capable of alerting people to heavily biased and potentially fake news — initially focusing on bias detection, but ultimately predicting the factuality of new media. “Imagine a product where you input a media URL, and it instantly provides a comprehensive profile, highlighting biases,” he said “This tool could be invaluable for journalists and platforms to filter content based on its reliability, helping prevent the spread of misinformation.”

The research is currently geared towards English-language media, but collaborative projects with other teams are exploring a multi-language approach, covering languages such as Arabic, German, Hindi, French, Spanish, Japanese, Italian and Russian. This will help create a framework that can be adapted to different languages and cultural perspectives, enabling cross-knowledge transfer models that will enhance the accuracy of fake news detection.

Mujahid also said that he and his peers plan to incorporate elements including images and video from news articles to increase the ability of their algorithms to detect and flag media bias.

A key reason that Mujahid has been able to deep dive into complex research is the quality of teaching at MBZUAI, enhanced by the student-faculty ratio of 4:1.

“This helps with effective knowledge transfer, and I appreciated the personalized attention,” he said. “All the supervisors are easily accessible. It’s a unique advantage having everyone in one place.”

Mujahid was also impressed with the opportunities that he had to engage in various projects at the university. He contributed to the development of Jais, the world’s biggest and most accurate Arabic LLM, and was also part of the student team that represented MBZUAI at COP28.

In addition, Mujahid mentored two student NLP teams at UGRIP – a student internship program that saw 34 of the world’s most promising STEM undergraduate students visit the Abu Dhabi campus for a one-month research internship in July 2023. One of the NLP projects led to a paper being accepted at the European Chapter of the Association for Computational Linguistics in 2023.

Away from research, Mujahid appreciated the rich social life and recreational facilities at MBZUAI that helped him find a healthy work-life balance. “With students from over 40 nationalities, making friends is easy and there are opportunities to engage outside of the lab. I’m a regular visitor to the gym and enjoy cycling with friends at Abu Dhabi F1 circuit.”

He also made good use of the UAE’s location to travel to numerous countries, including Azerbaijan and Saudi Arabia, where he visited Medina and Mecca.

Looking ahead, Mujahid plans to continue in the research field and secure a Ph.D. position, either at MBZUAI or a North American university. 

“I’d like to continue finding solutions to the problem of fake and misleading content,” he said. “MBZUAI has provided me with the ideal foundations to build upon my research.”