UA Little Rock’s Nitin Agarwal Shares Expertise on how Social Media Shapes What We Believe

Every time someone likes a post, watches a video, or scrolls through social media, they are providing information that helps artificial intelligence algorithms determine what they see next.

While these personalized recommendations can make social media more engaging, they can also create opportunities for manipulation, according to Dr. Nitin Agarwal, a University of Arkansas at Little Rock researcher studying the intersection of artificial intelligence, human behavior, and online influence.

Agarwal, founding director of UA Little Rock’s Collaboratorium for Social Media and Online Behavioral Studies (COSMOS) Research Center, recently shared his expertise with Apprenticely — formerly the Arkansas Center for Data Sciences — following a landmark court ruling involving Meta and YouTube.

Agarwal, who also serves as Maulden-Entergy Chair and Donaghey Distinguished Professor of Information Science, discussed social media addiction, algorithmic recommendations, and their broader societal implications.

The conversation explored how social media platforms are designed to capture users’ attention and how those same technologies can potentially be exploited to influence what people believe and how they behave.

“I felt it was an important conversation because the case raises questions that go well beyond how much time we spend on our phones,” Agarwal said. “It gets to the fundamental issue of how AI-powered platforms are engineered to capture and retain human attention and what happens when those same mechanisms influence what people see, believe, and ultimately do.”

At COSMOS, Agarwal and his team have spent years examining the relationship between artificial intelligence, social networks, and human behavior. Their research focuses on understanding how information spreads online, how algorithms shape what users encounter, and how malicious actors can exploit those systems to manipulate public perception.

According to Agarwal, social media recommendation systems operate through a continuous feedback loop. As users interact with content, algorithms learn what captures their attention and use that information to personalize future recommendations. Those recommendations, in turn, influence how users engage with the platform.

While this process can help people discover relevant content, it can also expose them to increasingly narrow information environments and reinforce particular narratives or beliefs.

“The same AI infrastructure that can keep someone scrolling can potentially be exploited to keep someone inside a particular narrative or worldview,” Agarwal said.

This potential for manipulation is one of the central concerns driving COSMOS research in cognitive security, a field examining how information and technology can be used to influence human perceptions, beliefs, and behavior.

Agarwal explained that malicious actors do not necessarily need to compromise a computer system or take control of a social media platform to influence its users. Instead, they can exploit human psychology and the algorithms themselves.

One way malicious actors can exploit these systems is by making certain posts appear more popular than they actually are. Through networks of automated or coordinated accounts, they can artificially boost engagement, potentially increasing the likelihood that a platform recommends the material to a wider audience.

“That is why our research approaches cognitive security as a socio-computational problem,” Agarwal said. “You cannot adequately understand the threat by studying human behavior or artificial intelligence independently. You have to understand the interaction between them.”

At COSMOS, researchers are using artificial intelligence and advanced data analysis to investigate suspicious online activity and uncover patterns that may indicate organized manipulation. Their work also examines how influence campaigns develop, spread across platforms, and adapt over time.

COSMOS has received more than $30 million in research support from federal agencies, including the U.S. Army, Navy, Air Force, Defense Advanced Research Projects Agency, and National Science Foundation.

Its research has also attracted interest from NATO and defense communities in Europe, Australia, and Singapore, reflecting the growing international importance of protecting information environments from manipulation. Tools developed by Agarwal to mitigate cognitive threats have been recognized by NATO’s innovation hub as among the top 10 globally for countering cognitive warfare. 

UA Little Rock students also contribute directly to federally funded COSMOS projects addressing issues involving artificial intelligence, misinformation, and national security.

Working alongside faculty, students analyze information gathered from social media platforms and gain practical experience with technologies such as machine learning, data analytics, and large language models.

“Students are integral to virtually everything we do at COSMOS,” he said. “They are not simply observing research. They are actively participating in federally funded projects addressing real-world problems of national and international importance.”

Beyond identifying threats, Agarwal believes researchers have an important responsibility to help make artificial intelligence and social media technologies safer, more transparent, and more resilient to exploitation.

Through his participation in the Governor’s AI Center of Excellence Task Force, Agarwal has also contributed research perspectives to discussions about responsible AI adoption, government effectiveness, safety, and public trust in Arkansas.

He emphasized that addressing the risks associated with artificial intelligence requires a shared commitment from technology companies, researchers, policymakers, and the people who use these platforms. For everyday social media users, he encourages greater awareness of how online content is recommended and how it influences their emotions and decisions.

“We should periodically ask ourselves: ‘Why am I seeing this? Why does this content make me angry, anxious, or excited? Did I choose to spend this much time here, or did the system successfully keep me engaged?’” Agarwal said.

For Agarwal, understanding the influence of artificial intelligence is an important step toward ensuring these technologies are developed and used responsibly.

“AI is extraordinarily powerful, but ultimately, we determine the values and objectives around which these systems are designed,” Agarwal said. “AI doesn’t define us, it reflects us.”