When you speak with the faculty and students in the Center for National Security Studies (CNSS) at Utah Valley University, there’s at least one thing they all have in common: They want to make a difference.

“My entire life, I've had a real interest in tech and how things work,” said Brandon Amacher, director of the CNSS Emerging Tech Policy Lab. “But simultaneously, I also have always had a really deep-seated desire to do something that I felt like was making a difference.”

That desire, coupled with the efforts of a committed CNSS student team and collaboration from the Gary R. Herbert Institute of Public Policy and the UVU Sales Marketing Applied Research Test Laboratory (SMARTLab), led to the publication of important research on the impact of artificial intelligence “deepfakes” on elections and political advertising. The team presented its findings at a news conference in October.

In working on the study, Amacher and his students united over their belief in the importance of their work and the enduring strength of the United States as a nation.

“Personally, intrinsically, I believe in what our country is supposed to be, even if it's not quite there right now,” said Hope Fager, who led the UVU student team on the deepfake study. “So I very much believe in the work that we're doing; this place is our home, right? And we can either roll over, or we can stand up and make it better.”

Amacher, who has a professional background in security technology and cyber intelligence, said he hadn’t planned on becoming an academic. However, he heard about UVU’s national security studies programs through contacts in the national security industry, and he jumped at the opportunity to help students tackle high-impact, real-world problems.

The election deepfake project became the perfect opportunity to leverage his experience, he said.

“Something that I really love about UVU, and one of the reasons why I was so enticed to come work here, is that UVU, across departments, has a very strong focus on practical learning,” Amacher said. “I think sometimes academia might have a tendency to get a little myopic because everybody is in their own silo researching their own topic. But I think that focus on engaged learning meant that synergy was just kind of there. It’s just plug-and-play. We were all ready to go.”

“Between CNSS, the Herbert Institute, and the SMARTLab, UVU had all the necessary pieces to put together a project like this,” Amacher said. He also pointed out how eager the CNSS students were to jump in and collaborate, even learning skills and software they had no prior experience with.

The student team had to learn how to train an AI model, write scripts for the audio and video to be produced, film real versions to compare to the fakes and use software to generate the actual images and sounds.

“Before we started any of this, I had a little bit of experience with AI, but I had to train a model from scratch,” Fager said. “So that's not necessarily something that I had done before, and I had to learn how to do that very quickly.”

Student interns from the Herbert Institute and SMARTLab also provided enthusiastic contributions to the collaborative project.

“This team was really willing to do whatever it took,” CNSS team member Leah Olsen said. “We had one night where it was 9 pm, and we had an issue with the software. Everyone came in and sat here for a few hours, just editing tiny points on the software. And we got it done.”

Once the deepfakes were created, the team used the facilities in UVU’s SMARTLab to track test subjects’ reactions to each piece of content. While biometric data collected during the study showed higher levels of engagement and confusion with the deepfake content, suggesting a potential “uncanny valley” effect, over 50% of participants rated deepfake content as “probably real” or “definitely real.”

In effect, the CNSS team had created completely falsified audio and video clips that were equally as believable as the real thing.

“If the entire objective of this AI is to deceive you, and you put that in the hands of somebody that knows how to use it, then one of these days, it's gonna happen,” Fager said. “We're going to get to a point where there aren't going to be any tells in the video anymore.”

Amacher said that, at that point, it’s important to think critically about what emotional response each piece of content is trying to provoke and why.

“The best thing that I could really tell people is, if something sounds way too good or way too bad to be true — if it's making you way too happy or way too angry — that is something you want to really double check and verify,” Amacher said. “And then, I think we need to change some of our foundational habits about media consumption, particularly social media consumption.”

“There's so much critical thinking that's going to need to go into media consumption,” Fager said. “You can't just turn off your brain and scroll through Instagram.”

“For myself, moving forward, I'm definitely going to be trying to step up my game of keeping myself informed and my loved ones and friends and everyone informed to the best that I can,” Olsen said.

Beyond the technical and critical-thinking skills the CNSS students acquired in working on this project, Amacher said it will give them a huge leg up as they move into careers in national security.

“As somebody who's hired new graduates and undergraduate students in my past jobs, if I had seen something like this [project] on somebody's resume, it would tell me so much about them,” he said. “It immediately shows me that they're thinking deeply about this. They're working hard on these issues. They know how to communicate this properly. It’s honestly difficult to overstate.”

“I think that skill, to be able to talk about AI and understand it, and then to communicate to other people in the profession, or consumers, or policymakers, or whoever it may be — I think that that is very valuable,” CNSS team member Lacie Jacobson said.