Artificial intelligence (AI)—from its potential as an academic tool to its dangers as a threat to human cognition—went on trial before a panel of George Washington University scientists, educators and philosophers at the latest installment of a conversation series hosted by the Columbian College of Arts and Sciences (CCAS).
In a wide-ranging discussion moderated by CCAS Dean Paul Wahlbeck, faculty experts explored how AI is changing the way we learn, why its output warrants skepticism and whether it can replicate—or even surpass—human intellect.
“There’s this ambivalence about whether [AI] is a tool or whether this is some…new form of consciousness,” said Associate Professor of Philosophy Tadeusz Zawidzki. “And [even] the people who are developing it are…not settled on that question.”
The Oct. 27 event, titled “AI in Teaching & Research—Helpful or Harmful?,” also examined the ethical dilemmas surrounding AI and the ways faculty are integrating it into their teaching and research.
In addition to Zawidzki, the panelists included Associate Professor of Cognitive Neuroscience Dwight Kravitz, Assistant Professor of Interior Architecture Douglas Crawford and Professor of Political Science Sarah Binder.
Wahlbeck opened the discussion by asking how society’s growing reliance on AI has affected learning and problem-solving skills. Kravitz, a member of the GW Mind-Brain Institute, observed that, while research is still in its early stages, the technology has already had “a profound impact on pretty much every aspect” of learning.
Calling it “the most powerful example of automation and knowledge indexing that we have ever created,” Kravitz outlined the central tension surrounding AI: its enormous promise versus the deep anxieties it’s unleashed. It would be foolhardy, Kravitz said, not to take advantage of AI’s capabilities. But researchers, students and the general public must also recognize its potential to “threaten the internalization of knowledge.”
“I would caution against reflexively saying, ‘No AI at all,’” Kravitz advised. But he also warned, “if in fact we use it to short circuit learning…then we have a real problem.”
AI in classrooms and labs
Throughout the discussion, panelists emphasized AI’s value as a classroom and research tool. Interior Architecture’s Crawford, for example, described how both his undergraduate and capstone students use AI to enhance renderings and develop sophisticated lighting and spatial designs.
Binder said she was “quite enthusiastic” about AI’s ability to rapidly analyze large datasets. In her research on federal debt and deficits, she leans on the technology to sift through hundreds of thousands of Congressional speeches to identify relevant references. “I can’t do that with human coding,” she said. Still, she stressed the need to scrutinize the validity of AI-generated data. “Right now [AI] methodology is the Wild West,” she said.
Both Crawford and Binder also grapple with incorporating AI into their classrooms while establishing boundaries around its use. In her classes, Binder institutes “a blanket ban on AI,” an imperfect solution, she acknowledged, that’s difficult to enforce. Crawford, meanwhile, accepts—and sometimes encourages—students’ AI use. But he requires them to distinguish their own work from AI-generated contributions. “Incorporating the technology also requires teaching students when its use is appropriate,” he stressed.
At the same time, the panelists noted that some students are becoming increasingly reluctant to use AI in class, either to protect their work’s originality or because they question the accuracy of its often “problematic results,” as Zawidzki put it. AI platforms frequently tailor their responses to flatter users and reinforce their existing preferences—a phenomenon Zawidzki called “algorithmic sycophancy.” As a result, the panel agreed, students must learn to discern when the technology can be trusted and when it warrants a healthy dose of skepticism.
“The true downside risk of AI is that it appears enormously credible—even though it tends to make mistakes. And it appears enormously certain about its responses—even though it tends to make mistakes,” Kravitz said. “If [students] are not trained to be incredulous at AI…we have not prepared them effectively.”
Claude comes to life
Zawidzki also raised an existential concern: whether AI could one day replicate human consciousness. Some of its own creators, he noted, are citing the 1950 Turing Test, warning against machines that exhibit intelligent behavior indistinguishable from a human’s. Others argue that distinction is already philosophically moot, given the existence of AI assistants like Claude, developed by Anthropic.
“If Claude considers the possibility that it’s conscious, then it can come to think of itself as conscious and act on that premise,” Zawidzki said—all the more ethically troubling given AI is not bound by the moral, political or logical norms that guide human interactions. “It’s purely transactional,” Zawidzki stressed.
Still, while acknowledging AI’s potential to diminish learning skills, the panelists doubted it would render humans obsolete in the workplace. In interior architecture, Crawford said, AI is an increasingly vital tool, but human designers still remain accountable. As long as “we have professionals creating those drawings…and putting their names on it, they’re the ones who are ultimately responsible,” he said.
Likewise, Zawidzki remains confident that, despite AI’s sometimes frightening ability to match human intellect, the “intrinsic value to learning” will endure. “AI has been able to kick our butts at chess for like 30 years now—but people still love to learn how to play chess,” he pointed out. “There’s something about human beings…that they want to know how to do stuff even if they know that there’s some machine that will always beat them at it.”