Dean’s Blog: Liberal Arts or STEM? It’s the Wrong Question–Ask the Jefferson Avatar

Image

Disclaimer: My own education which follows a long and winding road through the physical sciences, the life sciences, the computational sciences, and science policy equips me with a STEM biased view. My subsequent foray into data science these past 10 years has at least made me a T-shaped learner: a vertical in which I have years of STEM expertise and a broad horizontal which has enlightened me to the humanities and other fields.

I am forever reading articles that pitch liberal arts against science, technology, engineering, and mathematics (STEM) as if they are in competition. Clearly, many inside and outside of higher education see it that way, and the discussion seems more fervent than ever as higher education is under a period of greater public and federal scrutiny. 

There are many nuances to the liberal arts vs. STEM debate, but when cast in the age of AI, a liberal arts education gives a humanistic view of the world from which to start a career while a STEM education gives a mechanistic view of the world from which to start a career. Proponents of each side of the debate, will argue for their path, assuming they have achieved some level of success.

With some exceptions, higher education fosters the divide through a singular educational pathway already begun in high school. Those exceptions within university programs across the country have names like “Science, Technology and Society,” “Digital Humanities,” “Computational Media,” “Integrated Science and Humanities,” and “Human Centered Design and Engineering.” While these programs are to be commended, are they enough in an era where what it even means to be human is being questioned? I think not.

Data science as a discipline offers opportunities and challenges in the liberal arts vs. STEM debate. While data science is typically taught as if data were about genome sequences, climate data, energy consumption, traffic patterns, etc., perhaps different lines of questioning should be applied. Humanities students analyze a poem, play, or novel—texts that are fundamentally their own kind of data sets—in order to extract meaning about the human condition. Why shouldn’t data scientists adopt a similar outlook? 

My own assessment, perhaps naive, is that most of those doing data science are trained in gradient descent and linear algebra, not the works of Shakespeare or Molière. But imagine the insights if they were. In addition to improving lives through practical solutions and informed policy, the conventional domains of data science, might our students feel empowered to ask and answer the big questions about human nature, existence, and truth?  

What if students at a university were enrolled in both a literature course and a data science course? ould we do more to help them see how these fields might inform and activate one another? Rather than treating one as a general education requirement and the other as a professional pathway, we have an opportunity to encourage deeper integration—where interpretive skills and technical expertise work together to generate insights that are both analytically rigorous and profoundly human. 

The opportunity for data science, in collaboration with all other disciplines, is to create a higher level of understanding and awareness of the intersection of STEM and the liberal arts and train students appropriately. There are multiple challenges. We cannot do it alone, and traditional university organizational structures make such training difficult to nearly impossible. At UVA, we have created an environment and academic programs within our school to do better, but it is not easy. I know why and I have seen it before, albeit with a narrower focus: 

Years ago, Calton Pu and I wrote the first grant at Columbia University between a life scientist and a computer scientist. After the grant was awarded—the only grant I was ever awarded that gave more money than we asked for—I remember a meeting Calton and I had with our students talking about structures. We were 10 minutes into the discussion before I realized he was talking about data structures and I was talking about molecular structures. There was a semantic and cultural divide based on our prior training. Calton and I thought the future lay in an interdisciplinary field of biology and computation. History since those early nineties has proven us right, but that was just two disciplines. Fast forward to today where students are taught by faculty who speak both languages. I believe the future of data science and the humanities lies in being adept at both while still having the opportunity to specialize in one or the other. This requires a T-shaped approach to scholarship. To do this requires a merger, or at least closer cooperation, between multiple disciplines. Think of it as “lateral arts.” We are not there yet.  

The perception, nature of scholarship, and reward structures are different between disciplines and difficult to reconcile. I know this from evaluating faculty at the UVA School of Data Science each year. We have faculty who are anthropologists, economists, historians, computer scientists, environmentalists, systems engineers, biomedical researchers and more, each with their own scholarly expectations. It is a challenge to evaluate them under one umbrella since those performing the evaluation, myself included, come to the evaluation through the lens of our own scholarship. It is a credit to our faculty that we work it out together, through appreciation of the work of others. I see this as vital.

Increasingly the opportunities that exist through a closer cooperation between STEM and the liberal arts will be realized, as they were in computational biology. The needs of society, and the recognition by students and employers of that need will drive universities to change. Here is one example of that opportunity from the work of colleagues and myself: the Thomas Jefferson Avatar.

Avatars of historical figures either as textual responses to an AI bot, or including a visual component and possibly an immersive component are rapidly appearing and getting better as the supporting technologies improve. Where will we end up? It won’t be driven by the most realistic avatar, but by which avatar responds with the greatest historical accuracy. Special effects in the movies already tell us this. The whiz bang of special effects alone soon gives way to the best story told enhanced with special effects and usually produced at less cost. History will repeat itself. On the flip side, students of history and related fields from the Alpha Generation (b. 2010-) will want to learn the field by talking to a trusted avatar. Visitors to a historical site will want to talk to the people who lived there. Visitors to art galleries and museums will want to talk to the people who created the works on display. None of these scenarios can occur in the best possible way without close collaboration between STEM and liberal arts experts. Experts in each, but with a knowledge and passion for each other’s work. 

As we dig into the Jefferson Avatar project, the importance of cross-disciplinary training becomes clear. Jefferson was a complex figure whose words, actions, and beliefs evolved over his 83-year life. Capturing that nuance in a lifelike conversation with him at a specific moment in time is no small task. It requires:

A comprehensive review and digitization of all available primary works and related criticism. 

• Temporal mapping of Jefferson’s life: when, where, and with whom he interacted.
• Input from architectural historians and stewards of Monticello, Jefferson’s historic home, and other historical sites.
• Accurate visual representation of Jefferson himself, supported by art historians.
• Immersive environmental design, involving visual artists and 3D-rendering specialists.
• An understanding of the ethical, cultural, and philosophical implications of digitally reconstructing historical figures.

In other words, this is an unprecedented collaboration that will engage historians, librarians, data scientists, speech experts, instructional designers, ethicists, philosophers, and more. But the goal is not simply to build a technically sophisticated avatar. It is to translate vast, complex bodies of knowledge into a human-centered experience, fostering meaningful engagement with history, inviting critical reflection, and deepening our understanding of humanity. 

You get the picture. A true collaboration across the STEM-liberal arts divide for which no one is yet suitably trained. We have much to do and this is only one example of the collaborative future that awaits us.

Acknowledgements: Thanks to Jason Nabi, the lead on the Jefferson Avatar project and author of the UVA University Seminar “All Bots Created Equal: Building an AI Jefferson for his insights. Jason has a Ph.D. in English Literature. Thanks also to Emma Candelier for her insights.