Building off of ideas sparked in Bethel’s Advanced Digital Humanities course, Zach Haala ’23 and Associate Professor of History Charlie Goldberg are conducting a research project to use artificial intelligence to probe historical archives for patterns. They are one of the student-faculty research teams to receive a 2021-22 Edgren Scholarship.
By Jason Schoonover ’09, content specialist
November 19, 2021 | 9:30 a.m.
Associate Professor of History Charlie Goldberg and Zach Haala ’23 are one of the student-faculty research teams to receive a 2021-22 Edgren Scholarship for their project, "A Picture's Worth a Thousand Data Points? AI-driven Machine Learning in Digital Humanities Analyses."
Zach Haala ’23 and Professor of History Charlie Goldberg noticed an anomaly in their data. Using artificial intelligence (AI), the two had tracked the presence of smiles over nearly 80 years and thousands of Bethel photographs. As expected, smiles grew more prevalent in the photos over time, matching cultural shifts after World War II. But then in the 1960s, the number of smiles decreased. At first, they were stumped. Then Haala noticed that’s when the yearbooks started featuring more sports photos. “It’s a good example of how data spits stuff out, but data needs to be interpreted,” Goldberg says. In the 1960s, male athletes rarely smiled in photos, and large teams like men’s football and basketball affected the research results. To Goldberg, it shows the promise of using AI to explore history and also raises questions. “What do we then do with this stuff? How do we interpret it and use it to tell a human story—which is what historians do?” Goldberg asks.
Those are the kinds of questions Goldberg and Haala explored in the research project, "A Picture's Worth a Thousand Data Points? AI-driven Machine Learning in Digital Humanities Analyses." They were one of the 2021-22 student-faculty teams to receive an Edgren Scholarship to support their research.
To some, history may feel a long way from artificial intelligence and programming. But Goldberg also directs Bethel’s digital humanities program, which explores cutting-edge, forward-looking methods to apply to history, literature, and philosophy. While teaching Advanced Digital Humanities last year, Goldberg got the idea to use AI to study history. The class explores advances in AI technology and how it’s often a double-edged sword—it yields many opportunities with data and research, but it also leads to things like deepfakes—or fake photos or video created using AI, often depicting world leaders or celebrities. Goldberg wanted to go deeper. As a historian, he uses data—usually text or photos—to look for patterns. He was interested in using AI to isolate the same patterns historians explore but on a larger scale, and wanted to see how well AI could recognize the same patterns in photos that historians look for. He knew he needed a student who was highly skilled at coding and programming—and who was also willing to dive into the deep end and take risks. Enter Haala, who is triple majoring in computer science: software project management, software engineering, and digital humanities—and he had taken Advanced Digital Humanities.
For their Edgren Scholarship research, Associate Professor of History Charlie Goldberg and Zach Haala ’23 built on concepts they first explored in Goldberg’s Advanced Digital Humanities course. They are using artificial intelligence (AI) to probe historical photos for patterns.
For a sample data set, Goldberg and Haala looked to Bethel’s digital library collections and the annual yearbook that Bethel published from the early 1900s into the 1980s. By using various programs and facial recognition technology, Haala extracted and sorted all the photographs from the scanned yearbook pages. Goldberg noted it can be easy to run a test once, but it’s much more difficult to run multiple tests using 20,000 photos and then keep all those photographs organized by year. “This is why big data is art,” Goldberg says.
Haala came up with systems for organizing and tracking the photos. Once they had a process for scanning and sorting their data, they could run individual tests more easily. From there, they trained their AI to recognize things in photos. Eventually, results are converted to numbers and expressed in a CSV file in Excel. “It’s a long process,” Haala admits. However, he was surprised how accessible it was to undertake this kind of research—you just need a laptop, some coding experience, and time. Haala loved combining his interest in computers and software with his interest in the past. “I also love history,” Haala says. “I wasn’t just here to be a lab rat and do the coding that he needs. It was an enjoyable experience to be able to see this stuff and be a historian for a little bit.” Eventually, they were able to make charts to visualize their numbers, helping them explore how their results matched with history.
Zach Haala ’23 and Associate Professor of History Charlie Goldberg are blending history and cutting-edge technology in their research project, "A Picture's Worth a Thousand Data Points? AI-driven Machine Learning in Digital Humanities Analyses." Using artificial intelligence (AI), they probed historical Bethel photos to see how patterns line up with historical data.
Using AI, Haala and Goldberg ran experiments to track not only the prevalence of smiles in photos, but things like gender, emotion, age, and ethnicity. Aside from the blip in the 1960s, they successfully tracked the prevalence of smiles. Other tests successfully tracked changes in age and gender in the photographs, again matching historical trends. While college students trended to be younger after World War II, the results also matched Bethel’s transition from a seminary to a four-year liberal arts college. As for gender, the ratio between men and women grew more even in the photos over time. Bethel predominantly taught men early in the 20th century, but more women sought opportunities in higher education after World War II and then during the feminist movement in the 1960s and 1970s. Though the technology showed a good deal of promise, it wasn’t perfect. The results were limited in their ability to track things like changes in racial diversity over time, illustrating a larger problem in asking AI to investigate race in an ethical way.
The duo still hopes to conduct another experiment to track the prevalence of objects like jewelry, neckties, and glasses through the years. They’re set to present their findings to the Bethel community in February. To Goldberg, their findings show potential at Bethel and beyond. Not only can historians use this technology, but Goldberg hopes to one day have his digital humanities students design their own experiments to look for patterns. This reflects the goal of digital humanities at Bethel, to take concepts and procedures that people associate with multi-million or billion industries and do them with students on a shoestring budget, whether it’s AI, video-game making, or 3D modeling. Goldberg also plans to write an article for The Programming Historian, a website dedicated to the merging of history and technology. Haala said he could easily envision using this kind of technology in his career, as object detection and facial recognition software are being used more and more for apps and beyond.
Study the digital humanities at Bethel.
Bethel offers a B.A. in Digital Humanities, where students collaboratively learn how to use digital technologies to ask and answer meaningful questions in the study of English, history, literature, philosophy, theology, and other subjects, and to communicate persuasively and winsomely with their audiences.
Share this story: