![Annette Vee](https://www.humanitiescareers.pitt.edu/sites/default/files/Annette_Vee_wide_1.jpg)
The course Automated Writing from Amanuenses to AI grows out of my research interest in natural language generation and processing (NLG/NLP) and how automated language tools are shaping our contemporary writing landscape. Any time we go online, we are engaging with automated writing systems (AWS): our Google docs and tweets contribute to datasets of natural language to be used for our email and text autocompletes; we click on bot-written headlines or stock reports; and infamously, our online political discourse is flooded with comments that are generated or amplified by algorithms. Writing studies has traditionally focused on people and how we write, circulate texts, and read. But anyone training in writing studies now has to contend with the fact that we write among algorithms. I designed the Automated Writing course to introduce graduate students to some technical research on NLP/NLG and prompt critical questions surrounding AWS in writing and rhetoric/composition studies.
I had already spent a lot of time thinking about the intersections of literacy, writing, and computation in my research on computer programming as a literacy (“Understanding Computer Programming as a Literacy” article; Coding Literacy book) and had explored how bots were contributing to the circulation of political information with a friend and colleague at Lafayette College, Tim Laquintano (“How Automated Writing Systems Affect the Circulation of Political Information Online”). In 2019, there were major headlines about GPT-2 and discussions about how much better automated writing systems had gotten because of developments in AI, specifically recurrent neural networks (RNNs). I gave a couple of talks at University of Connecticut and University of Michigan that helped me to think through the implications of algorithmic writing for contemporary literacy.
Post-tenure, I had begun to return to my interests and background in K-12 education, and I was able to connect to my coding literacy research to K-12 education when Luciana Correa, a visiting scholar from Brazil, came to work with me in Fall 2019. Together, we ran a reading group focused on teaching coding as a literacy in elementary education, a Coding Carnival at Pitt’s Center for Creativity, and the Children’s MuseumLab. I wanted to work through ideas about automated writing systems in the company of some smart grad students, as well as encourage them to consider some exciting new research areas and public engagements in writing studies. The Humanities Engage invitation to design a course that would bridge disciplines and reach out to the public came at a perfect time.
I used the grant this summer to explore more of the technical details about how natural language processing works and some applications for its use. Applications in journalism are particularly interesting, in part because NLP is so widely used in the profession now. NLP helps journalists dig stories out of massive datasets, streamline workflows, fact-check, automate breaking news analysis, and even generate stories in constrained genres like sports reporting (Underwood, “Automating Journalism…”). I’m still no expert in machine learning, but I learned a lot from reading about automated journalism; Janelle Shane’s hilarious book You Look Like a Thing and I Love You (titled after a computationally-generated pickup line); websites such as Towards Data Science; and the Electronic Literature Organization virtual conference in July. I was already familiar with some Twitterbot learning tools such as Cheap Bots Done Quick and colleagues at the ELO conference shared teaching strategies for NLG tools like Tracery. I was also able to connect with Prof. Mike Sell, who developed a Digital Storygame Project at IUP that works with regional high school teachers to integrate Twine and digital literacy into English classes. Along with colleague Rachel Schiera (now at Lander University), they work collaboratively with teachers to help them meet their teaching goals and curriculum standards. I found their responsive and respectful approach a great model for how to frame the public-facing work that students might do in this course. Finally, the events over the summer that have prompted a worldwide reckoning with racism helped me to consider connections between the impulses to automate writing now and a longer history of inequality and power dynamics in labor. The gendered division of work in the office early 20th century, the use of amanuenses to tell and shape the narratives of enslaved people in the 19th century, and the impulses to mechanize human labor in the 18th century are all threads in the history of automated writing.
Automated Writing is a historical and technical dive into why people have developed automated writing systems (AWS), what challenges AWS offer, and how to implement AWS using natural language processing and public data sets. The course brings gendered and racialized histories of office automation and amanuenses for the writing-down of narratives from enslaved people in conversation with contemporary questions in artificial intelligence, such as whose intelligence is being simulated and how. In the course, we will explore: what writing is; power dynamics in writing; the limits of what computers can do; and the relationship of human consciousness to computation. Hands-on work in AWS-related systems include basic programming, Tracery (using Javascript) and InferKit (using GPT-2), Twitterbots, Conway's game of life, and the Leibniz cipher machine held by Pitt ULS.
Assignments include reading histories of writing automata such as the Maillardet Automaton, writing short blog posts analyzing different historical and contemporary AWS or their products, and playing with algorithms and datasets such as GPT-2. The final project for the course asks students to break the boundaries of the class: develop a unit to teach AWS in Pitt undergrad English courses like Composing Digital Media or Narrative and Technology, an online creative project using automated writing, a public art project, or a community workshop. My hope is that graduate students in the course can strengthen their technical skills and research agendas and then pass on their knowledge by developing courses or public engagement events and workshops.
Automated Writing from Amanuenses to AI Syllabus (PDF)
Learn about all the courses faculty developed with Faculty Summer Stipends for Curricular Innovation.