Red Hat helps teach about tech writing

In Brno, Czechia, Red Hat's large local office has been collaborating with Masaryk University to educate future technical writers. Educating others often leads to personal learning as well. Red Hat's Brno site is situated in Brno, the second-largest Czech city and the home of Masaryk University. The office features a sizable documentation team, with one of its members participating in the project.

For the last four years, Red Hat has been working together with the university's Faculty of Informatics to provide a technical writing course unit. Jiří Herrman shared insights from the course tutors on what they learned from their students. The university's course, PV177 Technical Writing, is supported by Red Hat, which maintains a website with course materials.

Red Hat has been focusing on teaching technical writing in Brno for some time, but the course launch faced delays due to the global pandemic in 2020.

One of the most significant challenges for the course was the interaction with other humans, a crucial aspect of technical writing. The course aimed at students with some technical knowledge, but setting exercises with shell-based tools proved ineffective for some students who lacked command-line experience, had borrowed laptops without administrative access, or used only their phones instead of laptops.

Another challenge was grading students' work, which was unsuccessful due to the speed required. The team ultimately settled on teaching students how to grade each other's work without causing conflicts. Interacting with remote teachers during online presentations also presented difficulties, particularly with multiple tutors and changes in presenters for each class.

In 2023, a major hurdle for the course was the use of large language models (LLMs). Although students were instructed not to use LLMs during the course, they did so anyway. The team struggled with rewarding good work that involved LLM use, as it encouraged further cheating, and punishing students for it discouraged their use. Instead, the team taught students how to use LLMs responsibly.

They found some positive results, such as awarding points, distributing feedback through internet-style memes, and role-playing a scenario, "Dungeons & Dragons" style, to navigate a cross-functional team and create documentation for various processes. The course materials are open source, and graduates can showcase their work on a permanent FOSS record for future employers.

A "Birds of a Feather" session moderated by Herrman addressed the challenges of using LLMs in tech writing. Most attendees acknowledged using AI in their tech-writing efforts, but they were critical of the models' performance. They found that LLMs struggled with writing documentation, drawing diagrams, and reviewing them for review purposes.

However, they found uses for LLMs in change management, updating documentation for new code, documenting meeting decisions, collecting information resources, and summarizing notes. Despite the bots' poor performance in key areas, attendees trusted them for some functions that tech writers don't focus on as closely. The summary feature of LLMs, in particular, raised concerns since studies have shown that these models often shorten rather than summarize, which is not an adequate replacement for the skills of professional tech writers.