On February 25, members of the San Diego Chapter of CATESOL, a professional organization serving Teachers of English to Speakers of Other Languages, met virtually to discuss the latest developments in AI technology and the impact of these systems in ESL reading and writing pedagogy. The meeting was broken into four informative sessions with time for collaborative discussion.
Session 1: How AI systems Generate Texts
AI algorithms generate texts based on the natural language processing (NLP) model to create content. The programs rely on pattern + script systems and n-grams to produce a contiguous sequence of a specific number of words (n) to fulfill the prompt. AI systems generate content formed by word associations where the primary goal is sounding fluent, not solving a problem.
Session 2: Different from Human Writing and Limitations
The primary difference between human writing and AI writing is the intent of the “writer.” Human writers focus on information that they are trying to express. AI systems focus on which words sound good in a sequence of words. When AI technology is presented with a problem to solve or a prompt that requires critical thinking, the software will sound fluent while giving hallucinated facts, or more specifically, inaccurate facts to fill in the gaps for the sequenced pattern of words. Many educators believe that “A” writing is easily distinguishable from AI Writing. The difficulty is distinguishing “B/C” level human writing from AI writing. An overarching limitation of AI technology is that there is little attention to meaningful thought expressed by fluent sentences and paragraphs.
Sessions 3 and 4: Updating our approaches to teaching
Most attendees were resigned to the idea that AI technology, specifically natural language processing (NLP) tools like ChatGPT, is fast becoming a mainstay in society. Participants were given time to collaborate and strategize on how to ethically and effectively using these technologies in our teaching pedagogy. Notable suggestions included using NLPs to create outlines, generate drafts, help with revision, and to create study plans. Discussion also centered on AI detection programs such as ChatZero, and more specially why a college-aged student felt the need to create a tool to combat the new technology. In conclusion, all members of the panel agreed that showing students why AI text generators aren’t sufficient for their needs is the best practice in teaching and learning.
