Final Project: AI and Literacy in Education

An Exploration on the Impact of Artificial Intelligence on Literacy

Human knowledge has always been built on through our ability as a species to pass on our knowledge through written word. Every great societal change has begun with an invention or innovation that has improved literacy. From using papyrus, the printing press and ultimately the computer and smartphone. The next stage for societal development was to be led by the use of artificial intelligence, yet it seems to have had the opposite impact on human development. Artificial Intelligence (AI) began far before the first modern computer with the purpose to support and assist human ingenuity. More recently, OpenAI’s GPT software was built on a language learning model with its own limitations in written text and understanding our language and a society as whole. This paper will investigate the role of OpenAI’s ChatGPT, one of the largest and most common AI tools in use by students, on literacy and education in secondary and postsecondary school.

Artificial Intelligence can be traced back to the first analytical engine in the 19th century but we will start with the most significant contributor to AI with Alan Turing in the 20th century (Gryzbowski et al., 2024). By creating the first predictive machine learning device and creating the idea of the Turing Test, Alan Turing was able to open the door for a world where machines could become interchangeable with a human mind. The first introductions to these possibilities began in 1966 with the utilization of the first Large Language models (LLM) in the ELIZA program by Joseph Weizanbaum that simply would identify keywords and reflect back a question for the user (Foote, 2023). However this program and those to follow were limited with simple responses, only with the advent of the world wide web did LLM finally have the opportunity to access a vast reservoir of human language in order to learn. The first machine learning programs based on LLM were the Turning Natural Language Generation by Microsoft and the GPT-3 by OpenAI which eventually debuted in 2022 and has become one of the largest and most frequently used AI tool in Canada with 77% of postsecondary students reporting using ChatGPT the past year (Janzen et al., 2024). These modern devices are all trained on all the available resources online before they are released and the grammar, vocabulary, and human syntax continues to evade even the most advanced ChatGPT model. Students are especially susceptible to falling for AI writing as they often believe advanced vocabulary demonstrates high literacy. Furthermore, with our ever changing world the amount of knowledge being created changes yearly and with the limited GPT models being trained using only the available resources at the time of its inception, it will always be out of date and require consistent training and updating. Investments that require significant time, financial, and environmental investment in order to create and run these GPUs to train these new models. 

Academic writing has been a key part of the sciences and humanities in both secondary and postsecondary school. Often students have difficulty in researching and writing their academic papers yet ChatGPT can be a useful technological tool for all students. In a study of postsecondary students one group was to use ChatGPT to write their paper and one without, and Bašić et al. (2023) found that “Chatgpt could accelerate writing in the first phase” (p.4). Often the first phase of writing a paper proves the most time-consuming and difficult place to start as students need to have a literature review. By providing ChatGPT a few prompts it can create a list of relevant studies and types of literature to review for the topic. Furthermore, it has shown to be an effective tool for creating an abstract outline, introduction and background information, and suggesting methodology for a study (Mondal & Mondal, 2023). However the most challenging aspect for ChatGPT is its inability to provide sources for the information it writes or creates. Oftentimes the AI will create fictitious sources or ‘hallucinate’ scenarios. 

In Figure 1, Chatgpt was asked two simple prompts as inspired by Dogra (2023)  first a factual claim where it was correct that the English Channel is not crossable by foot, then with a single falsified claim which resulted in ChatGPT creating a scenario that was entirely false and even ‘hallucinated’ an entire story. These hallucinations and fabrications are common limitations of a language learning model as the system predicts what should come next in a chain of text rather than understanding authenticity and truth. A study on ChatGPT generated reference lists found that “47% were fabricated, 46% were authentic but inaccurate, and only 7% were authentic and accurate” (Emsley, 2023, p.1) The inability of ChatGPT to find sources for its own claims and generate references that are not authentic will hinder its use in Education. 

Figure 1 (OpenAI, 2025)

ChatGPT can transform language learning for students as a form of self-assessment, learning, and by tailoring learning to each individual student’s learning goals. Language teachers must first teach their students how to notice bias and hallucinations formed by ChatGPT by having them cross reference any queries they question with real sources. Furthermore, most sources that ChatGPT was trained on were written in English, with 45% of all training material being English and the remainder spread around other languages with none going above 5% (Howarth, 2024). The bias towards English means that the usefulness of ChatGPT as a learning tool for literacy for language learners and teachers is mostly restricted to English acquisition. Moreover, the greatest strength demonstrated by ChatGPT is its ability to be used as a lexical reservoir. The ability for language learners to quickly search up a definition of a word or a phrase rather than using the traditional dictionary has shown a significant increase and effectiveness for learners (Lew et al., 2024). Especially in creating new sentences for language speakers. But just like new learners, idioms and grammatical syntax were equally difficult for comprehension for language learners no matter if they were using a dictionary or ChatGPT as learners lacked the grammatical understanding to put the definitions together. Language teachers can prompt ChatGPT to create a conversation tool for their students and language learners will then be able to practice their conversation abilities. 

The impact of machine learning on Education has been a revolutionary tool for both educators and learners. School districts and postsecondary schools have approved AI tools around the world while others have chosen to ban any use of AI. In British Columbia, School Districts have been approved and at times encouraged to use AI in teaching students how to safely use these technologies as well as to use them to create teaching material and to provide formative feedback. Educators can also use ChatGPT to create summative questions and revise these assessments to fit the level of their students and personalize them with prompts to relate to their students’ needs. For example, in a physics classroom if students are specifically passionate about basketball the questions can be adjusted to relate to their interests rather than having generic interstellar and ramp questions. Lastly educators can use AI to create scenarios and situations for students to analyze, either linguistically or socially as discussions and open areas of critique for the same AI system that created these scenarios. Showing students the drawbacks and limitations of even the most sophisticated AI system to date. 

For students the benefits are even greater; students may use ChatGPT to revise and proofread their work. To improve or judge the caliber of their writing as well as to act as a soundboard if they are unsure about their theses. Secondly a significant use of ChatGPT in the sciences and math is to generate questions for self-testing and to act as a “learning buddy” (Ma et al., 2024, p.6). Higher age group students often wish for greater practice opportunities than their written texts provide or to create self-exams for their history or science courses. By inputting the content outcomes and requesting ChatGPT with the correct prompts they are able to create their own assessments to practice and check their understanding and able to adjust the difficulty with their own prompts. Yet a major concern is that over dependence on ChatGPT by students for written work generation is that it will lead to “students’ human unintelligence, unlearning, and deficiencies in students’ academic and professional development” (AlAfnan et al., 2023, p.66). Students need to be provided the proper guidance and given opportunities to learn and demonstrate their learning in application based assessments which cannot use ChatGPT to solve or write their papers. 

ChatGPT and AI powered learning tools have become the next step in revolutionizing literacy and education. Every search engine now provides an AI overview rather than the resources we are looking for which often suffer from the same drawbacks as ChatGPT in their bias, hallucinations, and fabrications. Students have harnessed these tools for their own learning as have educators, however impressionable students in secondary school have started developing a dependency on these AI tools and are using them as a crutch rather than a tool. Educators can also benefit from ChatGPT by utilizing its ability to synthesize new questions and assessments while providing personable formative feedback as long as no identifying material is provided to the program. Overall ChatGPT is the next step but it requires an active engagement by educators to teach the new generation how to ethically and authentically use this resource. An active engagement that needs to be taught professionally and developed by school districts for all teachers rather than as an option. 

References

AlAfnan, M. A., Dishari, S., Jovic, M., & Lomidze, K. (2023). ChatGPT as an educational tool: Opportunities, challenges, and recommendations for communication, business writing, and composition courses. Journal of Artificial Intelligence and Technology, 3(2), 60–68. https://doi.org/10.37965/jait.2023.0184 

Bašić, Ž., Banovac, A., Kružić, I., & Jerković, I. (2023). ChatGPT-3.5 as writing assistance in students’ essays. Humanities and Social Sciences Communications, 10(1). https://doi.org/10.1057/s41599-023-02269-7 

Dogra, R. (2023, September 25). Understanding AI Hallucinations : Exploring Causes, Risks and Prevention. Aiworldtoday.net; AI World Today. https://www.aiworldtoday.net/p/understanding-ai-hallucinations 

Emsley, R. (2023). ChatGPT: these are not hallucinations–they’re fabrications and falsifications. Schizophrenia, 9(1), 52. 

Foote, K. D. (2023, December 28). A Brief History of Large Language Models. DATAVERSITY. https://www.dataversity.net/a-brief-history-of-large-language-models/ 

Grzybowski, A., Pawlikowska–Łagód, K., & Lambert, W. C. (2024). A History of Artificial Intelligence. Clinics in Dermatology, 42(3). https://doi.org/10.1016/j.clindermatol.2023.12.016

Howarth, J. (2024, June 11). Languages Supported by ChatGPT. Exploding Topics. https://explodingtopics.com/blog/chatgpt-supported-languages 

Janzen, R., Church, C., & Paleja, R. (2024, August 2). Exploring AI: Students Share Their Awareness and Usage of Generative AI. Academica Forum; Academica. https://forum.academica.ca/forum/exploring-ai-students-speak-about-their-awareness-and-usage-of-chatgpt 

Lew, R., Ptasznik, B., & Wolfer, S. (2024). The effectiveness of ChatGPT as a lexical tool for English, compared with a bilingual dictionary and a monolingual learner’s dictionary. Humanities and Social Sciences Communications, 11(1). https://doi.org/10.1057/s41599-024-03775-y 

Ma, Q., Crosthwaite, P., Sun, D., & Zou, D. (2024). Exploring ChatGPT literacy in language education: A global perspective and comprehensive approach. Computers and Education Artificial Intelligence, 7, 100278–100278. https://doi.org/10.1016/j.caeai.2024.100278 

Mondal, H., & Mondal, S. (2023). ChatGPT in academic writing: Maximizing its benefits and minimizing the risks. Indian Journal of Ophthalmology, 71(12), 3600. https://doi.org/10.4103/IJO.IJO_718_23

OpenAI. (2025). ChatGPT (Mar 25 Version) [Large Language Model]. https://chat.openai.com/chat