Correction: Human vs. Artificial Approaches to Grammar

This study was conducted to compare ChatGPT-generated responses and human-written responses on verb-tense instruction. After receiving sample sentences with intentional verb-tense errors, this research team prompted the AI to provide feedback on the sample work to evaluate the AI’s ability to correct grammatical errors. Following the AI-generated ESL-friendly explanations for the verb-tense errors, the team compared and contrasted this feedback to their human-created explanations. The team analyzed 20 responses, with 10 provided from each party, in terms of accuracy and clarity. The focus remained on identifying the presence or absence of incorrect form, correct form, sentence context, clarity, and extra or irrelevant information. Ultimately, the team collaborated to analyze the composition of the human and AI-generated explanations for ESL students.

The following sentences contain intentional verb-tense errors. The team fed these sentences to ChatGPT for data and analysis purposes.

1. Since I have been a college student, I made many new friends.

2. Mario will graduate as a veterinarian in January of last year.

3. There are rumors that college tuition is higher next year.

4. I live in Los Angeles for 15 years.

5. At 8:00 PM last night, I studied in the library.

Based on the results, these were the two major differences we noted between the human vs. AI-generated responses:

• ChatGPT often included extra rules or extended explanations, which sometimes added unnecessary information, whereas human responses were more direct but occasionally less thorough.

• Human responses showed more variation in linguistic terminology, while ChatGPT responses were more consistent but sometimes inaccurate (e.g., “future tense”).

In the interest of future research, our research team presents a few more questions for ponderance:

• How could this be applied to a classroom setting, specifically with the intention of aiding ESL students in their writing endeavors?

• Taking into account the human and AI-generated responses aligned in all other criteria, with the exception of “Extra and Irrelevant Information,” how does the brevity impact the ESL students’ understanding of English grammar?

In cooperation with the NACE Readiness Competencies, we applied critical thinking, communication, and technological skills.

Critical Thinking

• Analyzed linguistic data to draw conclusions on whether AI or humans can more effectively emulate the nuanced and accurate feedback required for ESL students.

• Evaluated the accuracy of human versus AI-generated explanations from a pedagogical standpoint, assessing the reliability of Large Language Models as supplemental tools for English language instruction.

Communication

• Crafted concise responses specifically tailored to ESL students that accurately explain the nature of the grammatical error within each sentence, the correct versions, and why the corrected version is used with reference to the context of the sentence. Specific observations of patterns were noted in each category of responses.

Technology

• Manipulated and managed emerging large language models to identify its appropiate use.