TOM DACCORD

View Original

Navigating the New Frontier: AI-Driven Assessment of Student Writing

There are numerous excellent and reliable tools for creating and grading multiple-choice, true/false, and fill-in-the-blank quizzes. However, grading student writing with technology poses greater challenges, as accurately assessing writing quality involves various complex factors. Fortunately, AI is making steady progress in this area, with recent enhancements to GPT-4 enabling more possibilities in assessing student work.

Writing is a multifaceted skill encompassing grammar, syntax, tone, and style. Providing accurate and helpful feedback through AI is challenging, requiring advanced algorithms to analyze writing in detail and offer specific improvement suggestions. AI often struggles with evaluating writing quality based on context and audience, and assessing originality and creativity can be problematic, too, especially in literature and poetry.

As AI technology advances, many limitations are increasingly being addressed. Recent developments, such as the introduction of "GPT-4 Turbo" by OpenAI, enhance ChatGPT's performance and allow for larger uploads. This enables more detailed prompts that can better evaluate student writing.

Teachers can upload writing standards and rubrics as part of a ChatGPT prompt. For example, when assessing a high school English essay, one could upload state or provincial expository writing standards. Using a specific rubric, such as the B.E.S.T Writing Scoring Samplers provided by the Florida Department of Education for grades 7-10, can form the foundation of your assessment.

Additionally, you can include your own or a departmental rubric in the ChatGPT prompt, alongside the uploaded standards. Informing ChatGPT which aspects of the rubric to prioritize can further refine the evaluation.

ChatGPT's accuracy and precision are further enhanced when you provide samples of student work, particularly if accompanied by grading summaries. Seeing examples of graded work helps ChatGPT to better assess the quality of student submissions.

Finally, instructions are critical. AI thrives on clarity and precision, so be explicit in what you want ChatGPT to do. Provide context, including details about your course, the unit, the students' grade level and ability level (e.g., Advanced, ESL, etc.), and your assignment's specifics.

To recap, three foundational elements are crucial for creating an effective ChatGPT prompt to assess student writing:

  1. state/provincial writing standards and rubrics and/or your rubric or scoring standards

  2. samples of student work (with grading summaries)

  3. clear, detailed instructions.

Tools like Eazl.ai can assist in creating and managing personalized prompts. (You can access a sample 8th grade expository writing prompt here.) For those who find crafting detailed AI prompts for grading daunting, numerous AI graders are available at a low cost. I suggest you explore my AI Tools for Essay Feedback page for recommendations. It you do opt for an AI grader tool, it’s important to check whether the tool uses GPT-4 for enhanced accuracy and customization.

We are just beginning to explore AI's potential in teaching and learning, but new possibilities for assessment of student work are opening up quickly. The challenge for us is harnessing these advances effectively. While AI tools for grading are becoming more adept and accessible, personalizing them for all learners, particularly those needing accommodations, should be our ultimate goal.