Piqosity now integrates with ChatGPT to AI-grade student essays. Essay prompts are found throughout Piqosity including in standardized test prep courses like TSI, ACT, and ISEE plus our English Language Arts (ELA) courses for grades 5-11. Upon writing their essay and scoring the quiz, students will now see an AI-generated score and specific feedback. As of this morning’s deployment, we are currently using ChatGPT 4o-mini.

Note that the Essay on the ISEE is not scored. Simply use Piqosity’s new AI-scoring as a tool for helping students to improve their writing. Click here to read more about the ISEE Essay.

Piqosity scores essays across six categories:

  1. Purpose and Focus
  2. Organization and Structure
  3. Development and Support
  4. Sentence Variety and Style
  5. Mechanical Conventions
  6. Critical Thinking

Each of the six categories receives an AI-generated score of 1 to 9 where 1 is the lowest score and 9 is the highest:

  • Scores of 1, 2, and 3 are below average
  • Scores of 4, 5, and 6 are average
  • Scores of 7, 8, and 9 are above average

Piqosity will justify its score in each category including citations from the student’s response. For example, if students receive a low score in the category “Mechanical Conventions,” Piqosity may specifically point out mistakes in grammar and spelling.

Additionally, students will receive an Overall Score, which is simply the average of the six category scores. This Overall Score is accompanied by a summary justification of the score including:

  • what the student did well,
  • where they fell short, and
  • specific steps to improve their essay.

Fourth generation large language models (LLMs) like ChatGPT 4o-mini are powerful tools but they do have limitations. We spent dozens of hours engineering our AI prompts and comparing the results to human-scored essays.

The AI is good at providing detailed feedback and providing “ballpark scores” for about 85% of essays that most educators would score as decent to good. However, it’s less capable at identifying truly exceptional work according to a student’s grade level or recognizing when the student simply didn’t try.

Our current implementation does a good job of mitigating for the truly bad essays by feeding specific data from the app to the AI like word-count expectations and examples of what an automatic failure looks like. However, we have been admittedly less successful at getting it to score remarkable essays as 8 and 9 on our grading scale.

As a result of these known limitations, students could probably describe our essay scoring as a “fair grader” for poor to good essays but a “hard grader” of exceptional essays. We will continue to refine our scoring methodology and look forward to even better results in the future by updating to newer LLM versions and further prompt refinement according to specific courses.

AI Essay grading is just the first of much more AI integration to come at Piqosity! We’re actively working on AI-generated questions in both our math and ELA courses to provided unlimited, on-demand personalized practice. And for educators we’re looking at implementing AI to better explain complex student stats.

Reference our help desk article for more details about Piqosity’s AI Essay Grading