A return to AI feedback in Moodle

This summer physical science will again pilot AI feedback in Moodle. This summer the laboratory reports are dropping the explicit rubric and using essentially the same marking system but without an explicit rubric.

The original rubrics for the course appeared as above circa 2010. 

December 2016

By 2018 the rubric used in Schoology had shifted to a ratings scale to conform to the way rubrics are set up in a learning management system. With a shift in ratings terminology and points, this structure would carry over into Instructure Canvas.

March 2021 statistics rubric built on outcomes in Canvas

Canvas allowed course learning outcomes to be included directly in the marking rubric. Physical science rubrics also included content criteria.

January 2026 in Moodle

By 2026 rubrics in Moodle had become minimalist. Experience had indicated that students were not reading nor reacting to the rubric marking. Where in Instructure Canvas learning outcomes could be included in rubrics, in Moodle learning outcomes cannot be included in rubrics. 

January 2026

Feedback was handled by comments which were made in the Comments field, copied into the Feedback field, and sent via Moodle messages to the student. This triple copy approach deals with the issue that the Moodle app does not, at least on most cell phones, display the comments or feedback. To ensure that the student sees the feedback, all three channels are utilized. The rubric becomes a marking system, not a formative assessment system. The students just are not used to reading and responding to rubrics or comments put in the rubric comments field. 

April 2026

A one off rubrics was built for the Ohm's law laboratory in order to feed more information into the AI Feedback (AIF) system that was piloted on this laboratory. This partially replicated information in the assignment description field. This led AIF to produce redundant feedback: once for feedback on whether the student met the description requirements and a second time for whether the student met the rubric requirements. With the students not known to be referencing the rubric, and the rubric now not being a truly formative rubric (the ratings do not specify was distinguishes each rating level from another rating level), tossing the rubric seemed like a reasonable experiment. 

Another reason to drop the rubric is that the Moodle app cannot mark a rubric graded assignment. One has to log into Moodle with a browser to access the rubric. In the fast and furious pace of the summer, not having to switch to the browser to mark saves a stitch in time. 

Although one can mark rubrics in a browser on a cell phone, jewelers magnifying glasses are almost necessary to see the interface text. And zooming/dezooming in the marking interface can be glitchy in a cell phone browser. By not handling rubrics in the Moodle app, Moodle does a disservice to decades of learning assessment specialists encouraging the use of rubrics. 

Summer marking on mobile has been made easier by not using rubrics.

Experimenting with the AIF Expert Mode Template indicates that AIF separately pulls in the description and the activity instructions. 


Given this distinction in the template, the description was pared down to be just the description. 


The laboratory structure was moved into the Activity instructions.


The role was modified in the Expert Mode Template and the prompt was placed into the teacher instructions. The internal AI model being used does not process images, hence the attempt to exclude those three sections. 

Back when each rubric was custom designed for a specific laboratory, the deletion of the rubrics would have made restoring the rubrics a challenge. Now that the rubrics are essentially identical from laboratory to laboratory, restoring them should the need arise is less of a challenge. Summer session has always been a time to explore and experiment. Summer 2026 is no different. 

The prompt was as follows:

=== ROLE ===

You are an experienced teacher providing constructive feedback on student submissions in an introductory non-major science course.

=== ASSIGNMENT ===

{{assignmentname}}

{{description_section}}

{{instructions_section}}

=== TEACHER INSTRUCTIONS ===

Evaluate the laboratory report for the presence of the five sections. Provide feedback on the introduction and the discussion. Do not provide feedback on the data table, data graph, and data analysis sections. 

=== STUDENT SUBMISSION ===

{{submission}}

=== OUTPUT INSTRUCTIONS ===

Provide detailed, constructive feedback that helps the student improve.

Focus on both strengths and areas for improvement.

Be encouraging but honest.

Format your response in strict Markdown:

- Use ## for section headings.

- Use "1. " for numbered lists (NOT "1)").

- Use "- " for bullet lists.

- Always leave a blank line before and after headings, lists, and code/math blocks.

- Use **bold** and *italic* for emphasis.

- For math equations use LaTeX: inline with $$ ... $$ and display blocks with \[ ... \].

=== LANGUAGE ===

Respond in {{language}}.

After some backend work, first light for AIF would arrive with feedback for laboratory six. 


That feedback followed the directions in the prompt and only provided feedback for the introduction and discussion. In general the feedback was quite good. There was some feedback on mathematics notation, but the course does not include how to properly format mathematical notation in their laboratory reports. Impressive results. Clearly, getting the prompt right is going to be key to successful use of AIF. 

Section Evaluation

You have successfully included all five required sections of the laboratory report:

  1. Introduction
  2. Data Table
  3. Data Graph
  4. Data Analysis
  5. Discussion (labeled as Conclusion)

While all parts are present, there are specific areas in your writing and technical notation that need attention to meet science reporting standards.

Feedback on Introduction

Your introduction clearly identifies the system you studied (cooling water) and the goals of the experiment. You have a good understanding of what needs to be determined.

To improve this section: - Avoid first-person narrative: Instead of saying "I am experimenting" or "I have to check," use objective language. For example, "This experiment investigates the cooling of water..." - Reduce repetition: You mention that you are looking for the "mathematical relationship" several times. Try to combine these sentences into one cohesive statement. - Refine vocabulary: Instead of saying "seeing the changes," use more scientific terms such as "observing" or "analyzing the temperature fluctuations."

Feedback on Discussion

You did a great job answering all the core questions posed in the assignment description. You correctly identified the exponential nature of the cooling and linked it to Newton's Law of Cooling. Your identification of the ambient room temperature is also correct.

To improve this section: - Mathematical Notation: There is a significant error in how your equation is written. You stated the relationship is exponential, but you wrote the equation as: [ T_1 = (100 - 38.31773)(0.879668)t1 + 38.31773 ] In this format, $t1$ appears to be multiplied by the base. For an exponential relationship, the time variable must be in the exponent: $$ T_1 = (100 - 38.31773)(0.879668)^{t_1} + 38.31773 $$ Please ensure your final report uses proper superscript for exponents. - Sentence Structure: The final sentence is a "run-on" sentence. It combines two different ideas without proper punctuation. I suggest breaking it into two sentences to improve clarity. - Depth of Explanation: While you correctly identified that the cooling rate is related to the temperature difference, you could strengthen your discussion by explaining why this happens (e.g., referring to the direction of heat flow from the hot water to the cooler air).

Overall, you have a strong grasp of the concepts. With a few adjustments to your formal tone and mathematical notation, this will be an excellent report!

(This feedback was generated by AIF system.)

Comments

Popular posts from this blog

Setting up a boxplot chart in Google Sheets with multiple boxplots on a single chart

Experimenting with the PlantNet and iNaturalist apps

Traditional food dishes of Micronesia