Part of the process in designing an alternative assessment is to plan a means of evaluating the assessment. Since authentic assessments resist the idea of one correct answer, we need a way of providing feedback to students about their learning that acknowledges a diversity of perspectives but at the same time does not add significantly to our grading workload.
You may also want to consider the holistic dimensions of the assessment, or how it exists within the larger context of your course. If you are rethinking your assessment design for compassionate or equity-based reasons, it follows that your broader grading scheme or approach might reflect similar concerns.
Flexible grading practices, such as:
can lessen the pressure or anxiety students experience in relation to your course assessments.
That said, flexible grading practices can also result in an increased workload (e.g. regrading) or impact the ability of you or your teaching assistants to give timely feedback (e.g. grace periods). We appreciate that there are numerous factors to weigh when deciding on a grading strategy, and not all practices are realistic for every instructor.
You may have chosen a new-to-you assessment format and are not sure how to approach grading the assessment, if you will be assigning a grade. Rubrics can help to break the daunting prospect of evaluating students’ work into set of manageable criteria.
Rubrics are often rendered as a grid or table, where the rows of the table describe the various criteria, and the columns of the table contain the levels of achievement and/or the points allocated. The cells of the table typically describe the characteristics of the various levels so that it is clear to students and those grading student work (e.g. teaching assistants) where the work falls on the grid.
If you are unfamiliar with rubrics, you may find an example helpful:
Some additional examples can also be found in York University’s “All About Rubrics” guide.
Types of Rubrics
All of the example rubrics are analytic rubrics, the most commonly used type of rubric. But it is not the only option, and you may find another type of rubric more suitable for your assessment design needs.
An analytic rubric has two dimensions: criteria and achievement / performance level. There are usually descriptions for each performance level in each criterion that clearly communicates expectations to students.
Analytic rubrics tend to be the most labour intensive to develop when writing descriptions to fill out the entire table. As such, they may be best suited to larger assessments or a set of repeated assessments that can reuse the same rubric. The labour can be worthwhile, however, when the rubric is used in a learning management system to automate the process of giving detailed feedback on the rubric criteria.
Note: we used generative AI (Claude) to draft realistic rubric criteria for a generic writing example. While there are many good examples of rubrics on the web, finding a suitable one can be a challenge; large language models (LLMs) are better suited to semantic queries than conventional search engines (e.g. “an analytic rubric for a writing assignment that emphasizes qualities of good academic writing which generative AI does not do well”). Further documentation regarding the prompt and response can be found under “References” on this page.
Learn more about how to use rubrics in D2L / Brightspace
A holistic rubric considers all of the criteria at once when assigning a level; it is like an analytic rubric where the criteria have been consolidated into one row of the table. The criteria or level description should clearly communicate the differences between levels.
As a result, a holistic rubric can take less time to develop than an analytic rubric. The trade-off is that students receive more generic feedback, and it can be more challenging to assign a level when student work is inconsistent across the various criteria.
While a holistic rubric “flattens” the number of criteria to one, a single-point rubric reduces the number of levels to two: either achieved or not achieved. It is like pass / fail grading on each criterion.
A single-point rubric can make it easier to decide where a student’s work falls in the rubric, but you may find yourself needing to compensate for the lack of gradation through written comments.
Note: it is possible to include levels of achievement in a single-point rubric; refer to the rubric guide created by Brock University’s Centre for Pedagogical Innovation.
To create a rubric, start by identifying the key competencies you want students to demonstrate through the assessment. Your learning outcomes and ideas about what the assessment is meant to measure will be a guide.
You may find it helpful to look for an existing rubric that you can adapt, or you can ask generative AI for suggestions. You can also refer to the guiding questions below:
Of course, rubric design is also shaped by disciplinary norms and the form of the assessment that you choose.
There is no optimum number of criteria to include but try to find a balance between adequately representing the characteristics that define your expectations for student work and burying the most salient characteristics within a large list. If you have more than 5 criteria, consider organizing them into categories of related criteria.
Note: Not every criterion has to have the same weight. When creating a rubric in a learning management system, you can usually have multiple groups of criteria with different maximum scores.
The rubric levels indicate to students the degrees of achievement. They are generally expressed in both quantitative (as points or scores) and qualitative (as level descriptors) terms.
If you are creating a holistic or analytic rubric, aim for 3 to 5 achievement or performance levels (the columns in the grid). The levels do not necessarily have to be evenly distributed – for example, a four-level rubric might distribute the score as 5 points / 4 points / 2.5 points / 1 point.
The level descriptors communicate information about the level of achievement. A few examples of descriptors are given below:
Some educators critique rubrics for giving an illusion of objectivity to a fundamentally subjective act. If you like the structure that rubrics provide but want to introduce another dimension of fairness to assessment, you can incorporate anonymous grading, where the assessor does not know the name of the student who created the work.
Some learning management systems have a feature that allows you to hide student names during assessment when creating an assignment submission folder.
Reflection is a contemplative act of purposefully analyzing experience in a deliberate manner. It is an important dimension of active learning but also acts a record of learning when documented.
Here are some helpful resources for designing and evaluating reflective activities.
Students who are accustomed to traditional assessments like multiple choice exams may need guidance on the types of questions to ask themselves when reflecting. You might provide structured prompts or exemplars to communicate your expectations. For more guidance around incorporating reflective practice into your teaching and sample prompts, please refer to the Reflective Teaching Practices module.
Model: Claude Haiku 4.5 (least intensive model chosen due to simplicity of the task)
Human-authored prompt:
“For a second-year university scholarly essay assignment, please suggest 4-5 criteria to be used in an analytic rubric with four performance levels. Students are not permitted to use generative AI in writing the assignment and the rubric criteria should reward student work done independently of generative AI, and qualities of good writing in general. E.g. critical thinking, specificity of examples, student voice, etc.”
Generative AI response:
“Here are 4-5 rubric criteria designed to assess authentic student scholarship while discouraging reliance on generative AI:
Rewards students for developing their own interpretations rather than restating conventional wisdom.
Rewards careful selection and integration of precise examples—a hallmark of genuine research and close reading.
Rewards authentic expression, consistent perspective, and honest engagement with complexity and uncertainty.
Rewards logical architecture that reflects the student’s conceptual understanding, not formulaic templates.
Rewards demonstrated learning from lectures, seminars, and assigned readings—showing the student has done the actual coursework.
This rubric prioritizes qualities that emerge from genuine engagement with ideas rather than surface features that AI systems excel at mimicking.”
Human in the loop:
I reviewed the output from Claude closely and judged it to be more than adequate, even using the most lightweight Claude model. Because the rubric descriptions were merely for the purposes of illustration – and yet, ideally still helpful to module learners – I only made a few minor edits to increase the tenor or precision of the language and omitted the fifth criterion.