Providing feedback is one of the top 10 things an instructor can do. But what makes for useful feedback?

The following is a post that I wrote for the MTEI Teaching Tip blog during my term as MTEI’s associate director.

It’s been said that practice doesn’t make perfect — rather, practice with feedback makes perfect. Feedback to students on exams, papers, presentations, labs, problem sets, and other assessments is a crucial element of the learning process: providing feedback is one of the top 10 things an instructor can do [Hattie 2009]. But the effect size of feedback, as measured in meta-analyses, is substantially influenced by the information content conveyed in that feedback [Wisniewski et al. 2020].

So: What makes feedback useful to students? How can you offer it at scale in a large class? Will students engage with it? We have some suggestions based on research in the field.

Characteristics of Useful Feedback

Useful feedback is:

• Timely: the feedback comes soon enough for students to remember their thought process in constructing their answer.
• Noticeable: the feedback stands out and gets the attention of students.
• Relevant: the feedback needs to be tied to goals that the students understands and shares.
• Actionable: the feedback suggests opportunities for students to refine and improve upon their previous performance, perhaps by pointing to specifically relevant exercises or readings.
• Specific: the feedback helps students locate the discrepancies between their answer and a correct solution.

There is a spectrum of specificity:

• Binary: the feedback conveys only a summary right-or-wrong judgment without any supporting details — for example, ❌ vs ✔︎.
• Terse: the feedback identifies properties of an answer that are deemed erroneous — for example, “incorrect calculation.”
• Explanatory: the feedback explains where there are errors in a student’s answer, and why they are deemed to be errors — for example, “in the third step, Bayes’ Rule was applied incorrectly, because the conditional probabilities were swapped.”
• Overwhelming: the feedback is too lengthy or too complex for students to understand it, and might be demotivating — for example, a digression on axiomatic formulations of probability theory.

Advanced students might need only binary feedback. Novices need terse or explanatory feedback, where the optimum amount of detail depends on how disruptive the feedback will be in the midst of the task they are performing.

(Acknowledgment: the material in this section is adapted from The ABCs of How We Learn by Schwartz et al., Norton, 2016; and Hattie and Timperley, 2007.)

Strategies for Efficient Feedback

Providing individualized feedback at scale can be a challenge. Devoting office-hour time to customized feedback, from you and/or TAs, can help. Meanwhile, feedback to a large course can benefit from the following strategies [How Learning Works by Ambrose et al., Jossey-Bass, 2010]:

• Look for patterns: find common errors and misconceptions. These might be identified by you or TAs. Give feedback on those to the class as a whole.
• Prioritize: allocate your time according to what information will be most impactful to students. Focus on the most important learning objective(s), rather than all of the objectives.
• Incorporate peer feedback: students can constructively critique their peers’ work, especially if given explicit criteria and some practice with how to apply it. The frequency of feedback can increase, since you aren’t having to write it yourself. As a fringe benefit, students can become better at identifying issues in their own work.

Gradescope can help with providing feedback at scale. Rubric items capture terse feedback, which can be written to explain not just what is incorrect, but what a correct answer would convey. Customized feedback written by graders, and localized within a submission using Gradescope’s annotation tool, can convey explanatory feedback. The latter can even reduce regrade requests by clarifying why a grader has deducted points.

Student Engagement with Feedback

Unfortunately, students don’t always engage with feedback. This might be because accepting and integrating feedback is not part of (their) standard social practice [Jørgensen 2018]. Other threats to engagement include underdeveloped metacognitive skills, such as self-appraisal and self-management; and emotional factors, such as perceived threats to self-esteem or social embarrassment [Hattie and Timperley 2007].

To improve engagement, ask students to specify how they used feedback in subsequent work [Ambrose et al., op. cit.]. For example:

• On a project with successive deliverables, you could ask for a statement with each deliverable that explains how the student incorporated the previous round of feedback.
• After handing back an assignment or exam, you could offer students some points back if they submit an improved answer based on your feedback. Consider whether or not you want to release numeric scores and/or solutions before receiving the improved answers.
• Before an exam, you could assign a one-page paper on how the student is incorporating the feedback from problem sets into their study strategy.
• After a prelim, you could invite students to submit a study plan based on their performance, and to review that plan with a TA or you. That plan could involve reflecting on why certain mistakes were made, and how to avoid them next time.