Why feedback should come from humans, not algorithms

Where technology works in giving feedback and where it falls short.

Hey teachers… remember these?

To my fellow millennials, this is called a “gradebook,” back when it was an actual “book.” I’m told that teachers used to add each student’s score by hand, divide to calculate a grade, repeat 120 or so times, and then submit all grades to an administrator. Then all scores from all classes were transferred by hand onto each student’s report card, and each report card was put in a sealed envelope with a stamp and mailed to each family. Yikes.

Now that gradebooks look like this:

all stakeholders can automatically see the grade as soon as the teacher enters it. Phew!

The Impact of Technology in Classrooms

The electronic gradebook is one of myriad technological advancements that automate time-consuming processes and drastically improve teachers’ jobs. There has been an explosion of these products in the last five years. We now have apps that help teachers communicate with parents, track classroom behavior, share lesson plans, and provide students with supplemental resources.

“There’s at least one time-consuming part of a teacher’s job that algorithms can’t improve: giving feedback.”

Ed tech and the algorithms behind them have the power to save teachers time and effort. But there’s at least one time-consuming part of a teacher’s job that algorithms can’t improve: giving feedback. Though some companies are investing millions of dollars in this pursuit, here are three reasons why humans are better than algorithms at giving feedback:

1. The Speaker Matters

Let’s say we create an algorithm that gives the same feedback as a teacher would. Isn’t that the same thing?

Not exactly.

We interpret feedback differently depending on who the speaker is. For instance, “Eat this apple,” seems like reasonable advice from anyone who looks like this:

But the wisdom of this advice changes drastically when the speaker looks like this:

How students receive, process, and react to feedback depends not only upon the speaker, but also on a student’s relationship with the speaker. As Douglas Stone and Sheila Heen say, “All feedback is colored by the relationship between giver and receiver, and we can have reactions based on what we believe about the giver…or how we feel treated by the giver.”

Students are much more likely to respond to a teacher who they trust, who cares about them, and to whom they are held accountable. Students don’t have the same trust relationship with algorithms.

2. Algorithms lack empathy

Giving excellent feedback requires a teacher try and get inside of the mind of the student to understand their thinking. By reviewing a student’s work, teachers can uncover misconceptions and guide students to the right path.

Do algorithms work this way as well? Adaptive learning company Knewton claims that “this robot tutor can essentially read your mind.”

Spoiler alert: It can’t.

This is because Knewton is not trying to read a student’s mind in the same way a teacher is. It does not analyze a student’s work to uncover misconceptions. It simply looks at a student’s final answer on a multiple choice question and tries to retroactively determine where the student erred. The technology might know what content to present next, but it can’t give the student any real feedback on their misconception.

Empathy is natural to humans, but not to algorithms.

Algorithms are limited this way because they lack empathy. They don’t “think” the way students do, so they need shortcuts. Teachers on the other hand are good at describing misconceptions, because they are able to think through the problem as the student did. Empathy is natural to humans, but not to algorithms.

3. It’s not about giving feedback anyway

What if we programmed algorithms that could understand and identify student misconceptions. Problem solved, right?

Not really.

Scholar Dylan Wiliam reminds us that “Feedback should be more work for the recipient than the donor.” This follows the logic of the “teach man to fish” adage. Giving the misconception to the student limits their understanding to that specific problem. It is far more effective to teach students to uncover their own misconceptions.

The latter occurs when students engage in metacognitive thinking, questioning, self-assessing, and other practices that require students to discuss their work. Teachers or peers can give feedback in this interactive way: providing clues when needed, asking questions when needed, but ultimately putting the work of thinking onto the student. Algorithms that provide the misconception only rob students of the chance to think for themselves.

What does this mean for ed tech?

As a rule of thumb, we should use ed tech for processes that software can do more effectively than teachers. This applies undoubtedly for calculating and sharing grades, but not for giving feedback. Even if it takes us longer, humans dominate algorithms in the quality of feedback we can provide. So let’s not waste time and money on mind-reading apps.

“Technology should augment a teacher’s ability to do what they do best.”

Technology does have a place in feedback, but not in providing feedback content. Instead, technology should augment a teacher’s ability to do what they do best by accelerating the feedback process. Apps like Echograde, Formative, and Google Classroom are designed with this idea in mind and serve as models for how we can approach feedback and ed tech.