As a veteran of over 200 online courses and a long-time Coursera enthusiast, I am interested in how the platform continues to evolve. I like the flexibility of self-paced courses and Coursera Coach can explain concepts clearly if I don’t fully grasp them while watching the videos.
Peer reviews have been an intrinsic part of Coursera from the early days, but in recent years have become less common. In 2023, about 39% of courses included them; today, only around 25% do. In the last year, only 10% of new courses included peer-graded components. And in many older courses, peer-reviewed tasks have been made optional, meaning you can earn the certificate without completing them at all.
In an ideal world, all learners would adhere to the Honor Code when submitting and reviewing assignments and issues of poorly-prepared assignments and unfair grading would not arise. Sadly, we are not in a ideal world.
What Has Coursera Actually Changed?
The jury is still out on how effective reducing the percentage of courses with peer reviews and making them optional really are.
AI grading has reduced waiting times and ideally removed some grading inconsistencies, although dealing with peer-reviewed assignments continues to haunt Coursera. Can AI identify generic, plagiarised, or AI-generated submissions? That question would need further investigation.
Issues with peer reviews have been going on for years. Back in August 2012, (during The Year of the MOOC) when Coursera was less than 12 months old, Audrey Watters wrote The Problems with Peer Grading in Coursera. Issues of the time included language barriers, criticism of grammar, style, and opinions, and minimal comments. Some learners hid behind the anonymity of peer feedback to deliver scathing reviews.
In 2021, I highlighted problems with peer assessments. Reviews were no longer anonymous, but inconsistent reviews remained an issue. Learners can re-do low-scoring answers, which opened the door for cheaters to copy others’ work. Another issue following the COVID-19 pandemic was the delay in receiving reviews. In 2023 I wrote a follow-up investigation which showed that learners and reviewers who didn’t follow the Honor Code were still not being penalized.
To be honest, I hoped that the advent of AI grading (introduced in late 2024) would have solved the problem of empty or one-word assignments receiving a passing grade from uncaring or fraudulent peers. To an extent it has. And delays no longer happen when AI reviews your work within a minute or two.
It’s gratifying to see my result as soon as I finish reviewing the minimum number of my peers’ answers (often three, although some courses ask for five. IBM’s Computer Networks and Network Security requires only one review).
As an experiment, I submitted minimal answers to assignments in some current courses, and AI gave them a failing grade. Learners can select a box to have their answer manually assessed if they are not happy with the AI grade, so I did this. My answers still failed. For this reason, I was surprised to see the recent Reddit comments.
Why Keep Peer Reviews?
Given the issues, why keep peer reviews at all? Why do the work if the grading is unreliable or it doesn’t count?
I still believe the effort is worth it. Researching and writing a real answer deepens your understanding and helps the material stick far better than a multiple-choice quiz.
Also, these certificates need to carry weight. If you can earn one with barely any effort, employers won’t value them. When low effort succeeds, hard work feels meaningless. That hurts dedicated learners who actually complete the work, employers hiring people, the universities, and Coursera’s own reputation.
Instead, let’s all remember that the real value of taking courses is in gaining knowledge and completing projects that can be used in the real world. I’ve taken over 200 online courses since 2012. While I can’t remember much detail from some of those courses, others have been supremely useful. They’ve helped me learn more effectively, become a better writer, use AI, improve my creativity and health, and even pick up some basic Python programming. Many courses that stick in my memory are those with research, projects, and written work.
Interestingly, some newer courses such as Generative AI: Introduction and Applications and Critical Thinking for Better Decisions in the ChatGPT Era include ungraded items to practice with AI. In other words, you’re using AI for a hands-on exercise that doesn’t count towards your final grade.
The direction is clear: Coursera appears to be moving away from human-dependent peer assessment rather than improving it.
Why Hasn’t Coursera Fixed This Yet?
There are two main theories.
The first theory relates to the effort put in. Part of a credential’s credibility comes from the effort required to earn it. But if you make the process too demanding, fewer people will finish courses. Fewer completions mean fewer satisfied customers and fewer certificates sold. In other words, stricter quality control may conflict with Coursera’s scaling ambitions.
The second theory involves the cost of operating at a massive scale. Fixing peer grading is not just a UX tweak, it needs an effective system for reviewing disputes, staff to investigate plagiarism claims and user complaints, moderators to enforce quality standards, and humans who step in when automation fails.
People are eager to invest in the parts of a business that scale. Not everyone wants to invest in the humans to maintain that scale. Peer review is slow and can be messy and expensive to do well. Some learners refuse to grade the work of their peers because they either don’t feel confident in their knowledge or object to being asked to do work that they consider to be the work of instructors in a paid course.
Final Thoughts
In the early days of free courses with free statements of accomplishment, Coursera helped universities provide free access to high-quality learning, but recent white-labelled courses and paid content have eroded this ideal. Peer grading remains a weak link with years of repeated complaints and little structural change.