Teachers say they are spending hours reviewing tests that were scored incorrectly by auto-grading bots, which try to match a student's response to an answer key
Auto-grading software often marks answers incorrect when they are right, giving teachers and families more remote-school headaches Tweets: @sub8u , @hypervisible , @brittanystoroz , @wsj , and @mims Tweets: Subrahmanyam Kvj / @sub8u : Algorithmic degradation of everyday life comes in many flavors. Kids graded by bots which are optimized for precise field-matching and sold by edtech companies promising “auto-grading” is yet another! https://www.wsj.com/... https://twitter.com/... @hypervisible : “Auto-grading promised to save teachers time but now has many of them spending hours poring over answers to their students' every assessment and exam—if they're aware of the grading problems at all.” 😵 https://www.wsj.com/... @brittanystoroz : How many kids are just out there failing now despite all of their hard work studying and preparing? Grading is brutal on teachers but this is worse. https://www.wsj.com/... @wsj : Teachers say they have to spend hours reviewing tests that were scored incorrectly by the automated systems more schools are adopting. “So much for technology making our lives easier.” https://www.wsj.com/... Christopher Mims / @mims : Automated systems are being used to grade kids' schoolwork?? This is bad, reports @juliejargon https://www.wsj.com/... https://twitter.com/...
Context & Ripple Effects
This story lands at the midpoint of an arc the related coverage traces cleanly: in late 2019, outlets were still framing AI as a relief valve for teachers, taking over repetitive tasks like grading and coursework optimization. A year into remote schooling, the WSJ reports the flip side — answer-key-matching bots marking correct responses wrong, with teachers absorbing the cleanup.
The piece also prefigures the broader edtech accountability debate that followed: by 2022, [[a:979334|proctoring software like Honorlock was being criticized for flagging normal student behavior]], and by mid-2024 teachers were adopting new AI grading tools even as critics flagged glitchy, overly harsh scoring — a market that keeps selling automation faster than it can verify it.
First-order effects
- Teachers lose the exact time savings auto-grading promised, spending hours manually reviewing bot-scored tests during remote instruction.
- Students and families receive incorrect scores on graded work, compounding the communication and trust strains of remote school.
Second-order effects
- Edtech vendors selling 'auto-grading' face pressure to add human-review layers or accuracy guarantees, since the product's core value proposition collapses when every output must be checked.
- Schools weighing automation tools get a cautionary data point alongside the proctoring backlash, making procurement more skeptical of vendor claims about fully automated assessment.
Third-order effects
- If the pattern holds across grading and proctoring tools, classroom automation settles into an oversight-inversion model: software performs the task, but teachers are structurally recast as unpaid quality-control for algorithms they didn't choose.
- Persistent error rates in high-stakes automated judgment push toward explicit liability frameworks for algorithmic harm — the same question Christopher Mims raises about who bears the cost when large volumes of people act on machine-generated errors.
The trend: Education automation keeps transferring verification labor back onto teachers, turning promised time-savers into review queues and feeding a broader reckoning over who is accountable when algorithmic judgments fail.