Nanjing University scraps its 'head-up rate' after controversy: AI surveillance in the classroom needs consent

By WION · November 8, 2026.

Nanjing University, one of China's leading universities, has removed the feature that measured student attention from its "smart classroom" system. The move came after screenshots of the system went viral and sparked a wave of privacy criticism, WION reports.

The images showed live video of classes alongside data on attendance, late arrivals and a metric called the "head-up rate" (taitou lu). It measures whether students are looking at the teacher or are distracted by their phones, laptops or other activities.

In a statement on its official social media accounts, the university said the screenshots came from a test of the system carried out in May. It said the technology does not collect facial or biometric data, nor does it perform facial recognition during teaching without authorization. It also confirmed that the system will no longer calculate the "head-up rate."

**What the system is and what it cost**

According to the outlet, the upgraded version was reportedly switched on in the first week of November. It uses classroom cameras and AI algorithms to automate attendance taking, count people and monitor activity in class.

A tender notice on the university's website shows the project exceeded 30 million yuan (about $4.5 million). It included upgrading equipment and facilities in more than 100 classrooms and several common areas on one campus.

According to WION, some teachers reportedly warned students that attendance records could affect their grades and, in some cases, their eligibility to sit exams. This matters: it turns a technical metric into a real academic consequence.

Several students told local media they were uncomfortable that their biometric data could be collected and shared with the university or with private tech companies. They also asked why they had not been consulted beforehand. One lecturer told the South China Morning Post that the system was mainly intended to assess teaching quality, not to monitor individual students. The university maintains that the upgrade is meant to facilitate live streaming, interaction between campuses and the recording of classes for reuse.

**Our take**

The thesis is simple: an attention metric can measure something real and still be a bad product. The problem is not counting how many heads face forward. It is what is done with that number and who decides it gets measured.

There are two very different versions of the same system. In one, the aggregate helps the teacher see which parts of a lesson lose the group and improve their teaching. In the other, the individual data point ends up in a grade or an exam permission. The technology is the same. The governance is not. The university insists on the first version, but the warnings about grades and exams reported in the press point to the second.

It is also worth being cautious about what we do not know. The university denies unauthorized facial recognition, and we have no way of verifying that from here. Nor is it clear how widespread those warnings to students were. What can be verified is the outcome: the feature is being withdrawn under public pressure.

That is perhaps the most interesting point. In a country known for adopting surveillance technology without much friction, pressure from students and public opinion was enough to roll back a feature. According to WION, the case has reignited debate in China over transparency, consent and safeguards. As context for the sector, it fits a broader pattern: social acceptance of AI in sensitive areas depends on clear human oversight and consent more than on the algorithm's accuracy.

In the short term, we expect more episodes like this. Schools and universities are buying classroom systems worth millions, and the AI layer comes almost as standard. The loser is the student when nobody asks them. The winner, if done well, is the teacher with tools to teach better.

In the long term, AI in education has enormous potential: personalized tutors, classes adapted to each pace, access to the world's best teacher for anyone. But that future only holds if families and students trust it. Tracking a student's gaze to grade them does not build that trust. Asking their permission and explaining what the data is for does.