The Attendance Camera
A district committee is considering a facial-recognition attendance system for a high school with 1,100 students. The vendor claims 99.2% overall accuracy and says the system will save 12 minutes per class in roll-taking. An independent audit submitted by a parent group reports error rates of 1.1% for lighter-skinned students and 4.8% for darker-skinned students. The proposed deployment would use cameras in the cafeteria and main entrance, retain images for five years, and share data with a transportation vendor. Students and parents have not been asked for consent, and there is no published process for correcting a misidentification. A teacher representative on the committee supports a pilot because attendance data are already collected. The committee has one meeting before voting. The teacher must decide whether to support the pilot, oppose deployment, or propose conditions, and must explain the decision to students in a later lesson.
1. Identify the stakeholders and the main ethical harms in the proposal. Explain why overall accuracy is not sufficient evidence of fairness.
Read the worked answer
Stakeholders include students, especially students with darker skin tones who face higher error rates; parents; teachers whose time is being saved; the transportation vendor; and the district, which bears legal and reputational risk. The main harms are misidentification, loss of privacy, and function creep. A student who is incorrectly marked absent may face disciplinary or academic consequences, and the error is not randomly distributed: the 4.8% error rate for darker-skinned students is more than four times the 1.1% rate for lighter-skinned students. Overall accuracy of 99.2% can hide that disparity because it averages across groups and may be dominated by the majority group. The five-year retention and data sharing with a transportation vendor expand the harm beyond attendance. The absence of consent and of a correction process removes due process. The teacher should treat the audit disparity as the central ethical issue, not as a technical detail to be fixed later.
2. Apply the principles of data minimization, consent, and purpose limitation to the proposal. What conditions, if any, would make a pilot ethically defensible?
Read the worked answer
Data minimization would require collecting only what is necessary for attendance, deleting images after the attendance record is verified, and avoiding cafeteria cameras if the purpose is arrival and departure. Purpose limitation would prohibit sharing attendance images with a transportation vendor without a separate, specific justification and consent. Consent is complicated in a public school because attendance is compulsory, so an opt-in pilot with a genuine alternative for students who decline is more defensible than a mandatory rollout. A defensible pilot would need independent auditing by demographic group, a published error-correction process, human review before any disciplinary action, a short retention period, and a clear ban on secondary use. It would also need parent and student notice, an accessible explanation, and a kill switch if disparities persist. Even with these conditions, the teacher could reasonably oppose deployment because the core function can be achieved with less invasive methods. Conditions reduce harm but do not erase the power imbalance.
3. Recommend a decision and outline how the teacher should explain it to students in a lesson on computing ethics.
Read the worked answer
The teacher should oppose immediate deployment and support only a time-limited, opt-in pilot with the conditions above, or oppose the pilot entirely if the committee will not accept independent auditing and deletion. The recommendation should be explicit about the tradeoff: saving 12 minutes per class is a real benefit, but it does not outweigh a documented disparity, five-year retention, and data sharing without consent. In class, the teacher should not present the decision as a simple technology-is-bad lesson. Students can map stakeholders, compute what a 4.8% error rate means in a school of 1,100 students, and compare an overall accuracy claim with group-level error rates. They can role-play the committee and write a data-minimization policy. The teacher should disclose her own position and the evidence behind it, then ask students to identify what additional evidence would change their minds. This models ethical reasoning rather than asking students to guess the teacher's opinion.