keepcalculator

06 Numbers and Powers

AI Proctoring: What It Sees and Can’t Prove About Cheating

Card No. 06·81914 · Updated · Standard formula, shown below

The calculator on this card is switched off while it is checked. The formula is below, and the worked example shows how to do the sum by hand.

AI proctoring software monitors students during online exams by analyzing video, audio, and screen activity, but it cannot definitively prove cheating—it only flags suspicious behaviors for human review. This article explains what the technology actually detects, its limitations, and why its findings are treated as evidence of suspicion rather than proof of misconduct.

Background: Why AI Proctoring Exists

Online education and remote testing grew rapidly, and with them came concerns about academic integrity. Traditional in-person proctors cannot watch every student at once, so institutions turned to automated systems. AI proctoring tools promise to scale oversight: they can monitor hundreds of test-takers simultaneously, flag unusual behavior, and generate reports for instructors. The technology emerged in the mid-2010s, and by the early 2020s it was widely used in universities, certification exams, and corporate training. However, its accuracy and fairness have been debated, with critics pointing to false positives and privacy concerns.

How AI Proctoring Works

AI proctoring typically combines several data streams:

  • Webcam video: Tracks eye movement, head position, and body motion. It can detect if a student looks away from the screen frequently, leaves the frame, or has another person appear.
  • Microphone audio: Listens for voices, background noise, or sudden sounds that might indicate someone is speaking to the student.
  • Screen recording: Captures what is displayed on the test-taker’s monitor, including other tabs, applications, or attempts to open unauthorized files.
  • Keystroke and mouse patterns: Some systems log typing speed, click patterns, or copy-paste actions to spot anomalies.
  • Browser lockdown: Software may restrict access to other websites or disable right-click and print functions.

During the exam, the AI analyzes these streams in real time or after the fact, assigning risk scores to behaviors. For example, a student who looks down at their lap for several seconds might get a “suspicious” flag, as might someone who reads questions aloud.

What AI Proctoring Sees

AI proctoring can see a surprising amount of detail, but it is limited to what the camera and microphone can capture. It sees:

  • Your face, including eye movements, blinking rate, and facial expressions (though it cannot read emotions accurately).
  • Your hands and desk area, if they are in frame. Some systems ask you to do a 360-degree room scan before the exam.
  • Your screen, including the browser window, other applications, and any external monitors (if the system is configured to record them).
  • Your surroundings, such as other people, books, or notes visible in the background.
  • Audio from your environment, including conversations, typing, or ambient noise.

It can also log timestamps for every action, so it knows exactly when you typed an answer or switched tabs.

What AI Proctoring Cannot Prove

Despite its capabilities, AI proctoring cannot prove cheating in a definitive sense. It can only flag behaviors that correlate with cheating, but correlation is not proof. For example:

  • Looking away could mean reading a physical note, but it could also mean thinking, looking at a window, or having a distracting pet.
  • Another person in the room might be a family member who is not helping, or a roommate who simply walks by.
  • Background voices could be a TV, a conversation elsewhere, or the student muttering to themselves.
  • Opening another tab might be an attempt to access notes, but it could also be a calculator, a translation tool, or a system pop-up.

AI systems do not have context or intent. They cannot know why a behavior occurred, only that it did. That is why most institutions require a human to review flagged incidents before taking action. The AI’s output is a suspicion score, not a verdict.

Common Problems and False Positives

False positives are a major issue. Students with disabilities, such as ADHD or anxiety, may fidget or look around more. Those with poor internet connections may experience lag, causing the camera to freeze or the audio to drop. Bright lighting can make a student appear to be squinting, which the AI might interpret as looking at something off-screen. Cultural differences also play a role: in some cultures, avoiding eye contact is a sign of respect, but the AI may flag it as evasive. A 2020 study by the University of Michigan found that AI proctoring systems disproportionately flagged students with darker skin tones, due to facial recognition errors. These issues have led to lawsuits and policy changes in some institutions.

How to Choose an AI Proctoring System

If you are an educator or administrator considering AI proctoring, consider these factors:

  • Transparency: Does the vendor explain what behaviors are flagged and how? Can students see the report?
  • Human review: Does the system require a human to confirm flags before they are reported to instructors?
  • Accessibility: Does it accommodate students with disabilities? Can they request alternate arrangements?
  • Data privacy: How long is video and audio stored? Who has access? Is the data encrypted?
  • Cost and scalability: Is the pricing per student or per exam? Does it handle large classes without performance issues?
  • Integration: Does it work with your learning management system (e.g., Canvas, Blackboard)?

Always run a pilot test with a small group and gather feedback before committing.

Alternatives to AI Proctoring

Some institutions are moving away from AI proctoring due to privacy and fairness concerns. Alternatives include:

  • Open-book exams: Design questions that require application and analysis, not memorization, so external resources are less helpful.
  • Oral exams: Conduct live video interviews where students explain their answers, which is harder to fake.
  • Timed, low-stakes assessments: Use frequent short quizzes instead of a single high-stakes exam, reducing the incentive to cheat.
  • Lockdown browsers only: Use software that restricts access to other websites without recording video or audio, which is less invasive.
  • Proctored centers: For high-stakes exams, students can go to a physical testing center with human proctors.

Each approach has trade-offs between integrity, convenience, and privacy.

Costs and Privacy Considerations

AI proctoring services are typically subscription-based, costing anywhere from a few dollars per student per exam to thousands of dollars per year for institutional licenses. The price often includes technical support, data storage, and reporting features. However, the true cost may include student backlash, legal risks, and the time spent reviewing flags. Privacy is a significant concern: video and audio recordings are sensitive data. Check that the vendor complies with regulations like GDPR or FERPA, and that they do not sell or share data with third parties. Students should be informed about what is recorded and how long it is kept.

Frequently Asked Questions

Can AI proctoring detect phones?

It can detect a phone if it is visible in the webcam frame or if the student picks it up, but it cannot detect a phone hidden under a desk or in a pocket. Some systems use radio frequency detection, but that is rare and not foolproof.

Does AI proctoring record your screen?

Yes, most systems record the screen, but only if you have granted permission. The recording captures the active window, and sometimes all monitors if the software is configured to do so. It cannot see what is behind other windows unless it takes periodic screenshots of the entire desktop.

Can AI proctoring tell if you are using a virtual machine?

Some systems attempt to detect virtual machines, but it is not always reliable. A virtual machine can sometimes bypass browser lockdowns, but the AI may flag unusual system behavior. However, using a VM is not proof of cheating by itself.

What happens if the AI flags you?

Typically, a human proctor or instructor reviews the flagged incident. They will watch the video segment and decide whether it warrants further investigation. You may be asked to explain the behavior. If the instructor believes cheating occurred, they may follow the institution’s academic integrity policy, which could include a hearing or a penalty.

Can AI proctoring detect eye movement accurately?

It can track eye position relative to the screen, but accuracy varies. It cannot determine exactly what you are looking at, only whether your gaze is directed at the screen or away. Factors like lighting, glasses, and camera quality affect accuracy.

Is AI proctoring legal?

It is legal in most jurisdictions, but it must comply with privacy laws. Some states or countries have restrictions on biometric data collection. Institutions are required to inform students about the monitoring and obtain consent. Legal challenges have arisen over privacy and discrimination, so the regulatory landscape is evolving.

Related

Formula used

Each line is one input (in, sel, list) or one result (out: label, unit, formula, decimals). How the calculators work

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.

More cards in this drawer

The whole Numbers and Powers drawer →