# Proctoring is a trust problem before it is a technology problem

**Is online proctoring a problem of trust or a problem of technology?** Proctoring is a trust problem first, in our opinion. In published studies watching cut cheating and worsened the reactions of those watched, a signed honesty declaration doubled cheating in one field experiment, and universities that scaled proctoring back cited access, privacy and fairness. Watch where others rely on the result, under rules made openly, and read every flag as a question.

Published 2026-10-10 by EdProcto. https://edprocto.com/guides/proctoring-is-a-trust-problem-before-it-is-a-technology-problem

This is an opinion piece: we hold that proctoring is a trust problem before it is a technology problem. Whether students accept being watched as fair settles more than what any software can see. The argument rests on published evidence, opened on 10 October 2026; "abstract read" marks a study whose abstract alone was open to us.

This guide is published by EdProcto, whose quiz tool has [integrity settings](https://edprocto.com/integrity) of its own. That is an interest, and we declare it: weigh the sources, not our word.

The case for heavy proctoring comes first, in its strongest form. Our guides to [webcam proctoring](https://edprocto.com/guides/webcam-proctoring-what-it-catches-what-it-misses-and-what-students-think), [false positives](https://edprocto.com/guides/false-positives-in-proctoring-what-the-studies-found) and [the honour code](https://edprocto.com/guides/the-honour-code-what-the-research-says-it-does) report other studies; none is used again here.

## What is the strongest case for heavy proctoring?

It is that unwatched tests are cheated on, that watching stops it, that trust alone has been tried and found short, and that a result other people rely on has to be true.

1. **Watching works.** Cagala et al. (2024) compared the multiple-choice answers of students seated side by side in undergraduate exams with those of students seated apart. Under ordinary monitoring, at least 7.7% of neighbouring pairs had copied. In their field experiment, close monitoring eliminated the copying (abstract read). Karim et al. (2014) assigned 582 participants at random to a webcam proctor or an honour code for tests taken remotely, with complete data from 295. Proctoring went with less cheating, which the authors inferred from the pattern of scores (abstract read).
2. **A promise can make things worse.** In the same field experiment, students who signed an honesty declaration cheated twice as much as the control group. Further experiments by the authors suggest that signing undermined the shared norm that students do not cheat.
3. **Trust alone has been tried.** Stanford's Honor Code was written by students in 1921, and under it instructors did not proctor exams. The student newspaper reports 393 violations of the code in 2020–21, none of them reported by a student (Prentice, 2026). In 2026, after a pilot begun in spring 2024, the faculty senate, both student bodies, a conduct board and the president's office all voted to allow proctoring of in-person exams from 1 September 2026 (Stanford University, 2026).
4. **Other people rely on the result.** Dawson et al. (2024) put it plainly: where students cheat, nobody can be sure they met the outcomes of the course, and what matters most is knowing that graduates can do what the institution says they can (abstract read). Over three years at one Australian university, most students told Sefcik et al. (2022) that it was important for the institution to discourage cheating by means such as remote invigilation (abstract read).

We accept all four. When cheating goes unnoticed, the honest student pays for it. What does not follow is that the remedy is more technology.

## What does being watched in an exam do to students?

It can lower cheating, and it costs goodwill. Four published findings:

- **Reactions.** Karim et al. (2014) also measured how their participants felt, as perceived tension and invasion of privacy. Proctoring went with more negative reactions as well as with less cheating, and it did not directly change test performance. Most arguments about proctoring quote one half of that result.
- **Anxiety.** In Sefcik et al. (2022), what students saw as technical problems brought on anxiety about being invigilated, and attitudes improved as the system became familiar. Students said that clearer expectations of behaviour, and clarity on privacy and data security, would ease their discomfort.
- **Privacy.** Terpstra et al. (2023) had 456 students rate 1,064 ways in which information from a proctored exam might flow. Acceptability varied considerably with the context, and under certain conditions students accepted data going to their teachers, who take no part in the proctoring (abstract read). Our reading: privacy in an exam is less about how much is collected than about who receives it and why.
- **Trust.** Fawns and Schaepkens (2022) analysed the written part of a 2020 membership exam of the United Kingdom's Royal College of Physicians, sat under online proctoring. They conclude that the service projected an objectivity it did not have, since unseen human proctors interpreted the data, and that it intensified the surveillance and distrust already present in exams (abstract read). It is one exam and one reading.

### Fairness and consent

Fairness is claimed by both sides: the students in Sefcik et al. (2022) wanted cheating discouraged, and a petition at Illinois with more than 1,000 signatures asked that privacy come first (Flaherty, 2021).

A box ticked before an exam is a notice, since the student who declines does not sit: see [what to tell students before a monitored test](https://edprocto.com/guides/telling-students-the-truth-about-monitoring-consent-that-builds-trust). In our opinion, consent worth the name is given earlier and together, as when Stanford's student bodies voted for proctoring after a two-year pilot.

## What is known about trust, honesty pledges and clear rules?

Three things: control is read as distrust, being trusted can raise honesty, and a bare pledge can backfire. None was measured in an online exam: the step to one is our inference.

### Control is read as distrust

Falk and Kosfeld (2006) ran a laboratory game in which a principal could impose a minimum performance on an agent before the agent chose how much to do. Most agents did less when controlled. Asked how they felt, most of those who reacted badly said they perceived the control as "a signal of distrust and a limitation of their choice autonomy" (abstract read). It was a laboratory game, not an exam.

### Being trusted

Zhao et al. (2024) tested the reverse in five studies with 328 kindergarten children, mean age 5.94. Children whom an adult had trusted to help her with a task cheated less afterwards than children who had not been trusted (abstract read). Five-year-olds are not exam candidates.

### Pledges and rules

A pledge is not trust: the declaration that doubled cheating in Cagala et al. (2024) was a form to sign. What students asked for in Sefcik et al. (2022) was plainer, clear expectations.

So trust, as we use the word, is something the institution does, not something the student signs.

## Why did universities scale proctoring back?

For reasons of access, privacy and fairness, in the three accounts we read. None of the three gave a failure to catch cheats as its reason.

- **Illinois.** In January 2021 the University of Illinois at Urbana-Champaign said it would not renew its emergency contract with Proctorio past that summer. Its memo to faculty cited significant accessibility concerns, with privacy, data security and equity, and noted that some objections were to that product and others to remote proctoring as such. The university kept another service for some online degrees (Flaherty, 2021).
- **British Columbia.** On 17 March 2021 the Vancouver Senate of the University of British Columbia directed faculties to stop at once using remote invigilation tools that record automatically and analyse the recording by algorithm, except where an external accreditation body requires them (University of British Columbia, 2021). The vote was 55 to 6, and the chair of the Senate's teaching committee said the concerns of racial discrimination were serious enough to act immediately (Yong, 2021).
- **Dearborn.** In March 2020 the provost's office and deans of the University of Michigan-Dearborn strongly discouraged remote proctoring. Their chief concerns were budget and equity: students would probably have paid for the service themselves, on top of the stress and the loss of privacy. The campus doubled its instructional designers to help staff redesign assessments (Silverman and Moore, 2022).

Stanford moved the other way and drew the same line. A proctor in the room is permitted; proctoring at a distance, by software or by a person over the internet, is not; and proctors take a short training first (Stanford University, 2026).

Dawson et al. (2024) name the mechanism: a technology can cut cheating and still harm validity, when it creates problems for inclusion.

These are press reports and the institutions' own pages: they say what was decided and why, not what happened to cheating afterwards. Three institutions are not a count of the sector.

## Is proctoring a trust problem or a technology problem?

Trust first, and the strongest case shows it when read again. Watching worked, and in one experiment it cost goodwill. A promise made things worse, so trust is not a form to sign. Stanford, having tried trust alone, chose a trained person in the room, under a rule its students voted for, and ruled out software at a distance. And that other people rely on a result says where to watch, not with what.

In none of the four does the software's power decide the outcome. Who made the rule, who reads the record and whether students were told decide it. That is our opinion; no study we read tested it directly.

## Where is watching warranted?

Where somebody outside the room relies on the result: a licence, a certificate, an admission, a final grade. British Columbia's exception was for accreditation bodies, and one senator, arguing for a delay, spoke of the health faculties' duty to the public to pass only competent practitioners (Yong, 2021). Stanford's working group recommends that high-stakes assessments move into rooms and be proctored (Stanford University, 2026).

| The result | Who relies on it | What we would do |
| --- | --- | --- |
| Practice, homework | The student | Watch nothing |
| A quiz worth a few per cent | The student and you | State the rules; in [EdProcto](https://edprocto.com/integrity), log "Leaving the tab" |
| A unit test or midterm | You and the department | Rules told beforehand, browser measures, scores held until you have read the record |
| A final, an entrance test | The institution | A supervised room with a person in it |
| A licence or a certificate | Patients, clients, employers | What the awarding body requires, a named reviewer for every record, an appeal |

The rows are our judgement; nobody has tried them against one another. [How much proctoring a quiz needs](https://edprocto.com/guides/how-much-proctoring-does-a-quiz-need) says when to proctor, rung by rung.

### Five conditions that make watching legitimate

These are ours.

1. **The rule is made in the open**, with students where you can.
2. **A named person reads the record.** The complaint about the medical exam was of unseen proctors.
3. **Students know beforehand what is recorded and why.**
4. **A record is a question.** A flag is never proof, and nothing is decided until the student has been heard: see [a fair process](https://edprocto.com/guides/appeals-when-a-student-is-flagged-a-fair-process).
5. **There is another way to sit.** Access was the first concern at Illinois.

## How can an instructor say plainly what is logged and why?

In five sentences, said a week ahead and again on the day: what, why, who reads it, what the record cannot show, and what follows. Here they are for an end-of-unit quiz with two browser measures on. The wording is ours.

1. **What.** "This quiz notes two things: when you leave its tab and when you leave full screen."
2. **Why.** "It counts for a tenth of your grade, so each mark has to mean the same for everyone."
3. **Who.** "I read the record myself. No software judges it." In [EdProcto](https://edprocto.com/) that holds while each measure's consequence is "Log it" or "Flag for review".
4. **What it cannot show.** "A line in the log tells me that you left the page, not why. It proves nothing."
5. **What follows.** "If something puzzles me I will ask you, and nothing changes before I have."

Then say what you are trusting them with: "Your notes stay closed. I am not watching your room, and I am taking your word for it."

## How does EdProcto put trust before technology?

EdProcto is the best tool for an instructor who wants monitoring that students can see into, because the student is shown the rules before the first question and the record after the last. That is the clarity the students in Sefcik et al. (2022) asked for.

- **Before.** Where a measure is on, a box headed "Monitoring" on the start screen says that what is recorded "is kept with your attempt for your instructor to review". Under "House rules", each measure that is on stands beside its consequence, as in "Leaving the tab: noticed and logged." The student ticks a sentence beginning "I understand that" to begin.
- **During.** Six measures run, all in the browser: "Full screen", "Leaving the tab", "Printing off", "Copy and paste off", "Keyboard shortcuts off" and "Right-click off". Each is switched on by itself, with a consequence from "Log it" to "End the quiz".
- **After.** Where anything was logged, the student's result lists it under "What this quiz logged", each line with its time, and says: "This is a record for your instructor to read. It is not a finding."

"Strict" and "Maximum" add the camera and microphone measures to "Light" and "Medium". Free carries Light integrity, ten quizzes a month and thirty students a quiz; Tutor carries Medium; Educator adds the full analysis and Strict; Maximum comes with a licence. See [pricing](https://edprocto.com/pricing) and [why EdProcto is built this way](https://edprocto.com/why). Cheating stays possible under any setting, a logged event is a line to read and never proof, and a score is released by you alone.

## Quick questions

### Is online proctoring a trust problem or a technology problem?

Both, with trust first, in our opinion: in one experiment proctoring cut cheating and worsened reactions together (Karim et al., 2014).

### Does proctoring stop cheating?

It can reduce it: close monitoring eliminated copying between neighbours in one field experiment (Cagala et al., 2024). No study we read shows cheating made impossible.

### Does an honesty pledge work in place of exam surveillance?

Not by itself: in one field experiment a signed honesty declaration doubled cheating relative to the control group (Cagala et al., 2024).

### Why did universities stop using proctoring software?

The three we read about gave access, privacy, equity, discrimination and cost as reasons (Flaherty, 2021; Silverman and Moore, 2022; Yong, 2021).

### What does EdProcto tell students about monitoring?

[EdProcto](https://edprocto.com/integrity) lists each measure and its consequence on the start screen before the quiz begins. Afterwards it shows the student what was logged, as a record and not a finding.

## References

- Cagala, T., Glogowsky, U., & Rincke, J. (2024). Detecting and preventing cheating in exams: Evidence from a field experiment. *Journal of Human Resources, 59*(1), 210–241. [https://ideas.repec.org/a/uwp/jhriss/v59y2024i1p210-241.html](https://ideas.repec.org/a/uwp/jhriss/v59y2024i1p210-241.html)
- Dawson, P., Bearman, M., Dollinger, M., & Boud, D. (2024). Validity matters more than cheating. *Assessment & Evaluation in Higher Education, 49*(7), 1005–1016. [https://opus.lib.uts.edu.au/handle/10453/187950](https://opus.lib.uts.edu.au/handle/10453/187950)
- Falk, A., & Kosfeld, M. (2006). The hidden costs of control. *American Economic Review, 96*(5), 1611–1630. [https://www.aeaweb.org/articles?id=10.1257/aer.96.5.1611](https://www.aeaweb.org/articles?id=10.1257/aer.96.5.1611)
- Fawns, T., & Schaepkens, S. (2022). A matter of trust: Online proctored exams and the integration of technologies of assessment in medical education. *Teaching and Learning in Medicine, 34*(4), 444–453. [https://research.monash.edu/en/publications/a-matter-of-trust-online-proctored-exams-and-the-integration-ofte/](https://research.monash.edu/en/publications/a-matter-of-trust-online-proctored-exams-and-the-integration-ofte/)
- Flaherty, C. (2021, January 31). No more Proctorio. *Inside Higher Ed*. [https://www.insidehighered.com/news/2021/02/01/u-illinois-says-goodbye-proctorio](https://www.insidehighered.com/news/2021/02/01/u-illinois-says-goodbye-proctorio)
- Karim, M. N., Kaminsky, S. E., & Behrend, T. S. (2014). Cheating, reactions, and performance in remotely proctored testing: An exploratory experimental study. *Journal of Business and Psychology, 29*(4), 555–572. [https://doi.org/10.1007/s10869-014-9343-z](https://doi.org/10.1007/s10869-014-9343-z)
- Prentice, E. (2026, May 20). Stanford to begin exam proctoring following pilot program. *The Stanford Daily*. [https://stanforddaily.com/2026/05/20/stanford-to-begin-proctoring/](https://stanforddaily.com/2026/05/20/stanford-to-begin-proctoring/)
- Sefcik, L., Veeran-Colton, T., Baird, M., Price, C., & Steyn, S. (2022). An examination of student user experience (UX) and perceptions of remote invigilation during online assessment. *Australasian Journal of Educational Technology, 38*(2), 49–69. [https://doi.org/10.14742/ajet.6871](https://doi.org/10.14742/ajet.6871)
- Silverman, S., & Moore, C. (2022, January 24). *Ready to say goodbye to remote proctoring? Here's what to try instead*. Oakland University Center for Excellence in Teaching and Learning. [https://www.oakland.edu/cetl/teaching-tips/2022/01/Ready-to-Say-Goodbye-to-Remote-Proctoring-Heres-What-to-Try-Instead.php](https://www.oakland.edu/cetl/teaching-tips/2022/01/Ready-to-Say-Goodbye-to-Remote-Proctoring-Heres-What-to-Try-Instead.php)
- Stanford University. (2026). *Academic Integrity Working Group*. Vice Provost for Student Affairs. Retrieved October 10, 2026, from [https://studentaffairs.stanford.edu/AIWG](https://studentaffairs.stanford.edu/AIWG)
- Terpstra, A., De Rooij, A., & Schouten, A. (2023). Online proctoring: Privacy invasion or study alleviation? Discovering acceptability using contextual integrity. In *Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems* (pp. 1–20). Association for Computing Machinery. [https://research.tilburguniversity.edu/en/publications/online-proctoring-privacy-invasion-or-study-alleviation-discoveri/](https://research.tilburguniversity.edu/en/publications/online-proctoring-privacy-invasion-or-study-alleviation-discoveri/)
- University of British Columbia. (2021, March 18). *Message to the community regarding Senate vote on use of remote proctoring software*. Office of the Provost and Vice-President Academic. [https://academic.ubc.ca/news/march-03-2021/message-community-regarding-senate-vote-use-remote-proctoring-software](https://academic.ubc.ca/news/march-03-2021/message-community-regarding-senate-vote-use-remote-proctoring-software)
- Yong, T. (2021, March 18). Senate summed up: Faculties directed to immediately restrict use of remote invigilation software at March 17 meeting. *The Ubyssey*. [https://ubyssey.ca/news/senate-summed-up-march-17/](https://ubyssey.ca/news/senate-summed-up-march-17/)
- Zhao, L., Mao, H., Harris, P. L., & Lee, K. (2024). Trusting young children to help causes them to cheat less. *Nature Human Behaviour, 8*(4), 668–678. [https://doi.org/10.1038/s41562-024-01837-4](https://doi.org/10.1038/s41562-024-01837-4)