Unproctored testing offers employers greater flexibility and reduced costs than proctored testing. However, unproctored testing raises a number of concerns, most importantly cheating. Recent research has examined differences between proctored and unproctored test performance. That research shows that there are very few differences between proctored and unproctored settings. In fact, most shows that individuals in a proctored setting actually perform slightly better.
A recent study compared how candidates for a supervisor position performed in a proctored vs. an unproctored environment. They were evaluated using a web-based assessment battery that had been validated to predict supervisory job performance. The differences between the proctored and unproctored conditions were very small. In every case, the average score in the proctored environment was slightly higher than the unproctored environment. This runs counter to the concern that individuals in an unproctored environment would have an advantage – and potentially cheat the system. Some hypothesize that individuals in a proctored environment are freer from distractions and take the assessment more seriously and therefore do better.
What About Cheating?
What if someone gets help when they’re taking an unproctored test? Some tests lend themselves to cheating more than others. Tests with clear correct answers, such as a simple math test are more susceptible to cheating than tests that rely on judgment or self-report. The most problematic is a test of knowledge, where someone could research the answers. These types of tests are best administered in a proctored environment.
However, many assessments used in pre-employment testing or for development are not knowledge-based, nor do they have clear right or wrong answers. So, in those cases, how much does cheating really help?
A recent laboratory study looked at this specific question. The study used a group of 130 individuals and broke them into 4 groups: working alone or with a partner and in a timed or untimed environment.
The results indicated that:
Cognitive Ability can be improved by working in pairs, but those differences can be almost completely eliminated by including a timer.
Situational Judgment had very little differences. The partner timed group actually performed worst of all.
Personality showed almost no differences. Where there were differences the paired time groups performed worst.
These results are interesting because they show that cheating may be less helpful than originally assumed. Even for sections where there are correct answers, such as cognitive ability, adding a strict timer to the section can almost eliminate the benefits of cheating.
What It All Means
At a minimum, we can say that there are very few differences in terms of performance on proctored vs. unproctored tests, at least with regard to self-report type assessments such as biodata and personality. In addition, adding a strict timer to tests that include correct/incorrect answers can greatly reduce the impact of cheating.
This doesn’t mean that test proctoring is outdated and has no value. For tests that focus on knowledge and can be researched, then proctoring is clearly important and probably necessary. A lot depends on the purpose of the testing in the first place. Certification testing will likely always require proctoring.
These studies do not address the equally important issue of test security. Maintaining the security of test content is essential to ensure the ongoing usefulness of a test. Again, tests that focus on self-report need less protection than knowledge-based tests. Once a test of knowledge is offered online in an unproctored environment the integrity of that test has been compromised. That issue needs to be considered when deciding on whether to offer an unproctored administration.