Reference: Fiechter, J. L., Fealing, C., Gerrard, R., & Kornell, N. (2018). Audiovisual quality impacts assessments of job candidates in video interviews: Evidence for an AV quality bias. Cognitive Research: Principles and Implications, 3(1), 47.
Biases in hiring and admissions practices affect nearly all sectors of our economy. Although most concerns about bias stem from immutable characteristics of those being interviewed, such as race, religion, or age, Fiechter, Fealing, Gerrard, and Kornell (2018) hypothesized that the audiovisual (AV) quality of video interviews could also bias ratings against candidates who experience AV hiccups (temporary pauses or disruptions in the video feed).
Fiechter et al. note two fundamental reasons for investigating this hypothesis. First, the use of video interviews has become a commonplace part of the hiring process. Second, audiovisual artifacts, such as video freezing, distorted audio, and hissing, are frequently experienced during these video interviews.
Although poor audiovisual quality is common during internet interviews, AV artifacts affect each interview with varying levels of magnitude and duration. Thus, some interviews may not experience much in terms of AV interruption, while in more extreme cases interviews may need to be cut short if the AV quality is degraded too extensively.
In order to understand if varying levels of AV quality negatively bias evaluations of candidates, Fiechter et al. conducted two experiments using participants drawn from Amazon’s Mechanical Turk service.
Fiechter et al. created videos showing four actors and actresses interviewing for a position in the legal field. Each video was a little over a minute-and-a-half long and showed a single actor or actress answering basic interview questions. They then created two versions of each video. One “fluent” version was left untouched, while the second “disfluent” version of each video was degraded using editing software to add in hissing, freeze frames, and other commonly-experienced AV artifacts. Importantly, the sound associated with these interviews was never degraded in a manner that would prevent a watcher from hearing the entire contents of the interviewees’ answers. This resulted in a total of eight interview clips (available on Open Science Framework’s repository).
In the first experiment, participants were each shown four videos – one for each of the four actors and actresses. Half of these videos were the original, fluent versions. The other half of the videos were replaced with the modified, disfluent versions with decreased AV quality. The participants were not told that the audio-visual quality of the videos was being manipulated.
After watching each video, the participants in this study rated the hirability and likeability of each job candidate. The candidates in the disfluent AV quality videos were rated as less hirable than those in the unmodified videos. Interestingly, the candidates in both versions of the video were assigned approximately the same likeability ratings. This seems to indicate that AV quality problems lead candidates to be seen as less hirable, even if their personality and manner of presentation is equally professional as their competitors with better AV quality.
In a second experiment, Fiechter et al. ran essentially the same study, however, they gave participants the following warning before they began their session:
Please read carefully: You will be watching videos that are of good and poor quality. Research has shown that the quality of video or audio can impact assessments of job candidates. As you watch the interviews, try not to let video quality bias you for or against any of the candidates. (p. 3)
Essentially, the researchers sought to find out if the AV quality bias could be reduced by making the evaluators explicitly aware of their tendencies to hold the quality of candidates’ videos against them.
Interestingly, this warning did not lessen the impact of AV quality on hirability ratings. Rather, the evaluators were equally as biased against the versions of the videos with reduced AV quality as they were in the first experiment. Once again, likeability ratings were not as strongly affected by the AV artifacts introduced in the disfluent condition.
Taken together, these experiments show that AV quality can impact the way in which interview candidates are evaluated in terms of hirability. Both those being interviewed and those doing the interviewing should be aware that although the internet may enable a more geographically-diverse applicant pool to be interviewed, it may also facilitate biased evaluations of candidates experiencing AV interruptions. In this study, the researchers made sure to never remove the sound entirely from the videos, as this would reduce the amount of information that the evaluators would hear from the candidate. Nevertheless, I have found that losing the audio feed completely is a common experience with video chatting software. As such, it is possible that the bias against candidates with AV artifacts in their video interviews may be even greater in reality than was found in this study.