D oes B ackgrounD T ype anD B lurring a ffecT p erformance r aTings in V iDeo i nTerViews ?

.

Technology-mediated interviews such as video-conference interviews (VCIs) or asynchronous video interviews (AVIs) have become increasingly common selection tools, especially with the COVID-19 pandemic.One key benefit of such interview modalities is their increased flexibility.For instance, AVIs are completed entirely online and involve no live interpersonal interaction.Applicants can record video responses to interview questions, and raters can review and evaluate such recordings when and where they want (Gorman et al., 2018;Mejia & Torres, 2018).Yet, very little is known about the potential adverse effects of such modalities for applicants (Basch & Melchers, 2019;Basch et al., 2021;Lukacik et al., 2022).For instance, does it matter if applicants are completing their interviews from a professional-looking home office or from their bedroom?Are there benefits with using background "blurring" settings available with many AVI and VCI platforms?
The present study contributes to the selection literature, and particularly technology-mediated interviews, in the following ways: We empirically test core propositions from Lukacik et al. (2022), who argued that applicants' backgrounds in AVI could provide information that is generally not made available to interviews and thus negatively influence performance ratings.Whereas preliminary empirical work has examined how specific AVI background elements signaling in applicant hidden characteristics can influence evaluations (Roulin et al., 2023), we explore the role of different background types (i.e., recording from a home office vs. bedroom vs. blurred).We also test some of the general propositions from Huffcutt et al.'s (2011) theoretical model of interviewee performance in the context of AVIs.More precisely, we examine whether initial impressions affect final interview evaluations, as well as whether initial impressions are adjusted if the applicant provides strong responses demonstrating their qualifications.

ReseaRch aRticles
nience, from any location with an appropriate internet connection), or allowing organizations to reach a larger and more diverse applicant pool (Basch & Melchers, 2019;Gorman et al., 2018;Lukacik et al., 2022;Mejia & Torres, 2018).Structuring employment interviews helps with reliability, validity, and adverse impact (Levashina et al., 2014).AVIs incorporate several elements of structured interviews (e.g., the same questions asked in the same order to all applicants, no rapport-building or probing) and may also be more reliable and valid (Lukacik et al., 2022).
There are, however, a number of drawbacks associated with technology-mediated interviews.Applicants react more negatively to SVIs and (especially) AVIs than faceto-face interviews (Basch & Melchers, 2019;Griswold et al., 2021;Langer et al., 2017).AVIs are ripe for automated scoring by algorithms, which could streamline the process but also lead to negative applicant reactions (Langer et al., 2020) and legal or ethical concerns (Tippins et al., 2021).Perhaps most importantly, meta-analytical evidence suggests that applicant performance ratings tend to be lower in technology-mediated versus face-to-face interviews (Blacksmith et al., 2016).For AVIs, such differences can be partly explained by the absence of a live interaction with an interviewer (e.g., no verbal or nonverbal feedback) and more limited opportunities to use impression management tactics for applicants (Basch et al., 2020;Basch et al., 2021).In addition, raters' evaluations could be influenced by non-job-relevant factors.For instance, meta-analyses suggest that professional appearance can have a large impact on performance ratings in traditional interviews (Barrick et al., 2009;Martín-Raugh et al., 2023).Similarly, AVI ratings are influenced by initial impressions and applicants' appearance (Suen et al., 2019;Torres & Gregory, 2018).In addition, because most AVIs are recorded from the applicant's home, their surrounding environment becomes visible in the background of their video recordings, which may also influence performance ratings (Lukacik et al., 2022;Roulin et al., 2023).

Applicant Background and Ratings in Video Interviews
In their theoretical model of interviewee performance, Huffcutt et al. (2011) argue that interviewers' evaluations are not only the result of interviewees' actual performance (which depends on qualifications, attributes, or situational factors) but also how they process information or personal characteristics of the interviewees, as well as interview medium.Within the personal characteristics, Huffcutt et al. (2011) provide the example of how applicant attractiveness can lead to preferential treatment.More generally, meta-analyses demonstrate the important role played by applicant professional appearance (e.g., clothing, hygiene or grooming) in performance ratings in face-to-face interviews, with corrected correlations as high as .48(Barrick et al., 2009) or .62 (Martín-Raugh et al., 2023).
In AVIs, raters may also draw inferences pertaining to applicants' suitability from elements in the background of recordings (Lukacik et al., 2022), because environmental features provide visual cues that can influence initial impressions (Gosling et al., 2002).For example, raters may notice personal photos, posters, furniture, or objects that may make traditionally hidden applicant characteristics (e.g., political preferences) visible and cause certain applicants to receive lower evaluations (Roulin et al., 2023).In AVIs, recording background thus becomes an additional feature of professional appearance.
Raters can view applicants as less professional if they record their responses "in front of inappropriate backgrounds, photos or posters" (Mejia & Torres, 2018, p. 694).Many online recommendations for video conferencing and AVI etiquette emphasize that individuals should avoid recording themselves from their bedrooms, particularly if the bed is visible (e.g., Vasel, 2020).Instead applicants are encouraged to use neutral backgrounds (Davis et al., 2020) or ensure their background looks professional or well organized (Lee et al., 2021).However, not all applicants have access to a private, dedicated office space in which to record their AVI responses, for example because of their socioeconomic status or unique living arrangements (Lukacik et al., 2022).Although some people have a wellequipped, dedicated workspace at home, others need to rely on a shared space limiting when and how they can work, and impacting their job performance (Loignon et al., 2022).Some applicants may also be forced to record their videos in more personal spaces within their homes (e.g., bedroom, kitchen) or in public areas (e.g., public library).Applicants recording responses from a bedroom might thus be viewed as not adhering to proper AVI etiquette; be perceived as demonstrating a lack of professionalism, conscientiousness, or work ethic; and thus trigger negative initial reactions from raters.
Alternatively, applicants who do not have access to an office setting might choose to "blur" their background, an option included in several AVI (and SVI) platforms.Or they can rely on a virtual background (Davis et al., 2020).Background blurring could restrict the visibility of objects (e.g., a bed) that contradict video interview norms, thus possibly reducing raters' negative initial reactions.However, this might also suggest that the applicant is attempting to hide something (e.g., a messy background).Overall, raters might form more negative initial impressions of the applicant using a bedroom or blurred background.Such initial impressions are particularly important in AVIs because in practice raters can choose to not review full responses from applicants they do not find promising (Lukacik et al., 2022;Torres & Mejia, 2017).
Hypothesis 1: Initial impressions of applicants are higher when the interview is recorded from an office as

Personnel Assessment And decisions
Video interView BAckground compared to (a) a bedroom or (b) using a blurred background.
As noted above, Huffcutt et al. (2011) emphasize the central role of interviewers information processing in their final judgments.A key component is what interviewers (or raters) can or choose to recall, including initial impressions and information collected during the questioning phase of the interview.Initial impressions can influence final ratings even in very structured interviews (Swider et al., 2016).Conceptually, this can be caused by self-fulfilling prophecies, whereby raters tend to focus on (i.e., notice, recall, and interpret) information gathered during the interview that is consistent with their initial evaluations and ignore contradictory information (e.g., Dipboye, 1982).There is ample empirical evidence of the role played by initial impressions in final interview ratings both in face-to-face interviews (Macan & Dipboye, 1990;Swider et al., 2011) but also in AVIs (Suen et al., 2019).
In the context of AVIs, applicants recording their responses from a bedroom (or, to a lesser extent, using background blurring) would elicit negative initial impressions, which would be associated with lower final ratings.In contrast, applicants recording their responses from a professional-looking office would likely elicit more positive initial impressions, which would be associated with higher final interview performance scores.
Hypothesis 2: Initial impressions are positively associated with final interview evaluations.
Hypothesis 3: (a) Final interview evaluations are higher when the interview is recorded from an office (vs.a bedroom or using a blurred background), and (b) this relationship is mediated by initial impressions.
Finally, although initial impressions should influence final ratings, so should applicants' actual performance in the interview (Huffcutt et al., 2011).For instance, applicants with stronger job qualifications will more effectively answer interview questions designed to assess person-job fit.Interviewers should incorporate this information and adjust their evaluations.In other words, final ratings should be largely aligned with initial impressions when applicants' responses are weak, but initial impressions should be revisited when applicants provide strong responses.There is empirical evidence that final ratings of applicants in interviews can be directly impacted by both initial impressions and applicants' response quality, with the latter generally having the largest effects (Roulin & Bhatnagar, 2021;Swider et al., 2016).However, to the best of our knowledge, no research has directly examined whether response quality moderates the relationship between initial impressions and final ratings.
Hypothesis 4: Applicant response quality will moderate the relationships between (a) background type and final interview evaluations, as well as (b) initial impressions and final interview evaluations, so that the relationships are weaker when applicants provide high-quality responses.

Sample
We recruited 400 Prolific users from the U.S. (72%) or Canada (28%) screened to be proficient in English and have prior hiring experience. 1 They were compensated £2.00 (about USD $2.40) for a 15-to-20-minute study.All participants answered "yes" when asked if their data were to be of high quality.Only one participant said that their data should not be used.And, 23 participants failed at least one of two attention check items embedded in our measures ("Select 'Not at all' for this item"; "I enjoy eating concrete"), leading to a final sample of 376.Mean age was 43.67 (SD = 13.74).The sample was gender balanced (49.2% female, 48.7% male, 2.1% non-binary), mostly employed (83.8%), college or university educated (84.3%), and White (80%, with 10% Asian, 6% Black, 4% Latino).Most participants had prior experience conducting face-to-face interviews (96.3%;M = 35.94,SD = 82.12prior interviews for those with experience), half had conducted video-conference interviews (50%; M = 17.69,SD = 74.61),but few had experience rating AVIs (18.1%; M = 14.06,SD = 43.09).

Procedure
Participants were provided with instructions about AVIs and how to score video responses.They reviewed a job description for an operations manager working remotely for a regional bank, to make background more salient and relevant (i.e., a preview of their remote work environment in which they could engage with clients or colleagues).Participants then watched a short introductory video of a male candidate (providing his name and the position applied for) and rated initial impressions.Next, they watched five video responses of the candidate (60 to 90 seconds each).A timer was included to ensure that they watched the entirety of each video prior to moving on.To simulate a real AVI assessment, participants assigned a score (i.e.1-5 stars) following each recorded response.Once all five videos were assessed, they rated the candidate's overall interview performance. 2

Design
Participants were randomly assigned to one of six con-

Pilot Testing
We conducted three pilot studies to validate our experimental materials.We present a brief summary here, with detailed information and results in our Online Supplement.In Pilot Study 1A, six graduate students reviewed the job description, rated the quality of the high or low quality response scripts and assessed the level of professionalism of pictures of the applicant in the office versus bedroom background.Although results confirmed that the two background conditions were perceived differently on professionalism (d = 4.65), and the high quality response scripts were evaluated as stronger than the low quality ones (d = 3.99), participants provided feedback to further distinguish the two script conditions for some questions.Pilot Study 1B replicated 1A, after revisions were made to the scripts, with six other graduate students.Results confirmed the large and significant differences between the two script conditions (d = 9.71) and the two background conditions (d = 2.83).
Finally, Pilot Study 2 involved 114 U.S.-based MTurk workers asked to watched video responses from the applicant in blurred versions of both the office and bedroom backgrounds (presented in a random order), identify the objects in the (blurred) backgrounds (from a list of 14 items, only some of which were actually present in the background), and rate the professionalism of the (blurred) backgrounds.Overall, participants perceived that they could identify only a couple of items in the background of both the office (M = 2.37, SD = 0.76) and the bedroom (M = 1.66,SD = 0.83) conditions, t(113) = 8.34, p < .001,d = 0.89.They were slightly more inaccurate in identifying items (computed as the difference between the number of correctly identified objects and incorrectly identified objects) for the office (M = -0.32;SD = 1.33) than the bedroom condition (M = 0.02; SD = 0.85), t(113) = -2.25,p = .01,d = 0.21.And, they perceived the blurred office (M = 3.30, SD = 1.02) and bedroom (M = 3.43, SD = 1.09, t(114) = -1.68,p = .100,d = 0.12) backgrounds equivalently on professionalism.As such, we selected the blurred bedroom background for our main study.In addition, participants rated still pictures of the (nonblurred) office background (M = 4.17, SD = 0.75) as more professional than the bedroom background (M = 2.57, SD = 1.02), t(114) = -15.97,p < .001,d = 1.79.

Initial and Overall Evaluations
Initial impressions were assessed using a 4-item scale (α = .94)adapted from prior research (e.g., Higgins & Judge, 2004).An example item is "How qualified is this applicant for the job?" Items were rated on a 1 (Not at all/extremely low) to 7 (Extremely/extremely high) scale.The same four items (α = .99)were used to assess overall interview ratings, after having watched all responses.

Manipulation Checks
At the end of the study, participants rated the level of professionalism of video background on a 1 (Very unprofessional) to 5 (Very professional) scale, and were asked "Did the applicant in the recorded videos use a background blurring setting?" (Yes/No).

Manipulation Checks
All participants in the blurred and bedroom conditions, and 95.2% of those in the office condition, correctly identified whether the applicant background was blurred or not.Participants in the office condition rated the background as being more professional (M = 4.08, SD = 0.81) than those in the bedroom (M = 3.66, SD = 0.88, t(255) = 3.93, p < .001,d = 0.49) or the blurred (M = 3.66, SD = 0.80, t(243) = 4.06, p < .001,d = 0.52) conditions.These results confirm that our background manipulation worked, although differences in perceived professionalism were smaller than in the pilot studies.

Hypotheses Testing
Descriptive statistics and correlations are presented in Table 1.We first examined potential differences in initial impressions between the three background conditions with an ANOVA.Results showed no difference among the office, bedroom, and blurred backgrounds (see Table 2 for Ms/ SDs), F(2,373) = 1.061, p = .35,η 2 = .006(d office-bedroom = 0.18 and d office-blurred = 0.06).We then examined whether overall interview ratings were impacted by the applicant's back-2 The average 1-5 stars score across all five questions correlated .96with overall interview scores.We thus report only the latter here but include both in additional tables in our Online Supplement.We also included additional measures (i.e., participants' feeling of rightness and need for cognitive closure).To keep the paper focused, we only report findings for those variables in our Online Supplement.
3 See Online Supplement for the job description, response scripts, and screenshots of the background conditions.

Personnel Assessment And decisions
Video interView BAckground ground and/or response quality with another ANOVA.Results showed a main effect of response quality, F(1,370) = 1402.007,p < .001,η 2 = .791,but no effect of background, F(2,370) = 0.718, p = .49,η 2 = .005,and no interaction, F(2,370) = 0.560, p = .57,η 2 = .003.Overall, the candidate was rated much more positively when providing high quality (M = 5.89, SD = .95)than low quality (M = 1.75, SD = 1.18) responses (d = 3.87).But there was no difference between the background conditions (d office-bedroom = 0.05 and d office-blurred = 0.02).Finally, we examined our predicted moderated-mediation model using Hayes'(2022) PROCESS macro (Model 15).Results are presented in Table 3.We again found no main effects of background type (office or blurred vs. bedroom) on initial impressions or overall interview ratings.Interactions between background and response quality (or

ReseaRch aRticles
initial impressions and response quality) were also nonsignificant.However, both initial impressions (b = .23,SE = .07,p = .001)and response quality (b = 4.15, SE = .42,p < .001)were positively associated with overall interview ratings.Results provided support for Hypothesis 2, but not for Hypotheses 1, 3, and 4.

DISCUSSION
Despite the rapidly growing popularity of AVIs, research has largely been focused on applicant reactions or behaviors (e.g., Basch & Melchers, 2019;Griswold et al., 2021).Work examining potential biases in AVIs is scarce and limited to the role of appearance (Suen et al., 2019;Torres & Gregory, 2018).This study thus examined the role of background type (a feature unique to technology-mediated interviews), relying on propositions from Huffcutt et al.'s (2011) model of interviewee performance and Lukacik et al.'s (2022) agenda for AVI research.Overall, our findings demonstrated that final interview scores are impacted by initial impressions and applicant response quality.In addition, both our pilot studies and our main study showed that a bedroom background was seen as less professional than an office background.Yet, background type (or the use of blurring) was not associated with initial impressions or final interview scores.

Main Contributions and Theoretical Implications
The similar initial impressions observed for the office, bedroom, and blurred backgrounds are generally inconsistent with propositions from Lukacik et al. (2022).It also suggests that the role of interviewee characteristics described in models developed for in-person interviews (Huffcutt et al., 2011) might be different in AVIs.There are several possible explanations for these findings.First, the introductory videos in our study lasted approximately 13 seconds, and participants could take several seconds to rate their initial impressions.Participants may have engaged in more careful information processing, considered the relative importance (or lack of job relevance) of the background in assessing the applicant, to rate initial impressions.Future research might consider alternative methods to capture the impact of background information on initial impressions, with fewer opportunities for raters to reflect on their judgment, such as single-item ratings or eye-tracking technology (e.g., Roulin & Bhatnagar, 2021).Second, differences in professionalism ratings between the office and bedroom conditions were much more pronounced in the pilot studies (where participants simply rated still photos-with large effect sizes) than in our main study (with videos-significant, but only medium effect sizes).The background manipulation was possibly less salient because participants rating videos focused on the applicant's verbal responses and professional demeanor, which were equivalent across conditions.It is also possible that the differences would have been more pronounced if the bed was more visible or the background was less neat (e.g., unmade bed, a pile of unwashed clothes in the corner).Effects could have perhaps also been larger with a female (vs.a male) applicant.For instance, female applicants tend to be more penalized than male applicants when they have parental responsibilities (Cuddy et al., 2004) or emphasize the importance of work-life balance (Proost & Verhaest, 2018).Such gender discrepancies were particularly salient during the COVID-19 pandemic, where female employees were more likely to work remotely and deal with childcare, whereas male employees made fewer adjustments to their work roles (Shockley et al., 2021).As such, a female applicant recording their AVI from a bedroom could have activated traditional gender role stereotypes, or a blurred background could have suggested that the applicant was hiding evidence of how childcare was impacting their work life, both of which could have led to lower evaluations.Future research should thus explore gender differences in how AVI background impacts interview ratings.Third, our findings might also reflect changing normative behaviours and perceptions of control or visibility associated with AVI background due to the COVID-19 pandemic.Indeed, recording an AVI in a bedroom has been traditionally described as inadequate in popular media (Vasel, 2020) and by researchers (Mejia & Torres, 2018).Yet, the pandemic forced many employees to work remotely, adapt their personal living spaces into work areas (Loignon et al., 2022), and use technology to communicate with coworkers or clients (Kniffin et al., 2021).It might similarly have altered perceptions as to what constitutes an acceptable space to record AVIs.The use of blurred backgrounds has also become more common since video-conferencing platforms (e.g., Zoom, MS Teams) started integrating blurring features during the pandemic.In addition, individual characteristics are more likely to lead to negative reactions when they are controllable and visible (Summers et al., 2018).But the pandemic may also have changed perceptions of background controllability (e.g., understanding that some applicants might have no choice but to record from a bedroom) and visibility (e.g., accepting that some applicants prefer to hide their personal environment through blurring), thus reducing the negative effect of background type.Future research could explore the effect of the pandemic on normative behaviors and perceived background controllability/visibility in technology-mediated interviews.
Our findings provide some support for other components of Huffcutt et al.'s (2011) model in the context of AVIs.For instance, the relationship between response quality and interview evaluations suggests that AVI raters incorporate information about applicant qualifications obtained during the interview.This should be reassuring for applicants worried about the limited opportunities to impress interviewers in AVIs (Basch et al., 2020) and could be used by organizations to improve applicant reactions (Basch & Melchers, 2019;Griswold et al., 2021).These findings are also consistent with preliminary evidence about the poten-tial validity of AVIs (Gorman et al., 2018), likely because AVIs are highly structured (Lukacik et al., 2022).Yet, the positive relationship between initial impressions and final interview ratings is consistent with past work on the role of initial impressions in face-to-face interviews (Macan & Dipboye, 1990;Swider et al., 2016) and preliminary AVI findings (Suen et al., 2019).These findings show that AVIs are certainly not bias proof, thus emphasizing the need for research to explore other potential biasing factors in that context (see Roulin et al., 2023 for a recent example).

Practical Implications
Both applicants and organizations should be reassured that background type alone does not influence raters' initial impressions or final interview judgements.Applicants might thus enjoy the increased flexibility and convenience to record their AVIs from any location with internet access, or use a blurred background, without placing them at a disadvantage.This flexibility may benefit job candidates coordinating work and personal schedules or those forced to record AVIs from a bedroom (Lukacik et al., 2022).Yet, this recommendation might be limited to the kinds of backgrounds used in our study (i.e., clean, tidy, quiet, without any elements signaling protected status).Our findings about initial impressions in AVIs suggest that organizations should avoid introductory statements by applicants, or not provide them to raters, to minimize non-job-related information available (Levashina et al., 2014).

Limitations and Future Research Directions
Although we created videos that resembled real AVI recordings to enhance external validity, the study involved a simulated selection with an actor playing the applicant role and online panel participants.Because effects are larger in field versus lab studies of technology-mediated interviews (Blacksmith et al., 2016), future research could explore the effect of background using real applicant videos and/or in high-stakes selection contexts.In addition, researchers could replicate and expand our findings using within-subject designs.For instance, raters could watch and evaluate the responses of several applicants with different backgrounds, which could more closely resemble a real selection process (where hiring managers evaluate the AVI recordings of several applicant for the same job).Yet, such studies should consider threats to internal validity (e.g., ensuring that all actors and response scripts are equivalent, how rating several applicants could influence initial impressions).Prior research noted that applicant appearance can influence AVI ratings (Suen et al., 2019;Torres & Gregory, 2018).Our "applicant" was well-groomed and wore professional attire (a fact noted by several participants in open-ended comments), which could have dispelled the background effects.Future studies could examine the combined roles of applicant appearance and background (e.g., a casual attire magnify a bedroom background).The high and low quality response scripts were very polarized and made it easy for raters to judge the applicant as (un)qualified.Perhaps video background plays a larger role when responses are of average quality, with more room for rater interpretation.We included a question to check that participants recognized the blurred (vs.real) background, but we did not directly ask them whether they identified the office versus bedroom.It is possible that some individuals failed to distinguish (or ignored) the background type presented to them.In addition, although our pilot studies demonstrated differences in perceived professionalism, both visible background conditions were very tidy and clean.A less tidy or clean bedroom may indicate lower conscientiousness (Gosling et al., 2002) and represent a greater deviation from the norm (i.e., proper interview etiquette).This might trigger more negative emotional reactions and initial impressions.Future research may also consider other background locations (e.g., recording from a bathroom, a public library, or outside) or virtual background, or examine the interaction of location with specific background elements.For instance, a bedroom could be seen more negatively if it includes objects signaling of protected group status (e.g., religious affiliation, sexual orientation) or values that could be polarizing (e.g., political beliefs; Roulin et al., 2023).

TABLE 1 .
Average Initial Impression and Overall Interview Ratings Across Conditions Note.Based on PROCESS Model 15, 95% confidence intervals are computed using 5000 bootstrapped samples.N = 376.** p < .01.