Using Transcripts to Judge Job Interviews Could Reverse Bias Against Autistic Job Seekers
Could "neurotype-blind" interviews raise the shockingly low employment rate of autistic people?
A 2022 study by Christopher Whelpey and Cynthia May finds evidence that neurotypical people prefer non-autistic job candidates — even when they see autistic ones as more qualified. The study also suggests a way to reduce or reverse the bias against autistic job candidates: removing visual and auditory cues, and evaluating text transcripts.
When neurotypical evaluators read mock interview transcripts, they do not perceive autistic candidates as more awkward than non-autistic ones. In fact, they view autistic candidates more positively and prefer to hire them.
The transcripts removed visual and auditory cues that neurotypical people find off-putting, so the raters’ biases against autistic ways of interacting weren’t triggered and couldn’t interfere with judging the candidates. When raters focus only on the content of what job seekers say, autistic people’s qualifications and enthusiasm come through.
Autistic People are Disadvantaged in the Job Market, and Job Interviews Act as a Barrier
Autistic people have high rates of unemployment and underemployment — even more so than people with other disabilities. When hired, they are often overqualified for the position. They also work fewer hours and earn lower wages than both the general population and people with other disabilities.
These disadvantages exist even after controlling for IQ and level of education. In fact, a college or higher level of education doesn’t help much. It may even hurt, because well-educated, high-IQ people may not qualify for government and other programs that give employment support to people with disabilities.
One reason for these unemployment rates is bias against autistic people’s ways of interacting during the hiring process. Job interviews require many skills that are hard for autistic people, such as:
intense, rapid back and forth interaction;
answering open ended questions;
determining what interviewers want to hear; and
projecting a confident, friendly persona
appearing enthusiastic, but not in an awkward way.
Autistic people often leave negative first impressions, even if they are perceived as qualified and competent.
In short, many qualified autistic people are not being hired because they are disadvantaged in job interviews.
As with people in general, unemployment reduces autistic people’s independence; denies them social support; hurts their mental health; and lowers their quality of life. Workplaces are also missing out on their skills and commitment.
Thus, discovering how to make job interviews more equitable would benefit both autistic people and workplaces in general.
Research Questions and Hypothesis
Whelpey and May investigated whether and how job interviews disadvantage autistic people. First, they conducted mock job interviews with autistic and non-autistic people. Next, they had people who did not know participants’ neurotype evaluate the interviews. Raters judged how qualified each candidate was, how positive an impression they left, and how well they performed in the interview. Finally, raters said how likely they would be to hire each candidate.
One group of raters watched videos of the interviews (getting both audio and visual information). The other group read transcripts (not seeing or hearing the candidates).
In the video condition, impressions of the candidates could be based on differences in style, such as body language, eye contact, and tone of voice. In the transcripts, only words were available. Thus, comparing how autistic people were judged in these two conditions shows whether autistic people were evaluated based on differences in style or in “the meaningful content of the interview.”
Researchers predicted that:
Autistic candidates would be judged to be as qualified as non-autistic candidates.
Despite that, when candidates could be seen and heard, neurotypical candidates would be viewed more positively and would be more likely to be “hired.”
In transcripts, where candidates can’t be seen or heard, this group difference would be reduced or eliminated.
Neurotypical candidates would be judged similarly in video and transcript conditions. Autistic candidates would be judged differently.
What the Researchers Did
Who Was Interviewed
Participants in the mock job interviews were 30 college students, aged 18-25. Fifteen people with a formal autism diagnosis and no intellectual disability [1] were recruited from their university’s office of disability services. Fifteen non-autistic people were recruited through Introduction to Psychology classes.
Note: The researchers called these groups the “Autistic Spectrum Disorder (ASD) group” and the “Neurotypical (NT) group.” No one in the “NT” group had a diagnosis of ASD. However, that doesn’t mean they were actually neurotypical. The researchers don’t say they checked whether these participants had another sort of neurodivergence (like dyslexia or ADHD), or believed themselves to be autistic or otherwise neurodivergent. For that reason, I am describing these groups as “autistic” and “non-autistic.”
Note: Interestingly, both groups had more females than males, which is common in Introduction to Psychology classes but unusual for autistic groups. In the ASD group, there were 9 females and 6 males, while in the NT group, there were 11 females and 4 males.
What Participants Did
Participants gave a 5 minute explanation of why they should be hired for their dream job. They were told to explain “why you want this job and what strengths you have that may help you succeed.” They were given 5 minutes to prepare and told they would be videotaped, and the videos would be rated along various unspecified dimensions.
Interviewers listened to the candidates’ explanations with flat affect. To avoid biasing the process, they did not give encouraging verbal or nonverbal feedback, such as smiling, nor did they verbally prompt participants. However, if the interviewee stopped before the full 5 minutes had passed, the interviewer said, “please continue.”
Participants were not told the real purpose of the study. Instead, they were told researchers wanted to understand the job interview process in general. (I think this “blinding” would have worked for the non-autistic group, but not the autistic group. After all, being recruited from Disability Services might tip people off that their disability status is relevant).
Who Rated the Interviews
314 college students at different US universities, age 18-25, were recruited to rate the interviews. (Yes, the raters were college students, not hiring managers. We’ll come back to this issue later). 151 people rated videos and 158 rated transcripts. Each person evaluated 5 autistic and 5 non-autistic interviewees.
Raters did not know that any of the candidates were autistic. They were not told the purpose of the study.
At least 50 people rated each video or transcript.
What the Raters Did
Participants were asked to rate how much they agreed with the following statements:
This candidate is TRUSTWORTHY
This candidate is LIKEABLE
This candidate is STRAIGHTFORWARD
This candidate is QUALIFIED FOR THE JOB
This candidate is ATTRACTIVE
This candidate is AWKWARD
This candidate is CONFIDENT
This candidate is ENTHUSIASTIC
This candidate is CAPTIVATING
The overall performance of this individual was good
I would hire this individual for the job they described.
Participants rated their level of agreement using a 7-point Likert scale, where 1 meant “strongly agree” and 7 meant “strongly disagree.”
As in most survey studies, there was also an unrelated “foil question” used to check whether participants were paying attention.
The researchers do not explain why they chose the qualities in questions 1-9, besides that they have been used in previous studies.
Ratings were “reverse scored,” meaning that:
For all questions but awkwardness, higher scores mean better ratings (eg, more trustworthy).
For awkwardness, lower scores means better (less awkward).
What the Researchers Found
Video
In the video condition, autistic candidates were judged equally qualified, but were rated as performing worse in the interview.
Non-autistic candidates were seen as better in every characteristic besides job qualifications.
Raters preferred to hire non-autistic candidates.
The graph below shows how strongly raters agreed or disagreed that job candidates had positive social qualities, such as “trustworthy” and “likeable.” Autistic candidates are shown in black and non-autistic ones in gray. Candidates were rated on a scale of 1 - 7, where 7 means strong agreement and 1 means strong disagreement. The asterisks show that group differences were statistically significant (at a p value < .001, meaning that there was less than 1/10th of a percent likelihood that these results occurred by chance).
The graph shows that non-autistic raters saw autistic candidates as equally qualified, but:
Less trustworthy (surprisingly, given the stereotype that autistic people don’t lie)
Less likeable
Less straightforward (surprisingly, given the stereotype that autistic people are too blunt, but it could reflect difficulty speaking under stress)
Less attractive
More awkward
Less confident
Less enthusiastic (which might be surprising if one considers autistic people’s passion about their interests, or expected given the stereotype that autistic people are not expressive)
Less captivating
Performing less well in the job interview, overall.
Transcripts
The results were quite different in the transcript condition.
The graph below again shows how strongly raters agreed or disagreed that job candidates had positive social qualities, such as “trustworthy” and “likeable.” Autistic candidates are shown in black and non-autistic ones in gray. Candidates were rated on a scale of 1 - 7, where 7 means strong agreement and 1 means strong disagreement. The asterisks indicate which group differences were statistically significant (at a p value < .01, meaning that there was less than 1% likelihood that these results occurred by chance). These differences are smaller and less statistically significant than in the video condition.
Autistic candidates were judged to be more qualified and to perform better overall in the job interview.
Autistic candidates were seen as:
Equally trustworthy
Equally likeable
More straightforward
More qualified
More attractive
Equally awkward
Equally confident
More enthusiastic
More captivating
Better performing in the interview, overall.
Raters were more likely to want to hire autistic candidates!
Video vs. Transcripts
The graph below shows how positive the overall impressions were of autistic candidates (black) and non-autistic ones (gray), comparing videos versus transcripts. Again, candidates were rated on a scale of 1 - 7, where higher scores are better.
With video interviews, non-autistic candidates were judged as performing better than autistic ones. With transcripts, autistic candidates were judged as performing better than non-autistic ones. The asterisks show that both of these group differences were statistically significant (at a p value < .001, meaning that there was less than 1/10th of a percent likelihood that these results occurred by chance).
Interestingly, using transcripts changes what matters to raters when deciding whether they’d want to hire candidates.
The researchers analyzed which characteristics best predicted whether raters wanted to hire a candidate. Would these differ between autistic and neurotypical candidates? Would they differ between videos and transcripts?
In the video condition, raters used different factors to judge autistic versus non-autistic job candidates. For non-autistic candidates, the best predictor of intent to hire was the level of qualification. For the autistic candidates, judgments of trustworthiness, likeability, and straightforwardness were more important.
In other words, neurotypical candidates are judged as they should be — on their qualifications — whereas the equally qualified autistic candidates are instead judged by their social weaknesses. As a result, autistic candidates are systematically disadvantaged.
“In the transcript condition, the importance of being qualified becomes more important for ASD candidates relative to NT candidates. Thus, in the absence of visual cues, not only are ASD candidates rated as more qualified than NT candidates, but the relative importance of being qualified also increases in terms of hiring intentions.” — Whelpey & May, 2022
Strengths and Limitations of the Study
Strengths of the Study
First, researchers made an effort to ensure instructions were clear to autistic participants. That ensures that any group differences weren’t just because the non-autistic group understood instructions better.
“We worked with Disability Services to ensure that our consent materials were composed in plain language (Plain Writing Act, 2010) and were clear and appropriate for use with individuals with ASD.”
“Plain language” means writing that lets readers:
find what they need,
understand what they find the first time they read or hear it,
use what they find to meet their needs.
The resulting instructions would’ve been much clearer than real-world interview questions.
Second, the effects were probably underestimated, for several reasons.
Evaluators were college students rather than real employers. Because younger generations have more exposure to autistic people in school and in public, and more anti-bias training at work, they could be less biased against autistic candidates.
Autistic people were applying for their dream job, for which they were genuinely enthusiastic. Furthermore, they probably knew what strengths made someone a good fit for the job, and were not just trying to guess what mattered to interviewers and employers.
Autistic people did not have much back-and-forth interaction with interviewers, so there were fewer moments when their interaction could be seen as atypical and negative.
I suspect the interviewers’ neutral affect and lack of response could have made non-autistic participants think they were doing poorly, distressing them. The interviewers’ behavior reminds me of the “still face” experiment used in developmental psychology, where a parent’s lack of response distresses babies. In other words, the interviewer’s style could’ve actually put the non-autistic job seekers at a disadvantage.
Note for those interested in statistics and study design: The researchers determined how many participants they needed to recruit using a power analysis. This method finds out, given how strong you expect the effect to be, how many participants you will need to detect any significant effects. Although it is best practice to perform a power analysis before conducting a study, not all researchers do so.
Limitations of the Study
Like all studies, this one had several limitations, which reduce the likelihood that the results apply to real world job interview situations.
In this study, autistic participants were college students. Many autistic young adults do not attend college. Previous generations may be less likely to have gone to college, and may be perceived more negatively, because they grew up with more stigma and less public understanding of autism
Autistic participants did not have intellectual disability (ID). People with ID (low IQ) are perceived differently, apply for different jobs, and may get different amounts and types of job skills training. Thus, it’s not clear how much the results will translate to them.
Autistic participants could speak well enough to convey their qualifications and enthusiasm. I’m not sure how well results will translate to those who have less speaking ability, or lose their skills under stress.
Participants could list whatever dream jobs they wanted. The researchers did not say which ones, or whether non-autistic and autistic candidates applied for the same jobs. So, we can’t know for sure whether results would be the same for an autistic and a neurotypical candidate applying for the exact same job.
Raters were assumed to be neurotypical, but the researchers don’t seem to have confirmed it. They don’t report checking that none of the raters were diagnosed with autism or another developmental disability, and that none of them identified as autistic or otherwise neurodivergent.
This study did not explore the characteristics of individual autistic people that predicted how they were rated. It also did not examine characteristics of individual raters that predicted how biased they were, such as their knowledge and attitudes about autism. To be fair, the researchers couldn’t have asked such questions without giving away the purpose of the study. Then, raters’ behavior might have changed to fit what they thought researchers wanted to know and how they wanted the researchers to see them.
Does the answer to the question “how likely are you to hire this person?” predict actually hiring the person in real life? The answer isn’t as straightforward as you might think. People express intentions to vote for particular candidates, donate to charities, diet, exercise, and more in studies, then do something entirely different. However honestly they try to answer, study participants can be bad at predicting what they will actually do. Studies of people’s implicit biases about specific groups have been criticized for failing to show whether results in the lab translate to differences in real world behavior.
What Should Workplaces Do to Eliminate Anti-Autistic Bias from Job Interviews?
If the results of this study translate to real world workplaces, we can draw some interesting conclusions.
First, the obvious:
Neurotypical people’s biases may prevent qualified autistic candidates from being hired.
That may be one reason for autistic people’s high unemployment/underemployment rate. This bias may also prevent businesses from hiring the best candidates for the job, especially for jobs where in-person social interaction is not very important.
The conclusion that makes this study fascinating:
When NT people focus only on the content of what job seekers communicate, autistic people are actually more socially adept than neurotypical people.
Autistic candidates are no longer seen as unusually awkward. Instead, they are viewed not only as more trustworthy, enthusiastic, and worthy of hiring, but also as more captivating and even attractive!
It makes sense. Autistic people are highly aware of and used to relying on verbal content, whereas neurotypical people rely on, and are used to paying attention to, style and nonverbal cues.
Unfortunately, the researchers don’t talk much about the fact that they not only reduced bias, they actually reversed it.
What should workplaces do to reduce bias against qualified autistic job applicants?
The researchers spend a lot of time discussing whether and how to make neurotypical interviewers become aware of and change their biases.
Initially, researchers focused on changing how autistic people present themselves in interviews. This approach is challenging. Imagine trying to have a difficult conversation while changing everything about your tone of voice, body movements, and facial expressions. The authors argue that it’d be easier and more effective to address neurotypical interviewers’ biases than to change how autistic people express themselves.
However, to address neurotypical interviewers’ biases, one must make them consciously aware that they are interacting with an autistic person. Yet, disclosing an autism diagnosis can backfire and lead to even more discrimination.
The study results suggest a simpler answer: just have businesses make decisions based on interview transcripts, instead of in-person judging.
Using transcripts is “neurotype-blind” and “neurotype-deaf,” similar to the “blind” auditions orchestras use. They prevent any bias against autistic applicants from influencing hiring decisions.
With this approach, there is no need to directly change interviewers’ biases about autism. There’s no need for the candidate to disclose their diagnosis. Using transcripts is simpler, and less risky, than bias training.
However, if we do try to change interviewers’ attitudes, we should teach them to ignore the style and focus on the content of what candidates say.
In this study, even without explicit instruction to focus on content, doing so led raters to appreciate the communication strengths of autistic people.
Citation for the Research Study
Whelpey, C. E., & May, C. P. (2022). Seeing is Disliking: Evidence of Bias Against Individuals with Autism Spectrum Disorder in Traditional Job Interviews. Journal of Autism and Developmental Disorders, 1-12. https://doi.org/10.1007/s10803-022-05432-2
Edited for clarity 2/28/23; reference list changed to in-text links 7/6/23.