Reducing bias through training
Does accent bias awareness training make any difference?
We tested several interventions designed to make recruiters more aware of the potential for accent bias. We found that accent-based differences in ratings of professional competence can be reduced through such training.
It’s perfectly normal to have stereotypical reactions to accents. Psychologists have shown that intuitions help us navigate our daily lives. When we have a lot of information to process quickly, stereotypes and snap judgments help us to free up cognitive effort and focus on what we need to.
But if left unchecked, these biases and stereotypes can lead to discriminatory behaviour. We may have stereotypical ideas about social groups, but it is important not to allow those to influence our decisions about whether someone is suitable for a job in a law firm, for example. These should be based on objective criteria and deliberative thinking about their strengths and weaknesses.
The ability to ignore stereotypes and process information more consciously is what psychologists call cognitive control. Over the years, a variety of strategies have been proposed for enhancing such control, to reduce discrimination. We tested 5 of these strategies to see whether they were effective in controlling people’s intuitive accent biases.
The five control strategies we tested were as follows:
Strategy 1: Raising Awareness. Recruiters are alerted to the existence of accent bias.
Strategy 2: Identifying irrelevant information. Recruiters are asked to commit to ignoring irrelevant information when making their decisions, e.g. If I hear that the candidate has an accent, I will pay no attention to it.
Strategy 3: Committing to fairness and objectivity. Recruiters are asked to commit to an agreed set of objective criteria before making judgments.
Strategy 4: Increasing accountability. Recruiters are told that they will have to justify their decisions.
Strategy 5: Appealing to multiculturalism. Recruiters’ attention is drawn to diversity and its positive benefits.
Testing the strategies
We designed an experiment to test whether any of these strategies are effective in reducing the kinds of accent bias that we found in our nationwide survey on attitudes to voices.
In the experiment, we focused on three accents : Received Pronunciation (RP), Multicultural London English (MLE) and Estuary English (EE). We chose these because, in our survey, RP speakers were rated the most employable, while MLE and EE speakers did significantly worse. We tested whether any of the above strategies can reduce these differences.
Our procedures were similar to those for the nationwide survey. We asked 480 members of the general UK public to rate three candidates for an entry-level job at a major British law firm. Each candidate was a native speaker of either RP, MLE or EE, and listeners heard each candidate respond to one interview question. Before hearing the candidate’s response, listeners were given information of the five types above, and a sixth (control) group had no intervention.
Participants then listened to each of the three candidates and rated them on the same five evaluation scales used in the nationwide survey
After rating all three candidates, participants provided information about their own personal background (including gender, ethnicity, age, region of origin, highest level of education, occupation, English accent, languages spoken), their exposure to different UK accents, the diversity of their own social networks, their beliefs about bias in Britain, and a set of psychological measures such as their level of oncern about being perceived as prejudiced.
What did we find?
All strategies had some effect, but Raising Awareness (strategy 1) had the strongest and most consistent impact on reducing the difference in rating between a non-standard accent (MLE or EE) and RP.
This means that when people were alerted to the existence of accent bias, differences between their ratings of job candidates with different accents were smaller.
The other four strategies (Identifying Irrelevant Information, Committing to Fairness, Increasing Accountability, Appealing to Multiculturalism) all showed some positive benefits, but more inconsistently: some were effective for only certain participants, and some for only certain candidates.
None of the strategies were shown to have a harmful effect (i.e., none increased bias), but their positive impact was more limited. Note that all of the other four rely on first making people aware of accent bias. Our results show that simply raising awareness about accent bias may be the simplest for recruiters.
We have shown that raising awareness about accent bias is an important tool in helping to reduce bias in professional hiring contexts.
Our other results have shown that accent bias is still prevalent, so this intervention is needed.
We would therefore encourage recruiters and HR professionals to inform themselves about accent diversity and the potential for bias in hiring. We recommend incorporating the brief wording below (shown to reduce accent bias) to remind recruiters not to rely on accent in judging competence:
Recent research has shown that, when evaluating candidates’ performance, interviewers in the UK can be influenced by the candidates’ accents of English. In particular, they tend to rate candidates who speak with a “standard” accent more favourably than candidates who speak with “non-standard” accents. This is an example of so-called “accent bias”. The focus should be on the knowledge and skills of the candidate, not their accent. Please keep this in mind when assessing the suitability of candidates.
In training, it may also be helpful to remind recruiters that candidates’ accents are the natural result of their social background. It is difficult for anyone, of any social class, to change their accent, and nobody should be asked to reject their own social group. Candidates from under-represented social backgrounds often also have under-represented accents, and should not face double-discrimination for this reason. Job candidates should be assessed on their knowledge and skills, not their accents.