Written by:Abdul Wahaab MSc GMBPsSOrganisational Psychologist
Matsumoto and Hwang (2011) eloquently summarise the study on facial expressions, with these said to consist of various states ranging from Joy to Sadness to Anger. Did you know that facial expressions may be measured as part of recruitment processes for large organisations? Read on to find out more.
Large organisations get swamped with thousands of applications every single year. The competition for roles always seems to be increasing, with Covid appearing to make every application process that bit harder to pass.
As such, organisations are always faced with the challenge of implementing quicker yet legally defensible strategies to hire the best candidates in a shorter period of time.
The challenge arises with the onset of technology and its impact on recruitment. Technology moves quicker than many other domains given its format and broad reach. As such, it could be argued that recruitment methods reliant on newer technologies are not as tried and tested, as say a psychometric test which has been modelled and used to predict job performance for decades (e.g., Bertua, Anderson, & Salgado, 2011).
Unfortunately, this seems to be the case in some domains. Not only do organisations want to hire the best talent, but there are also a growing number making a conscious effort to make the recruitment process more engaging. As Virgin have experienced first-hand, a negative experience with the application process may lead candidates to associate Virgin with something negative, and perhaps avoid Virgin-related products in the future.
Conversely, a positive experience with the recruitment process helps two-fold:
Thus, a company strives to make their recruitment process, quick, engaging and legally defensible. Technology plays a part in helping to realise this dream, though the desire to be both engaging and legally defensible is often at odds with the other.
Take HireVue. A large proponent of video interviewing software, HireVue recently came under scrutiny during the Black Lives Matter movement for its use of AI software to map facial expression among other cues to determine a candidate’s suitability for the role. The challenge is related to the parameters that predict candidate suitability. It could be argued that if these are created in a particular country by a particular group of people, then these parameters are subject to the suitability criteria of that country and those people. In the above piece, HireVue state that efforts to combat this include hiring Autistic professionals to account for Autistic representation in the general population. Indeed, this is a step in the right direction and the road to bias-free A.I. is a challenging one.
In addition to the suitability of A.I. itself, the other challenge in using Artificial Intelligence to map facial expressions relates to whether extraneous factors influence a candidate’s suitability. Often, candidates’ unfamiliarity with a video interview can result in unfavourable expressions more attributable to the experience itself rather than their lack thereof. It is of no surprise then that organisations themselves now encourage candidates to practice assessments beforehand (e.g., Morgan Stanley). There is no secret formula that will help you ace an assessment, but the familiarity aspect (for AI-powered video interviews especially) is key to put forward your best self, or best expression rather.
That’s not to say objective, empirical and bias-free methods of quantifying candidate suitability through A.I. is an impossibility – rather, how can we be absolutely sure of the latter when we are reaching toward more and more engaging recruitment methods? Indeed, there is a challenge to meet both these bits of criteria.
Facial expressions are often measured using plug-ins that are developed to post-process a video recording to detect and interpret different emotion points. For instance, software can detect a mouth and its inflexions to contribute toward a potential output in which the individual is said to be happy.
The example of using Artificial Intelligence in video interviews is a popular one and the primary focus of this piece. Sonru is another leading provider, with the basic idea that not only do you speed up the hiring process by avoiding face-to-face interview appointments, but you also get a bit of extra information to help with selection.
The onset of Covid-19 and virtual recruitment in particular, has resulted in many more companies turning to video interviews as part of their recruitment process.
Before we delve into this, we need to first consider the process by which an application is reviewed. Take the above discussion; if Artificial Intelligence is being used to determine one’s video interview performance, it may be subject to inherent biases. Of course, these are not at all intentional, and efforts are being made to address these. However, the issue remains: Artificial Intelligence is not necessarily a superior means of eliminating bias when compared to human beings. There is still work to be done here.
Thus, if Artificial Intelligence is being used to determine video interview performance, facial expressions that do not conform to the parameters set by the software design team, are theoretically unfavourable. Preparation for interviews is key and can help alleviate some unpreferred expressions by bringing across confidence and a real know-how of the job role, organisation and field as a whole.
It is worth noting that some AI-based interviews use multiple cues to determine one’s expression or performance, such as gestures. Confidence in the role will help bring this across in one’s performance, though it can also be useful for candidates to visualise the ideal employee. Promotional videos from the company are also useful, as are recorded interviews, as one can gain a sense of their preferred communication style and needs.
Recruiters, especially those working for large organisations, are often required to sift through a large amount of video interview recordings to determine a candidate’s suitability. Potentially, one’s facial expression could be misinterpreted by the recruiter as say, a lack of confidence in one’s ability.
That’s not to say this is the norm, absolutely not. However, there is a likelihood that unfavourable or negative expressions may potentially paint the candidate in a less favourable manner. These nonverbal cues hold greater weighting for progressing two otherwise identical candidates to the next stage of the process.
The onset of technology to quickly sift the best candidates, alongside the Covid-19 pandemic, has resulted in a recruitment landscape in which recruiters look to remain legally defensible, but also provide an engaging experience. The competition in providing the latter brings many challenges, not least the fairness and suitability of newer technologies in candidate selection. As time progresses and technology evolves further, we must too evolve and make conscious efforts to address lapses in fairness and equality in a proactive manner.
