Colleague robot will (not) conduct your assessment tomorrow

BION 05.2020

The wonders of digitalization seem to replace the human workforce step by step. This development has not stopped at HR. There are numerous examples of the use of digital processes and instruments, especially in regard to the personnel selection process. Therefore, we ask the following question: Are artificially intelligent machines the better recruiters?

By Ruben Grasemann

Digitalization plays an increasingly important role in personnel selection processes. There are some prominent examples that have received media attention lately: Amazon’s use of AI in recruiting, as well as ambitious claims to replace traditional AI-based recruitment tests with faster and more valid procedures made by companies like Precire. The question that is increasingly being asked: Are artificially intelligent machines the better recruiters?
In order to answer this question, we first take a look at the present state of personnel assessment. To be able to make a statement on potential today, trained observers use competence-based, structured interviews and results, for example from work simulations, role plays and conceptual tasks (Melchers et al., 2020). Online performance and personality tests are also used as a complementary tool. According to Lochner & Preuß (2018), we are already “at the transition to the future”, because interviews are already being used digitally and remotely and sometimes time-delayed instead of live for years now. The research focus in regard to the future of assessments is now increasingly directed towards the topic of artificial intelligence (AI).

Recognizing data patterns

By AI we understand the ability of machines to think, decide and act humanly (Lochner & Preuß, 2018). For this purpose, machines learn to recognize data patterns and use them to predict future events (Lochner & Preuß, 2018). Davenport and Ronanki (2018) view the automation of complex corporate processes and the knowledge gained through data analysis as crucial driving factors behind the use of artificial intelligence in corporate contexts.

The aim of companies is to improve and develop existing products, make better decisions and give employees more space for creative work by automating routine tasks (Davenport & Ronanki, 2018). In the context of assessments, the use of artificial intelligence becomes meaningful in the analysis of CVs and cover letters (Campion et al., 2016; Langer et al., 2018; Lochner & Preuß, 2018) and the automatic decoding of verbal and non-verbal behaviour (Chamorro-Premuzic et al., 2016; Langer et al., 2018; Lochner & Preuß, 2018).

Campion et al (2016) addressed the question of CV-analysis in personnel assessment by training an AI to emulate a human evaluator. They conclude that the machine proves to be as reliable and valid as its human counterpart and emphasize the financial benefits of such a solution. With regard to the AI-driven evaluation of cover letters, Lochner & Preuß (2018) argue, on the basis of previous studies, that this would also yield similar results as the human evaluator. They emphasize however, that there is a general lack of empirical findings regarding the valid prediction of job application letters on professional success and therefore pose the question “to what extent companies will demand cover letters in the future at all” (Kanning, 2016; Lochner & Preuß, 2018).

Analysis of the Big Five dimensions of personality

For the evaluation of interviews, AIs nowadays transcribe the spoken word and then conduct the analysis (Lochner & Preuß, 2018). Companies such as Amazon, Google, Microsoft and IBM, for example, offer AIs such as IBM Watson, an AI that enables the analysis of the Big Five dimensions of personality alongside the six corresponding facets (Costa & McCrae 1992; Lochner & Preuß, 2018). Chamorro-Premuzic (2017) describes other forms of interview analysis, for example the fully automated recording and analysis of speech characteristics, facial expressions and body movements. The latter being considered as valid predictors of personality traits. In this study, objectivity, reduction of the interviewer bias and cost efficiency are stated as possible advantages of using AI.

In consideration of the complexity of personnel assessments, however, different publications also reveal a variety of challenges that make the use of AI more difficult. Langer et al. (2018) note that there is a risk that applicants will be able to figure out the algorithms on which the AI is based. Accordingly, the AI can be manipulated, for example, by using a large number of positive emotion words in order to appear as open as possible.

Furthermore, AI is trained by using human examples. Thus, at certain points comes the risk that AI unintentionally learn human prejudices, which can lead to sexist or racist judgements. Langer et al. (2019) take an approach from a different angle as they focus on the participant’s experience: In a study with 123 participants they compare differences in the reaction to highly automated interviews compared to interviews in video conferences. In the context of the assessment, they found negative reactions and lower acceptance of the automated interviews due to a lack of perceived fairness and social presence.

More efficient and productive recruiters

With time, the selection patterns used by Amazon’s AI have been proven to be sexist (Wilke, 2018). And experts are warning about the usage of Precire’s speech analysis (Przybilla, 2017). The scientific implications for practitioners are clear: AI does not replace the recruiter. And what about tomorrow? Science agrees that AI will become an integral part of recruiting, as it increasingly develops into a valid, reliable and cost-effective tool that supports HR on its way to becoming a strategic partner (Lochner & Preuß, 2018).

However, this is not a foregone conclusion: science repeatedly emphasizes the transparency and controllability in the implementation, especially in terms of the General Data Protection Regulation (GDPR) and the traceability of decisions (Lochner & Preuß, 2018). Although the improvement of the HR practitioners‘ digital competencies is essential, it still falls short of the mark, as the change to AI-driven methods must be professionally accompanied, similar to a change process, in order to reduce reactance and emphasize the vision (Davenport & Romanki, 2018; Lochner & Preuß, 2018).

Lochner and Preuß (2018) are optimistic: “Automated assessment (…) by no means makes recruiters dispensable. Recruiters will become more efficient, more productive and more successful”. It must be clear, however, that in the highly functional symbiosis of human observer and artificially intelligent machine, the level of human responsibility will continue to rise: In terms of the ethical handling of data, in terms of transparent communication and in terms of empathetic, compassionate interaction with the human participant.