Artificial intelligence and cybervetting: Benefits and downsides

The gap that prevails between human resources (HR) practices and ethical concerns about the use of AI in the hiring process made us ponder whether and how AI could be beneficial in terms of cybervetting job applicants and whether AI could be designed in a more positive manner that would allow job applicants and hiring managers/organizations to trust the use of AI to cybervet prospective employees. We thought of a couple of ways in which AI could be helpful for cybervetting purposes and in the hiring process more broadly.

First, perhaps the use of AI to cybervet job applicants could take place earlier in the hiring process, to compare multiple candidates who are being considered for the position instead of waiting until there are only one or two finalists to cybervet them. For example, AI could be used to cybervet candidates based on their application documents, which is perceived as a more ethical way to use AI in hiring (M = 3.60; da Motta Veiga etal., Reference da Motta Veiga, Clark and Figueroa-Armijos2022). AI could thus become a component of the earlier steps of hiring process through which hiring managers can, for example, cybervetting whether applicants are being truthful with their application documents as well as with information related to their background and fit with the position and organization. This strategy would be conducive to collecting applicant data about overall characteristics relevant to the job and the organization.

A second way in which AI could be used for cybervetting purposes, while also mitigating some of the privacy concerns raised by Wilcox etal. (Reference Wilcox, Damarin and McDonald2022), would be to aggregate applicants’ social media data collected by AI. This score would then become one hurdle of the hiring process, along with other prominent factors such as job–person fit (Figueroa-Armijos etal., Reference Figueroa-Armijos, Clark and da Motta Veiga2022). More specifically, to be both ethical and efficient in using AI for cybervetting purposes, hiring managers could use AI-enabled tools to gather internet and social media information about multiple candidates (i.e., finalists for a job opening), keep that data and information at the aggregate level (e.g., use a scoring system about person–job and person–organization fit rather than just “looking for red flags”), and thus be able to compare job applicants in a more objective manner. Aclear opportunity to use AI in cybervetting is the potential to create scoring systems that would keep the social media and internet information private while only providing a summary of the activity or a score based on criteria determined vital by the organization and around which the AI will be designed. Unfortunately, one major problem remains in that the AI is created by humans who are inherently biased (Jasanoff, Reference Jasanoff2016) and by organizations who have their own profitability goals at stake (Brynjolfsson & McAfee, Reference Brynjolfsson and McAfee2014). Furthermore, as mentioned earlier, there are also risks related to inherent biases against women and minorities (Noble, Reference Noble2018; O’Neil, Reference O’Neil2016), which warrant an industry-wide collaborative approach.

To provide a more concrete example of the above-described approach would be for the organization to create a score that builds on the diversity of the data collected to determine a job applicant’s level of authenticity (i.e., an increasingly sought-after characteristic; Cha etal., Reference Cha, Hewlin, Roberts, Buckman, Leroy, Steckler, Ostermeier and Cooper2019). The organization would thus create an algorithm to allow hiring managers to capture that aspect of a job applicant’s personality, based on their social media and internet activity. This posits, though, another important question. Are people authentic on social media? How can this be accurately measured?

Although we proposed a couple of ways in which AI can be helpful for cybervetting purposes, we also acknowledge some of the inherent downsides of using AI in cybervetting—namely the increased lack of privacy and the instrument deficiency in what the AI is meant to capture. First, although we argue that AI can be leveraged to help protect one’s privacy by only providing aggregated data about job applicants, there is also the unspoken risk that organizations might use AI to get information about a job applicant’s personal and professional life, beyond what is strictly necessary. The gatekeeper in this case would be the person in charge of designing the algorithm to gather the required information while keeping it as private as possible. Second, and related to the issue around the person who designs the AI, is that organizations are likely to commission one of their employees or outsource this task to create an AI system that captures various characteristics about job applicants based on their social media and internet information. This critical step would unequivocally need to be scientifically tested to ensure that the AI is capturing what it is intended to and that there is no measurement deficiency or contamination (Murphy & Shiarella, Reference Murphy and Shiarella1997).