Key technologies and legal considerations when using AI in recruitment

The use of automated decision-making software has grown exponentially in the last decade due to its significant accuracy and efficiency benefits. However, as the practice has expanded, an unprecedented collection of legal issues has arisen along with it, making compliance on behalf of industry practitioners ever more nuanced. With disaggregated global regulation around AI technology, it’s essential for organizations to understand the legal considerations and how to best use the technology to drive compliant and positive outcomes. 

Key technologies and legal considerations

AI Matching Technologies

AI matching technologies are typically used to evaluate potential candidates on role-based criteria. They increase efficiencies and avoid human bias to create the best longlist based on the ability to perform the job alone. 

Key Legal Considerations

Because these technologies learn from existing data sets, ingrained and historical biases at an organizational level can taint them. Algorithms learn from this data, with the potential to replicate and amplify existing biases, regardless of any other emphasis placed on merit. This presents a significant risk of inadvertently increasing discrimination in hiring and contravening employment law.

AI Chatbots

AI Chatbots are an efficient way to discern a candidate’s compatibility, leveraging algorithmic and learning- based response technologies to act as an objective interviewer or responder to FAQs.

Key Legal Considerations

Privacy watchdogs are increasingly monitoring the use of AI Chatbots in relation to privacy and employment laws. There is a prevailing concern that these tools can access information about a candidate that would be unattainable by human evaluation, using it to make hiring decisions. Therefore, upholding the privacy rights of candidates is critical and must be considered in building your chatbots.

AI-Grading Software

AI-grading software automates the grading of applications throughout the recruitment process, from candidate compatibility to evaluating interview responses. 

Key Legal Considerations

While functionally similar to AI-matching technologies, these grading software bears different legal considerations. In the pre-employment phase, there is a privacy concern that the algorithms may discern information that human analysis could not. Furthermore, there is a risk that algorithmic solutions can have their own biases at all stages, stemming from a lack of human empathy, social awareness, and critical reasoning skills to make equitable assessments.

Facial Recognition and Voice Analysis

These technologies use biometric attributes to determine information about a candidate that was previously unattainable. During the interview phase, they are commonly used to derive additional insights from facial expressions, body language and verbal language choice, style and tone.

Key Legal Considerations

Existing employment laws were not designed to govern these technologies, hence are difficult to adapt. Furthermore, their use raises significant privacy concerns, as biometric data handling requires additional safeguards, and collection can be considered invasive. In addition, the process may be tainted by discriminatory algorithms, creating a complex employment law concern. In its infancy, facial recognition was primarily developed based on Caucasian males. This can add inherent bias as these were the first traits embedded by machine learning, demonstrated by a reported 34% increase in bias when analyzing women of color.

In Conclusion

The use of AI in recruitment poses a rapidly emerging legal challenge. Without comprehensive legislation governing the technology, the onus is on industry practitioners to apply traditional legal doctrines to complex systems. In relation to employment law, the primary concerns revolve around bias in algorithms. A large part of this risk is rooted in misguided and unsupervised mandates. Unsupervised learning systems that learn from a pattern of existing biases pose legal and employment concerns when used for recruitment purposes. It is incumbent upon organizations to ensure their algorithms are supervised, based on recognized organizational psychology principles, and routinely assessed. 

Concerning privacy, technology presents a novel consideration to the right to privacy itself. Advancements in artificial intelligence offer opportunities to collect data that was previously unattainable. In this landscape, organizations must have a compliant means to collect privacy consent, ensuring notifications are easily accessible, comprehensible, and up to date. When there is no legal requirement to institute such measures at the time of writing, proactivity might provide practitioners with retroactive indemnity and a competitive edge.

The impress.ai response

At impress.ai, we employ a combination of rules-based and supervised learning algorithms. Our approach sets rules based on widely recognized organizational psychology research that demonstrably combats bias. It does this by evaluating the candidate‘s merits within a prescribed mandate rather than replicating potentially problematic hiring norms.

In addition, we can help HR and recruiters refine and build the best privacy consent structure for their system and jurisdiction, including traditional notifications and just-in-time notices.

We also recommend and support impress.ai clients to seek bias assessments from accredited bodies. Just as an organization would perform employee evaluations, assessing intelligent decision-making systems should be routine to ensure their mandate remains impartial and in accordance with prevailing regulations.