Cameron Schilling
Companies are rapidly adopting artificial intelligence and automated decision-making (ADS) tools. They should be valuable tools for HR professionals and an integral aspect of today and future HR operations. However, companies that do so should be aware that ADS or Manager or HR experts are responsible for discriminatory HR decisions that rely on AI and are responsible for making these decisions.
AI and advertising offer a huge benefit to businesses. Here are just a few examples.
- Review and summary of job seeker materials for integrity and compatibility, including available jobs, employer culture, and expectations.
- Summary and analysis of interview content and performance.
- Evaluation of personality, traits, skills, aptitude, etc.
- Performance summary and analysis based entirely on employee work products, including email, phone, video phone calls, recorded meetings, interactions between business systems and databases, speed and accuracy.
- Recommendations for employment, progress, training opportunities, and more.
- Monitoring and warning about employee misconduct regarding safety and discipline.
The initial response of some managers and HR experts is that AI is not useful to them or that they don't trust AI to produce reliable results. However, as more HR operations use AI, managers and HR experts who refuse to do so will lose contact with common practices. Furthermore, it has been shown that business managers and HR experts make decisions that are influenced by potential variables and inherent biases and are consciously and unconsciously influenced by, and the results obtained by AI are not necessarily less reliable than human decisions. Finally, existing HR systems and applications already have AI and ADS built into them, making their capabilities more powerful. Employers who choose not to use it will be exposed to competitive disadvantage over time.
Like people who make HR decisions, AI can generate decisions based on inappropriate variables. This occurs because AI models are trained using historical data and embody the biases and trends that the data have existed throughout history. Similarly, AI is trained on HR data on employer historical practices, existing workforce, management structures, job descriptions, and more. AI can also unintentionally face a disadvantage when advertising disqualifies candidates due to availability and candidate availability is based on medical or family care requirements, or when AI can easily assess employee performance at meetings and such performance is based on impairments.
People like AI can produce discriminatory outcomes, so it's not surprising that employers who use AI are responsible for that discrimination. New York City, Illinois and California have adopted or amended existing laws to ensure the outcome. But do you really need a new law to communicate that? Existing laws, including RSA Chapter 354 A and the federal Civil Rights Act and Title VII of that Act, already prohibit and punish discrimination by employers, no matter how discriminatory decisions are made. In fact, AI and ADS, no matter how autonomous they behave, have not implemented such decisions. Rather, employers use AI and ADS to analyze the data and implement the results generated by those technologies. When humans rely on technology to make HR decisions, humans are still making decisions. Similarly, if an employer allows the implementation of the advertising outcome without human supervision, the employer decides to do so and is responsible for the outcome.
Perhaps the most prominent aspect of these new AI regulations for HR decisions is that they require employers to maintain records that can later scrutinize to determine whether HR decisions are legal or discriminatory. That requirement is generally consistent with other existing and emerging laws on AI, providing broader and more stringent requirements for using AI for HR capabilities.
The European Union and Colorado adopted a wide range of AI laws last year, while other states almost certainly will adopt similar laws this year and in the coming years. These regulations categorize the use of AI for employment decisions as high risk. That does not mean that the use of such AI is prohibited. Rather, in order to use AI for that purpose, employers must first perform risk assessments to identify risks inherent in AI use and implement measures to mitigate those risks. These measures ensure human control across the process, including safeguards such as thorough testing of AI before implementing AI, reliability of the data used to train models, limiting the use of AI to qualified managers and HR professionals, configuring AI, model training, data entry, process and process creation, and reviewing and auditing results.
The risk assessment process has the following results:
- Implement measures that eliminate or mitigate the possibility of discriminatory decisions that are being made and implemented.
- Follow new employment discrimination regulations.
- Establish defenses at events where future employees or existing employees claim that their use of AI or AD has led to discrimination.
AI errors in HR functions are not a reason to refrain from using them. After all, humans are wrong too, and technological advancements are essential for business development and competition. Rather, like other new technologies, AI in HR must be implemented based on appropriate risk assessment and mitigation processes.
Cam Shilling Founded and Chairs McLane Middleton's Cybersecurity and Privacy Group. A group of six lawyers and one paralegals will assist businesses and private clients to improve security, privacy and AI compliance and address any incidents or violations that arise.
