As artificial intelligence capabilities continue to increase, employers will contend with many issues surrounding the use of AI in the workplace. To prepare employers to address some of these issues, we have created a series of posts examining employers’ use of AI.
AI is relatively new, but it is certainly here to stay. For employers considering implementing AI processes, there are some general considerations they should keep in mind as they delve into the ever-growing world of AI in the workplace.
Link to The importance of accuracy The importance of accuracy
Whenever a process is automated, employers should ask questions about the accuracy and precision of the program being used. Such questions are particularly pertinent when the process being automated is traditionally reliant on human emotion and perception, such as job interviews. When incorporating AI into these processes, an employer should know what information has been used to educate the AI program. For example:
- Is the information source truthful?
- Has the data source been vetted for potential inaccuracy?
- How much data (and what kind) was used to educate the program?
One major point of potential weakness in AI is that, generally speaking, it lacks human perception and creativity. That means AI can often be inflexible and overly literal. Thus, it may not consider a biased or false news article or be able to distinguish whether information on an applicant’s resume is actually truthful.
Because AI programs use mass amounts of general data to evolve and become useful, the question arises: Can the AI program accurately identify what information in its data bank relates to the particular real world scenario in which it is applied? Despite how helpful AI seems in theory, employers should be aware of the potential for the program to get stuck in technicalities and produce unexpected results. One need not look further than a quick internet search of the terms “lawyer” and “Chat GPT” to see real-world examples of how problematic unchecked trust in AI can be.
Employers should consider whether using AI in the workplace and in their processes will improve efficiency—and at what cost.
Link to Privacy surrounding AI in the workplace Privacy surrounding AI in the workplace
Employers also should keep in mind the privacy of individuals being evaluated by an automated process. For example, when AI technology such as facial recognition or expression analysis is used in an interview, the applicant will likely want some sort of assurance that their unique facial features are securely captured.
At the very least, it is important to consider existing security and network storage privacy protections. In the event of a data breach, the type of personally-identifiable information gathered and stored by AI has the potential to present significant privacy and confidentiality concerns. Additionally, the integration of AI programs and related data collection software might, understandably, trigger concerns about the security of employer trade secret and proprietary information.
Link to Programming bias Programming bias
The typical process through which AI technology is developed also is an area of potential concern for employers because AI must be taught to identify certain features, terms or qualifications. This means that, generally, the AI technology is only as open-mined as it is trained to be.
There are certain behaviors an AI tool may learn depending on the number of scenarios it experiences before being implemented and depending on the terms or circumstances used in its training. Closely evaluating the learning process and testing the technology are important—before becoming reliant on AI—to avoid potential bias.
Link to Discrimination in employment decisions Discrimination in employment decisions
Perhaps the most commonly discussed point of concern, and the topic of forthcoming articles in this series, is the potential for discrimination in employment decisions made using AI programs. The use of personally-identifiable information, such as an individual’s face, or details contained in a person’s work history, like a gap in employment, are a potential hotbed for alleged discriminatory practices.
While liability for such allegations of discrimination is still evolving and has not yet been subject to extensive judicial analysis, it is likely we will see litigation over the alleged discriminatory use or impact of AI in employment decisions in the years to come.