Skip to content

Category: AI

Job Descriptions, Recruitment and AI

A "now hiring" sign to fit the theme of how companies engage in recruitment.
Photo by Free To Use Sounds on Unsplash

Finding good employees is tough. Unemployment is at a 50 year low (it is around 3.6%) There are around 1 million more job openings vs. potential applicants at this point in time, and the reality is that immigration and finding candidates that can come to the US to work is getting more difficult under the current administration.

So, how can companies find the right candidates? And why is an employment law blog writing about recruitment? Recruitment efforts (when not accurate or strategic) can have negative implications on employees in the workplace, and the use of particular recruitment software strategies can be sticky- so being well-informed of the pitfalls is key.

It Starts with the Job Description

Crafting job descriptions is an incredibly important part of any recruitment process. Why does it matter? There are 3 important reasons. It sets your expectations for recruiting for the position, it sets the essential job duties if an employee is disabled and needs an accommodation under the Americans with Disabilities Act, and it forms the basis of meeting the duties within the white-collar exemptions.

Setting Expectations

An advertisement for an open position will always contain a job description and expected duties of the position. The accuracy of these postings are critical. A poorly worded job description will not attract the right candidates will create a big waste of time and money for the business and may even expose the company to a discrimination claim. In the event a business chooses to hire an applicant who was recruited with an inaccurate job description,  when these candidates join the company, they will likely feel lied to, and  that the job does not match what they applied for. LinkedIn’s Talent blog has a great breakdown of what candidates actually review when they see a job description. According to LinkedIn, candidates believed that the most important parts of the job posting were the compensation, qualifications, and job duties. Clear and specific measures of what success in the position looks like were also very popular with candidates. The best job postings also let the candidate know exactly what they need to do to achieve success in the position.

Job descriptions and Essential Functions under the ADA

 The job description forms the basis of the essential functions of the position, which determine whether someone can be accommodated under the ADA if they have a disability. In many positions, it is critical for the company to have clear job descriptions to defend themselves in the event that a person has a disability and cannot be accommodated. Job descriptions should be formalized and written down. For example, attendance can be a regular requirement of a job

White Collar Exemptions

Finally, the job description can serve as evidence that a person meets one of the salary basis exemptions for overtime. In many instances white collar employees (think Doctor, Lawyer, Manager, etc.) will qualify as exempt from overtime which means even if they work over 40 hours they would not qualify for overtime pay.  A written job description stating the duties of these individuals will be used as evidence to see if the person meets the relevant exemption for the executive, administrative, professional, sales, or other exemption related to overtime pay.

AI and Recruitment

There is no doubt that technology can aid immensely in recruitment. One can consider all of the technological improvements that have made work so much better. The internet allowed job openings to be posted online, online applications allow people to apply at their convenience (without having to mail or hand deliver an application), scheduling software can make setting up an interview a breeze, and applicants can access reviews of the company through Glassdoor, equipping them with information that would have been much more difficult previously unless you knew someone on the inside. 

It is a big task for any company to try to select the right candidates, especially when the pool is large and diverse. Many companies have begun using software to screen out candidates and select the best candidates for the position, however, many of these tactics have their own pitfalls you should be aware of (some of them being blatantly asking for a lawsuit). These methods can include using software to screen resumes, (unlawfully) targeting individuals for applications and advertisements based on characteristics of individuals like zip code and age, and finally conducting interviews through video or using an electronic device, essentially streamlining the interviewing process and lessening the need for in-person manpower.

Screening Resumes with Software

Using software to screen resumes and applications can be efficient. If a job requires a certain certification, degree, or level of experience and a candidate does not have that and a company cannot or does not want to consider applicants without the skill or experience, then a resume screener can automatically reject these applications and perhaps even send out an email letting them know that they were not accepted for the position. It basically prevents a real human from having to review an application. Of course, you do need to be careful about machine learning and letting the app run on its own. Amazon had to adjust some software that used machine learning that it was developing because the software was discriminating against women. The software penalized candidates for having “women” in their resume and penalized candidates from 2 women’s colleges. Companies must be careful to avoid discrimination in the resume screening by any algorithm or AI (Artificial Intelligence)  and continuing to monitor the software to determine if it later causes a disparate impact on any group.

Targeted Advertising

Facebook recently settled several cases involving its targeted advertising that enabled it to exclude certain individuals based on age. Age is a protected characteristic and is protected under the ADEA (The Age Discrimination and Employment Act of 1967). It is illegal to discriminate against someone on the basis of age. The ads targeted young men, so women and people over 55 did not even see the ads, which means that it was also discriminatory on the basis of sex (another protected class). Facebook also had ads that were targeted based on the zip code that individuals were living in which also may be problematic as racial makeup differs from zip code to zip code especially in metropolitan areas, which could also allow companies to discriminate based on race and even national origin.

AI Conducted Interviews

Video interviews have been one of the big developments that have come out in the last few years, but they are already subject to some problems. One of the most interesting and relevant examples of this at the moment is with the company HireVue and their interviewing technology.

HireVue’s AI conducts video-based  interviews which measure “non-verbal cues – such as facial expressions, eye-movements, body movements, details of clothes, and nuances of voice.” It also collects all the responses from the interviewers to allow users to more easily compare answers. 

There are a few issues with this kind of software including the question of if this technology could even reasonably identify things like emotion or personality traits. It could also potentially discriminate against individuals who have a different primary language, or even have individuals “put on a show” for the software based on knowing what the software is looking for.  The Brooking Institution stated that “Scientific evidence suggests that accurately inferring emotions from facial expressions is very difficult and it stands to reason that inferring personality traits is even harder, if it’s possible at all.”

Angela Chen at Technology Review stated the complaints that some groups have had with the software. “As a result, applicants who deviate from the “traditional”—including people don’t speak English as a native language or who are disabled—are likely to get lower scores, experts say. Plus, it encourages applicants to game the system by interviewing in a way that they know HireVue will like.”

However, HireVue claims that it

has a well-developed process that involves testing the input data used to train our algorithms for bias, and then performing what organizational psychologists call “adverse impact” (bias) testing on the output data (predictions) produced by each algorithm. If we find that a certain factor studied by the algorithm is producing a biased result, we take it out of the algorithm, retrain it with all data except that factor, and test it again. This all happens before the assessment can go live in an interview, so you can rest assured that HireVue algorithms are de-biased long before you ever encounter them.

Title VII bans national origin discrimination and as HireVue measures language patterns it may be a problem for the software. The software may suffer problems from creating a disparate impact based on national origin (caused by the candidate’s speech patterns) and may discriminate against individuals with a disability that may have had a stroke or other condition that affects their speech or facial expressions.

As a result of these issues and as noted by Drew Harwell in the Washington Post, the Electronic Privacy Information Center (EPIC) filed a complaint with the Federal Trade Commission stating that HireVue’s use of “unproven” AI systems to scan people’s faces and voices is a threat to workers.  The FTC “regularly enforces ‘unfair and deceptive acts or practices’ statutes against companies found to be making claims to consumers without a ‘reasonable basis’ in a way likely to ‘cause substantial injury.”’ EPIC claims that the results from HireVue are “biased, unproven and not replicable.” Further, they claim that it could be biased against someone because of a protected characteristic.  “HireVue advertises that its technology does not use facial recognition technology” because its systems do not attempt to identify people.”

EPIC argues that this is misleading as the FTC ruled that facial recognition technology includes any ‘“technologies that analyze facial geometry to predict demographic characteristics, expression or emotions.”’ The result of the EPIC case is a case all recruiters should watch, and it will be interesting to see the arguments as it unfolds. Recruiters that use AI can especially benefit from following the case to see whether any changes are necessary to the software that they use to minimize any disparate impact it has only any protected class (like anyone with a disability or those that are not native to the US). The case and similar ones will answer important questions on how this software and others like it can address discrimination concerns, which will have a great, and likely positive, impact on how recruitment is done in the future.

Illinois has also taken a stand on the issue. It recently enacted legislation that requires the company to  disclose that artificial intelligence analysis will be used before the interview, and that the company must “provide each applicant with information before the interview explaining how the artificial intelligence works and what general types of characteristics it uses to evaluate applicants.” Companies must also get consent from the applicant before they use this software.

It will be interesting to see how the use of this technology evolves. There is likely a place for AI in recruitment, however there are still many problems to be worked out.

Conclusion

It’s a tough time for recruitment- with unemployment so low, and immigration work visas more difficult to obtain, finding quality candidates who can fill a job is a difficult task. To compound this issue, the HR Daily Advisor’s 2017 Annual Recruiting Survey found that 58% of respondents had no recruitment strategy. With no clear strategies at hand, and a smaller pool of applicants, getting clear about what you can do to strategically recruit is essential for companies to have the best chance of finding quality candidates in their search. Additionally, being aware of the implications of poorly vs. well written job descriptions, as well as the risks and rewards of certain software supports in recruitment efforts, will make your company better prepared to recruit moving forward. 

The information provided in this blog is for educational purposes only and is not legal advice. If you need legal advice, then you should speak with a lawyer about your specific issues. Every legal issue is unique. A lawyer can help you with your situation. Reading the blog, contacting me through the site, emailing me or commenting on a post does not create an attorney-client relationship between any reader and me.

The information provided is my own and does not reflect the opinion of my firm or anyone else.

Brett Holubeck (of Houston, Texas) is the attorney responsible for this site.