Note

Hired by a Robot: The Legal Implications of Artificial Intelligence Video Interviews and Advocating for Greater Protection of Job Applicants

Introduction

Self-driving cars, 3-D printers, and now Artificial Intelligence (“AI”) systems make decisions that affect human lives.1 Technology has led us into the future further than many science fiction novels projected. While fears persist that AI may one day exhibit more intelligence than humans and lead the world into a dystopian movie scape,2 in its present state the technology has a lot to learn before it will no longer need the oversight of humans3 and the legislative system.

Although AI has been integrated into many different industries,4 its use in the employment industry presents many unique legal issues.5 As employers integrate AI into their hiring, monitoring, and performance review processes,6 legislatures need to contemplate the possible ethical and legal implications of AI. While AI has many benefits such as removing human bias and improving efficiency,7 it also has risks such as algorithmic bias8 and data privacy.9 If programs do not properly code and monitor, the algorithm may unintentionally discriminate in the recommendations it gives to employers. To make its recommendations or decisions, the AI algorithm gathers data, and if this data is not protected it could be used in ways the job applicants, employees, or even employers never intended.

An intriguing AI development over the last few years is AI video interviews.10 AI video interviews provide benefits to employers by reaching more diverse and distant candidates.11 However, AI video interviews have the same risks as general AI such as algorithmic bias12 and data privacy issues13 with almost non-existent regulation.14 Without regulation, employers risk discriminating and not receiving the top job applicants, and the applicants have no protection over their video and data generated from their interview.

This Note argues for state legislatures to enact regulations for AI video interviews to protect job applicants and give employers and AI vendors guidelines. The regulations should provide for informed consent, protect less advantaged job applicants, and protect applicants’ privacy. Part II presents an overview of AI and how it has been integrated in the employment field. After explaining the AI video interview technology, this Section explains the benefits and risks regarding bias and data privacy. It then finishes with a survey of the current field of AI regulation with a focus on video interviews. Part III elaborates on the benefits and deficiencies of the Illinois Artificial Intelligence Video Interview Act. Finally, Part IV provides recommendations for legislatures in creating AI video interview regulations for their state’s employment market.

AI and Employment

AI has been rapidly developing over the last dozen years.15AI discoveries in a variety of fields have a great outlook for improving society.16 Across industries the AI developments are looked at with excitement for the potential improved efficiency and advancements in products and society.17 Alongside the improvements, people have voiced concerns about the ethics and sustainability of AI.18 One of their greatest fears is AI will take people’s jobs and reduce the workforce.19 However, rather than stealing jobs, many AI developments have focused on assisting employers to hire employees.

Businesses are implementing AI in their human resources departments. Employers now use AI to assist with employee training,20 to review productivity,21 and to augment job roles by taking over tedious tasks.22 Of note is the use of AI in the recruitment and hiring process. Effectively managing the hiring process is important not only for the candidates but for the company as well.23 In order to get the best candidates for the job, employers need to make sure they are effectively screening candidates and testing the potential employees’ fit for the role and the company.24 For the employee, the hiring process can become a barrier that may unintentionally keep qualified candidates away.25 The most qualified candidates may not get the job due to the hiring employees implicit bias or preference for inapplicable prestige markers such as alma mater favoritism or nepotism.

AI provides many benefits to employers’ hiring processes. AI developments in the hiring process include scanning resumes,26 tracking candidates throughout the process,27 game-based assessments (measure relevant skills through computer “games”),28 and interviewing candidates.29 The integration of AI within the hiring process has the potential of reducing hiring time by 75 percent.30 Besides time saving, AI can yield twice the hiring pool and allow the company to eliminate job applicants without requiring human intervention.31 Therefore, not only does AI in the hiring process allow employers to save time and money, but it has the potential to reach more applicants and provide a more diverse hiring pool.32

Besides reaching further to find applicants, using AI in hiring can remove the bias present in the interviewers to give a fairer shot to all interviewees.33 The AI evaluates the candidates scientifically, because AI does not have the same explicit or implicit bias as a human interviewer, who may unknowingly be vetting candidates based on their bias.34 The AI will focus on the merits and personality of the candidate and their fit for the role rather than possibly fixating on arbitrary characteristics such as sharing the same favorite sports teams or having a mutual connection.35

Section II.A explains the development of AI video interviews and how they are conducted. Section II.B examines the possible risks and issues of using AI algorithms in employment, specifically algorithmic bias and data privacy. Finally, Section II.C gives an overview of the current employment legislation for video interviews and AI data.

Development of AI and Video Interviews

“[A]rtificial [I]ntelligence . . . [is] the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings.”36 The computers are designed to embody human characteristics correlated with intelligence, “such as the ability to reason, discover meaning, generalize, or learn from past experience.”37 The first AI machine was developed in 1952 to play checkers,38 but has gained substantial ground with integrations in many different fields, from self-driving vehicles, to medical diagnosis, and facial and voice recognition software.39

The AI used for video interviews employs machine learning to analyze the candidates. Instead of requiring specific rules to process information, machine learning uses the data sets to learn and predict without explicit instructions.40 The AI is given an algorithm, the set of rules the AI will follow to gather and analyze the information,41 and then goes beyond the initial information provided, continually sending the new data it gathers back into the algorithm to generate more accurate predictions.42\

How AI is Employed in Video Interviews

Employer’s initial screening process of candidates is used to refine who is most qualified for the position by considering the skills they currently possess. At the interview stage, most employers are looking to discover the applicant’s personality and whether that personality will fit with the open role and the company culture.43 With society becoming more virtual and the difficulties scheduling in-person interviews, many companies have turned to asynchronous video interviews, a recorded interview the applicant can take at any time. However, research has found that humans are not able to accurately predict another person’s personality based on recorded interviews.44 The integration of AI in asynchronous interview technology is capable of “surpass[ing] humans in recognizing or predicting an applicant’s personality for screening job applicants because applying AI techniques to audio-visual datasets can achieve more reliable and predictive power than human raters.”45 Humans evaluating candidates may have bias, “whereas a computer does not have implicit biases: we can expect that a computer will evaluate all interviewees using the same criteria and make personality judgments more consistent and fair.”46

AI’s integration in the video interview process gives employers a better idea of the candidate’s abilities and whether they have the skills best suited to be successful in the job. The AI detects the interviewee’s facial expressions and tracks the changes in expressions.47 Additionally the AI captures data on the interviewee’s enunciation, word choice, and other speech patterns.48 The AI analyzes the facial expressions and other body movements to evaluate the applicant’s nonverbal communication skills and fit for the job role.49 Researchers have some evidence that nonverbal communication skills better indicate an applicants’ interpersonal skills than their verbal communication with an interviewer.50 “Unlike other forms of nonverbal communication, facial expression is universal and conveys human emotions that can be recognized by computers with a high degree of accuracy.”51 Research has shown that these nonverbal cues are better indicators of a job candidates interpersonal communication style than behavioral interview questions.52

There are many different models to determine a candidate’s personality, but many AI researchers and developers used the popular five-factor model (“FFM”) also referred to as the OCEAN model: “openness, conscientiousness, extraversion, agreeableness, and neuroticism”.53 The AI determines these traits based on how the person’s face moves or the inflection in their voice when answering questions.54 The video interview AI “system dissects the tiniest details of candidates’ responses — their facial expressions, their eye contact and perceived ‘enthusiasm.’”55 The questions analyze the candidate’s tone and movements to see “how excited someone seems about a certain work task or how they would behave around angry customers.”56 For example, the change in expression or tone can reveal their conscientiousness through how interested they are in answering about how they like to organize their daily tasks.57 Word choice can also reveal candidate skills, such as saying “we” rather than “I” indicates team skills.58

These AI-detected traits and characteristics then can be refined to fit the job role. An employer can tailor the algorithm to pick the most applicable traits to the role.59 For example, if an employer is looking to fill a customer service role, they would look for an applicant that is more agreeable and conscientious so they could best assist customers. While for non-customer facing roles such as developers, an employer might instead look for an applicant that rates high on openness and conscientiousness. AI in asynchronous video interviews give the employer the flexibility of conducting a less biased personality screening in a more time-effective and distant manner.

AI Video Interviews in Practice: A Closer Look at HireVue

A range of companies provide video interview capabilities with AI analysis.60 The market has grown large enough for competition to break out. The market leader of AI video interviews is HireVue, a hiring platform founded in 2004.61 HireVue provides AI assessment upgrades to their asynchronous video interviews platform.62 HireVue boasts having conducted over “22 million on-demand interviews” and claims to have “[m]ore than 700 enterprise customers worldwide.”63

When an employer uses HireVue for interviews, the platform sends the job applicant instructions on how to complete the interview through the HireVue platform on a computer or handheld device.64 The interviewee may start the interview at any time during the day when they have roughly 30 minutes available.65 The interviewee will be presented a question by HireVue and given 30 seconds to prepare. The interviewee stops the recording once they have answered the question.66 Most companies choose the unlimited answer feature, which allows interviewees to review their answers and decide if they would like to rerecord.67 Next the AI analyzes the interviewee and gives the employer “Insights Scores” for each interviewee based on how much they have the qualities and characteristics needed for the position.68 The employer uses the scores to decide who to hire or typically who moves on to the next portion of the hiring process.69 It is a relatively quick process, but despite the benefits, employers and applicants need to be aware of the potential risks in AI.

Rising Issues in AI in Employment Contexts

While AI provides many benefits to employers, there are some potential issues. Section II.B.1 first gives an overview of the general bias issues that may develop in AI. Section II.B.2 focuses in on what issues can develop specifically for AI video interviews. Finally, Section II.B.3 discusses issues that could arise with data privacy when using AI.

General Algorithmic Bias in AI

As previously discussed, AI provides many advantages to employers for efficiency and especially in removing some (but not all) human bias. One of the biggest advantageous of using AI over humans in evaluating candidates for employment, is the removal of human explicit and implicit bias.70 AI evaluates candidates based on skills and traits that a good employee in the role would possess. However, the human element is not completely eliminated from AI. There are potential for biases in two aspects: programming by humans and the initial algorithm’s data set.71

First, a human programs the initial AI algorithm. This programming could cause any implicit bias held by the programmer or employer to be accidentally, or unconsciously, inputted into the algorithm.72 For example, if the employer wants the algorithm to prefer candidates that live in certain zip codes because they believe the proximity to the office will make for greater employee retention, this appears to be a neutral criteria.73 However, the AI could unintentionally discriminate against race and low income depending on the area the business is located.74 As many neighborhoods are still experiencing strong racial segregation and many business districts are located around more affluent areas of cities, instructing an AI algorithm to favor specific zip codes can creates a discriminatory hiring process.75

The greater danger of bias lies in the existing data sets used to create AI algorithms.76 The developers give the AI data sets to learn and make decisions that follow the data set. As the AI algorithm “learns” based on the specific data set given to build it, this data set has a huge effect on how the AI algorithm makes decisions. If the developers give the AI a biased data set, the AI form a biased algorithm from that data set and make biased decisions.77 Within the employment context, the danger of bias comes when data for the algorithm is used based on previous employees and job candidates.78 While using these qualities would initially appear to be a neutral way of establishing skills and traits for desirable employees, the algorithm may correlate these indicators of “success” with other possibly discriminatory traits or qualities.79

This potential danger was illustrated when Amazon used an AI algorithm to sort through resumes.80 Programmers developed the algorithm on resumes of “successful” candidates who were hired from the past ten years.81 However, Amazon failed to note that most applicants hired over the last ten years were white males.82 From this data set, the AI algorithm started to favor white males and prioritized resumes that had words associated with males.83 Accordingly, the algorithm disfavored woman and would penalize the applicant for any words associated.84 For example it would rank lower resumes that had “women’s soccer” or if they listed in their education any all-women’s schools.85 Amazon cut the algorithm before it implemented it into any real hiring decisions.86 However, if programmers do not review the AI as it is in practice, these discriminatory algorithms could cause discriminatory hiring.

Potential Bias in AI Video Interviews

AI video interviews have the same potential for bias but has some unique problems that arise through the use of video analysis. AI video interview technology is created similar to other AI hiring technology and has human-programmed algorithms often based on a training data set, creating danger of bias.87 AI in video interviews use facial recognition to track the changes in expression and analyze the changes in the speakers tone.88 There have been multiple studies that show facial recognition has analyzed people’s emotions differently based on race.89 AI associated black faces more with negative emotions such as anger to a greater degree than white faces.90 A recent study also showed AI analyzed white men as more honest.91 These algorithmic biases are formed as most of the test subjects are white males and females. With a base of data from only part of your applicant pool, it can create a biased result in the application of the AI.

AI also presents the risk of discriminating against other groups, such as LGBTQ individuals, on the basis of sexual orientation.92 AI is able to more effectively determine a person’s sexual orientation than humans.93 After the Supreme Court ruling in Bostock v. Clayton County held Title VII does not allow employers to discriminate based on sexual orientation, AI vendors and employers may run into Title VII issues if sexual orientation bias is not factored into the algorithms.94

Disability bias is even more difficult for AI programmers to mitigate.95 As the data used to create the algorithm is typically either from past and current employees or from a set of the general population that do not have disabilities, mental and physical disabilities are not often accounted for in the algorithm.96 The algorithms use data such as “typical enunciation and speaking at a specific pace a[s] qualities that might correlate with effective salespersons.”97 If people with disabilities have qualities that manifest in an atypical way compared with what is considered a normal “successful” candidate, then they will be overlooked even if they could successfully perform the job.98 For example, most AI video interview algorithms track candidates eyes when they respond to a question. Eye tracking is used to determine characteristics such as openness and honesty.99 Many people with disabilities, however, have different eye motion patterns when talking.100 Those with autism, for instance, typically look toward speakers’ mouths instead of the eyes.101 These differences could lead AI to identify incorrect characteristics about the applicant.102

Accounting for disabilities within AI algorithms involves a lot of thoughtful and expensive additional work.103 Instead of pulling from past successful employees or the general public, the AI data requires a large range of people with different disabilities in order to get a good amount of data for the typical responses to the interview questions.104 Creating a data pool for each disability requires considerable time, resources, and effort that many AI-developing companies might not be able (or choose) to afford.105

Even HireVue, a popular AI video interview product, does not indicate that they account for the differences in disabilities in their algorithms.106 In the FAQ on their website they say that many neurodiverse individuals “excel in job roles that don’t require you to be chatty or socially charismatic in some particular way.”107 They claim their software only evaluates for the skills needed for the role.108 They also recommend that the best resource for neurodiverse individuals is to seek an accommodation from the employer.109 This disclaimer puts the responsibility for the AI bias against those with disabilities (the job applicant) to request an accommodation. This is only effective if the job applicant is aware of the potential AI bias against those with disabilities and is willing to disclose their disability to the employer.110 With the growth in video interview AI technology, employers must keep in mind this risk to protect against unintentional discrimination.

AI and Data Privacy

The second main developing issue for AI that employers need to monitor is the applicant’s data generated by the AI video interviews. To examine this issue, this Section looks at the current state of data regulation. The world is going through a “data explosion,” meaning “the universe of data doubles every two years and quintillions of bytes of data are generated every day.”111 The “billions of smartphones and other devices” generate the staggering amount of daily data from the constant user interactions.112 With data being constantly gathered from electronic devices, issues regarding data privacy are ripe for attention from policy makers.113

AI has already become a big player in data and has shown the additional risks that will develop regarding personal information privacy.114 Most analysis of private data, “such as search algorithms, recommendation engines, and adtech networks—[is] driven by machine learning and decisions by algorithms.”115 For example, AI analyzes the data gathered on websites and creates connections that develop a picture of user behavior that can predict a person’s buying habits.116 There is very little regulation for what rights the user has to the data and if there is any protection from exploitation.117

For AI video interviews, the AI extracts data from the videos in order to analyze the candidates.118 The extracted data from the video interview may clearly be considered personal information as it is about the interviewee’s personality, characteristics, and verbal and non-verbal language abilities. Employers will need to account for how the data is being used and what protections will be guaranteed to their applicants.119 Additionally, issues may develop around how the gathered data is utilized beyond the initial interview.120 For instance, will the data only be used for the specific purpose of determining an applicant’s fit for a company, or would it also be used to improve the interview process going forward?121 Would the data be used by the developers to further refine the AI algorithm? There are no current widespread guidelines for regulating if the interviewee has any claim over the extracted data.122

Current AI Legislation in Employment Law

As AI is strongly tied into data privacy issues,123 AI video interview products need to comply with data privacy statutes. Section II.C.1 discusses current data privacy legislation that could affect AI video interviews. On its own, there is very little AI legislation on a state level124 and a lack of any legislation nationally.125 While there has been movement over recent years for regulation,126 Illinois and Maryland are the only states with an AI video interview legislation.127 Section II.C.2, gives an overview of Illinois and Maryland’s AI video interview statutes

Data Privacy Regulation and Legislation

The first all-encompassing regulation for data privacy protection is the European Union’s General Data Protection Regulation (“GDPR”) which has been effective since 2018.128 The GDPR is considered “the toughest privacy and security law in the world.”129 Even though it is an EU law, it applies to “organizations anywhere, so long as they target or collect data related to people in the EU.”130 The GDPR has broad requirements and limitations on data collection and the potential uses for data collectors.131 The GDPR focuses on protecting personal data and manifests these protections through the following:

data subjects’ privacy rights:

  1. The right to be informed
  2. The right of access
  3. The right to rectification
  4. The right to erasure
  5. The right to restrict processing
  6. The right to data portability
  7. The right to object
  8. Rights in relation to automated decision making and profiling.132

With the risk of harsh fines,133 any company who collects data from EU citizens, including AI vendors, must comply with the GDPR.

California is considered the U.S. state with the toughest legislation on data privacy.134 The California Consumer Privacy Act135 gives similar broad rights as the GDPR, such as to disclose the gathered data, to have it deleted, to opt out of selling it, and not to be discriminated against for opting out.136 As one of the main functions of AI is to learn through gathering data, employers and AI vendors will need to be aware of the protections afforded to the interviewees through state laws like California’s or the GDPR if the applicant is from the EU.137 Job applicants should be aware that if their state does not have data legislation, the data from their video interview can be collected and used by the employer and AI vendor at those companies’ discretion.

AI Video Interview Legislation

At this time, only two states have passed AI video interview legislation: Illinois and Maryland.138 Illinois enacted the Artificial Intelligence Video Interview Act in 2019.139 Prior to the interview, the Act requires employers to make the following disclosures regarding the use of AI to candidates:

(1) Notify each applicant before the interview that artificial intelligence may be used to analyze the applicant’s video interview and consider the applicant’s fitness for the position.

(2) 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.

(3) Obtain, before the interview, consent from the applicant to be evaluated by the artificial intelligence program as described in the information provided.140

The Illinois statute is particularly concerned with the applicant’s awareness and consent to AI use.141 By having informed applicants, both the employer and interviewee will be on a more level ground regarding the interview process.142

The statute also addresses concerns regarding privacy and data by limiting the employer’s ability to share the video.143 The statute states, “[a]n employer may not share applicant videos, except with persons whose expertise or technology is necessary in order to evaluate an applicant’s fitness for a position.”144 Additionally the statute intends to protect the applicants by giving the interviewees the right to have the video recording destroyed.145 Once the applicant sends in a request to delete the video, “employers, within 30 days after receipt of the request, must delete an applicant’s interviews and instruct any other persons who received copies of the applicant video interviews to also delete the videos, including all electronically generated backup copies. Any other such person shall comply with the employer’s instructions.”146 The Illinois policymakers created these provisions to ensure the rights of destruction of the video remained with the applicant,147 especially as most AI video interview software providers only delete applicant videos based on the employer’s specified length of time or explicit request.148

Illinois has been working to further improve the bill through an amendment on demographics.149 The amendment requires any employers that exclusively use AI video interviews to determine who to bring in for an in-person interview to report demographic data to the State’s Department of Commerce and Economic Opportunity annually.150 Specifically they need to report “the race and ethnicity of applicants who are and are not afforded the opportunity for an in-person interview after the use of artificial intelligence analysis; and . . . the race and ethnicity of applicants who are hired.”151 “The Department [will then] analyze the data reported and report to the Governor and General Assembly by July 1 of each year whether the data discloses a racial bias in the use of artificial intelligence.”152 This amendment goes into effect January 1, 2022.153

Maryland’s statute went into effect in October 2020 and is focused on ensuring applicants consent to AI video interviews.154 The statute also prevents employers from using facial recognition technology during a hiring interview unless the applicant consents.155 An applicant consents by signing a waiver that includes: “(i) the applicants name; (ii) the date of the interview; (iii) that the applicant consents to the use of facial recognition during the interview; and (iv) whether the applicant read the consent waiver.”156

Urgency of AI Video Interview Legislation and the Issues
to be Addressed

With jobs moving more online than ever before,157 it is urgent to stay on top of any developments and changes with the workplace. Distant hiring is growing158 and the law needs to adjust to fit these new employment needs. Illinois has already taken a step forward by enacting the first law on AI in asynchronous interviews.159 The law is a good template for other lawmakers to formulate statutes for their states. Section III.A discusses the positive aspects of the Illinois statute such as transparency, consent, and privacy protection. Section III.B moves on to discuss possible areas of improvement within consent, data privacy, bias, and liability. While the Illinois statue is a good first step there is also room to improve even further so there is increased certainty no uncertainty in future legislation. Finally, Section III.C will explain the urgency in regulating AI video interviews.

Other States Should Adopt the Positive Attributes of
Illinois’ Statute

Illinois’s Artificial Intelligence Video Interview Act has outlined for employers how to address key issues that may arise from the use of AI video interviews.160 Three main issues that the statute addresses are: (1) transparency, (2) consent, and (3) privacy protection. Each of these parts of the statute have good points that should be used in other states’ legislation.

Transparency of AI Use and Functionality in Interview

The transparency requirements of the statute serve an important function of putting the applicants on closer-to-equal footing with the employer. The Artificial Intelligence Video Interview Act ensures transparency through two prongs: initial notice and explaining the technology.161 The employer must “[n]otify each applicant before the interview that [AI will] . . . analyze the . . . video interview and consider the applicant’s fitness for the position.”162 Employers’ use of AI in video interviews is a recent development,163 and jobseekers may not be aware of the technology. By informing the applicants, it keeps the field fair between applicants and allows them to feel more comfortable with the technology. Notifying the applicants of the AI, allows the applicant to know that the video recordings they make will be analyzed by AI and not simply watched at a later time by the employer as an applicant may otherwise assume.164

The second transparency requirement is for the employer to “[p]rovide . . . information before the interview explaining how the artificial intelligence works and what general types of characteristics it uses to evaluate applicants.”165 This provision ensures the applicant has the relevant information to know how the employer uses the AI and can understand what the AI is evaluating in the video interview.166 The employers have some protection as the employee will clearly have the information available before the interview and can ask any questions at that time instead of bringing up a complaint or claim after the interview. This transparency is particularly important for the next aspect of the statute: consent. Without full and fair transparency, the applicant would not be able to give their fully informed consent.

Job Applicant’s Consent to the AI Video Interview

Receiving the applicant’s consent to using AI video interview technology is a good policy to protect all parties. The applicants are protected as requiring consent ensures the employer will inform applicants of the AI. Consent protects the employer and the AI developer as the applicant is agreeing to allow AI to analyze their facial and vocal expressions.

Illinois’ statute clearly outlines the two consent requirements: Consent must be “before the interview,” and to the AI evaluation “as described in the information provided.”167 The consent prior to the interview will keep the applicant from feeling obligated to accept since they have already conducted or are in the midst of the interview.

The requirement to consent to the evaluation “as described” assures the employer is fully transparent about what they use the AI to evaluate.168 If the employer tells the applicant the AI is used to evaluate applicants’ attitude and honesty, but the company evaluates language competency, then the employee has not consented to the AI’s language competency evaluation. To receive full consent, employers will need to be fully transparent with applicants, otherwise they risk liability for any other uses.

At the end of the section, the Illinois legislature makes clear that, “[a]n employer may not use artificial intelligence to evaluate applicants who have not consented to the use of artificial intelligence analysis.”169 If the applicants do not consent, the employer may not use AI in the interview process.170 Illinois is showing a priority of protecting and informing jobseekers of the new technology employers use during the hiring process.

Privacy Protection of the Job Applicant’s Video

The Artificial Intelligence Video Interview Act protects applicant’s privacy in two provisions. First, it explicitly states that the applicant’s video may only be shared, “with persons whose expertise or technology is necessary in order to evaluate an applicant’s fitness for a position.”171 Limiting access to the video evaluators, restricts the use of the video for hiring purposes. Second, the statute requires, “[u]pon request [of] the applicant, employers . . . must delete an applicant’s interviews and instruct any other persons who received copies of the applicant video interviews to also delete the videos.”172 This section protects applicants by giving them some control over their video. While the applicant does not get to decide who views the video, though limited by law, the applicant can protect their privacy by deleting the video. It is an important step in balancing the applicant’s rights and the employer’s ability to evaluate candidates as they see fit.

Drawbacks and Possible Issues Left to Determine in the
Illinois Statute

The Illinois Artificial Intelligence Video Interview Act covers very important ground in protecting the job applicants and the employers. However, there are a few questions that remain that lawmakers or judges may need to address in the future. These questions include: (1) consent consequences, (2) lack of bias protection, (3) applicant’s data privacy rights, and (4) liability.

Consent Provision Shortcomings

Two issues that arise for consent are to the required time before consenting and consequences if no consent is given. While the employer must receive consent before the interview,173 the statute does not outline how far in advance it must be given. An employer could spring the consent on the applicant minutes before the interview. Such an approach could cause an applicant to feel pressured into consenting even though there was no time to fully understand the implications of the AI evaluation. The point of consent is to ensure the applicant is able to make an informed decision about the interview, and there may be a need to outline when is considered “before the interview.”174

If the AI information and request for consent are not in clear places, there could be a contract of adhesion issue. If there is a simple box to check and the option of whether or not to read about the AI video interview, the consent would be similar to that of checking “I understand” in a terms of service. Whether this would stop the applicant from being able to sue or provide the proper protection against liability for the employer would require future litigation.

In addition, there is no guideline for the possible consequences if an applicant refuses to consent to the AI video interview.175 Is the employer allowed to drop the applicant out of the hiring pool?176 Or does the employer need to provide a different screening process that does not involve AI analysis? This could draw up additional issues for applicants with disabilities who do not want to consent as AI does not properly accommodate to some neurological and physical disabilities.177 If employers are not required to use a secondary method it would result in candidates having to request an accommodation, revealing their disability, even if the disability would not otherwise affect their job performance.

Lack of Algorithmic Bias Protection

The statute does not include any provisions on how to handle any bias that may arise in the AI video interview program’s algorithms. With the law silent on this facet of AI, the developers and employers will be left to decide if they employ any mechanisms to prevent bias. If bias develops in the algorithm, it could lead to Title VII, ADA, and other discriminatory hiring litigation. By the delegating what is required to be in compliance now, state laws could prevent costly and time-intensive litigation in the future.

Applicant’s Data Privacy Rights are not Properly Protected

While the statute accounts for the applicants’ rights to the interview video, the statute does not account for any data extracted from the video.178 The statute is silent on the employer or developer’s responsibility to data extracted by the AI.179 Can the data extracted only be shared “with persons whose expertise or technology is necessary in”180 evaluating the candidate or is it able to be shared for other purposes, such as research or algorithm maintenance?181 Also, when an applicant requests to have her video interview deleted, is the employer required to delete the extracted data and tell the developer to delete it as well?182 There is not a proper designation within the statute about who owns the data derived from the AI, and this issue needs to be determined by lawmakers so applicants know their rights to the data and employers know how they may use the data.

Scope of Liability and a Missing Cause of Action

Finally, the statute does not regulate any possible liability that arises due to the employer using the AI evaluations.183 For example, if there is a bias that develops in the algorithm, is the AI vendor responsible for those who were discriminated against due to the algorithmic bias or the employer?184 Who is responsible for oversight and compliance of the AI software?185 Lawmakers need to make it clear to the developers and employers what steps need to be taken to remove them from liability from AI discrimination.

Additionally, the statute does not create a specific cause of action for applicants to bring claims against either the employer or vendor for improperly following the statute to the applicant’s detriment or misusing the applicant’s data. Without a specific cause of action, the statute does not effectively allow the applicant to sue for any damages the employer or vendor may have caused through their biased algorithm or appropriating their data. State legislatures creating a clear cause of action puts employers and vendors on notice while protecting the applicants.

Urgency of States Creating AI Video Interview Legislation

With all the possible risks associated to applicants and employers, regulating AI video interviews is of immediate importance for employers, employees, and AI vendors. As technology has developed,186 more and more jobs are able to be performed completely virtually.187 Through globalization, companies have increasingly looked to hiring workers from distant locations.188 Of recent urgency is the increase in virtual jobs due to the COVID-19 pandemic.189 Many workers preferred virtual work or flexibility even before the pandemic.190 As more companies have increased flexibility to allow virtual positions or are even requiring virtual work, the ability to recruit candidates is expanded to greater distances away from the company.

However, the virtual nature of the position requires companies to adapt to more efficient virtual hiring practices. Video interviews will continue to grow and more companies seeing the benefit and efficiency of AI video interviews to spot the most qualified candidates will decide to implement AI video interviews within their own hiring processes.191 With more employers using the products and more applicants going through an AI interview, regulation of the process is most helpful solution.

As of this Note, there is not any current litigation regarding AI’s use in video interviews; however, the risk of litigation increases as the products are being increasingly employed in hiring. The first hint of future litigation has started in the form of an administrative complaint; the Federal Trade Commission (“FTC”) has received a complaint by the Electronic Privacy Information Center (“EPIC”) asking the agency to investigate AI’s use in hiring pre-screens, specifically naming HireVue.192 In the complaint, EPIC asserts HireVue “is flouting national and international standards of transparency, fairness and accountability.”193 The complaint argues that the FTC can exercise its authority over the matter as HireVue is using unfair and deceptive trade practices.194 These practices include, as represented by HireVue, that the company “builds algorithmic models for employers based on data from top performers, a method which can perpetuate past hiring biases.”195 Additionally in the complaint, because HireVue keeps their algorithms secret, the job applicants cannot effectively challenge the algorithms for bias.196

Creating legislation now will reduce the risks mentioned in EPIC’s complaint and will promote consistency in litigation as judges will have legislation to guide them in their decisions. Additionally, legislation can be enacted to keep candidates informed about the AI video interview process and help both employers and AI vendors meet regulations now before too much capital is put into the product, requiring more resources to later be made compliant.

AI Video Interview Legislation

First, Section IV.A discusses this Note’s recommendations for what lawmakers should include in future AI video interview legislation. The recommendations draw from the Illinois statute but expand the requirements. Next, Section IV.B explains why it is important to create legislation on AI video interview technology rather than allowing the employers and AI vendors to regulate themselves.

Proposal for State Legislature AI Video Interview
Legislation

When state legislatures develop their own AI video interview legislation, they should use the Illinois Artificial Intelligence Video Interview Act as a template but add additional requirements. New statutes should emulate the Illinois statute provisions on transparency, consent, and privacy protections.197 State legislatures should add additional provisions on bias protection, disability protection, data protection, and expanding the consent requirements.

Expanding the Applicant Consent Requirements

Legislation should follow the Illinois policy of requiring transparency and consent before the video interview.198 The applicant should be aware of the process, so they are on the same footing as the employer. Additionally, the statute should specify how soon before the interview the interviewer must request consent. The applicant needs to receive the information about the AI, how it is used in the interview, and what it evaluates199 with enough time to make an informed decision. Applicants need the time to review the material and be able to ask questions before giving consent, and complete their own research about the AI video interview if they wish. Giving a sufficient time period, such as 48 hours ensures the applicant is able to inform themselves as they wish and will give an informed consent.

However, the advanced time period only needs to apply to providing the necessary information to an applicant and requesting that applicant’s consent. If the consent is received from the applicant one minute before the interview, it is acceptable as long as the material was provided sufficiently in advance. The policy purpose is to provide the applicant an opportunity to inform themselves. To fit this purpose, the statute could require employers disclose the AI video interview in both the job posting and sent to the applicant when scheduling the interview. Disclosing the AI in the job posting will allow for all parties to save time; if someone does not wish to participate in the AI video interview, they can choose to skip out on applying to the position. As most AI video interviews are asynchronous,200 so long as the information and ability to ask questions is provided with instructions on how to complete the interview, and an immediate deadline is not set for conducting the interview, the employers would be in compliance.

States should require disclosing to the applicants what data the AI gathers through the evaluation201 The applicant’s consent should be contingent on what specific information was provided about the process and to the specific characteristics and traits the AI evaluates in the video.202 This requirement serves the policy goal of protecting the applicants and employers.203 The employer is not hiding what AI-determined qualities in a candidate they are looking for and it assures the applicant that the AI process is a fair and accurate evaluation.

Similar to Illinois, other states should require that if the applicant withholds consent, the employer cannot use the AI to evaluate the candidate.204 While it is important to allow employers’ decisions on what tools or practices are used to make hiring decisions,205 this ability needs to be balanced against protecting disadvantaged applicants. The statute should require the employers to provide alternatives to qualified candidates who do not consent. While there may be many different reasons a person does not consent to the AI video interview, those who have a disability should not feel forced to disclose when the job otherwise would not be affected by their disability. The employers should provide for an alternative for all promising applicants to eliminate any issues with bias for applicants with disabilities.

Data and Privacy Protection

The statutes should include provisions that protect the applicant’s privacy regarding the video and the data extracted. The provision should follow Illinois’s lead in restricting who has access to the video.206 Access to the video should only be given to those within the company who make hiring decisions, and should only be shared with those outside of the company whose technology is necessary to evaluate the candidates,207 i.e., the vendor’s developers. This will protect the applicant if there is any information shared in the video the applicant only expected to be seen and heard by those hiring. The applicant is protected from their video being spread or seen by others not involved in the direct evaluation of the interview.208

States’ statutes should include a provision requiring the employer to delete the video interview upon the applicant’s request.209 The applicant should have some power over their video and if the applicant does not want it to exist any longer, whether or not they are hired by the company, the video should be deleted. The provision should include the requirement to delete “within 30 days after receipt of the request”210 to guarantee the video will be deleted within a reasonable period. As the videos may also be with the AI vendor in order to evaluate the candidate, legislation should include Illinois’s requirement of the employer, after receiving the request, to tell other parties with the video to delete it within the 30-day period.211 The other parties should be required to “comply with the employer’s instructions.”212 Adding this instruction is important as the AI vendor often require the request to come from the employer or else they do not delete a video.213

The legislature should add into these provisions about restricted sharing and deleting the videos the same requirements for any data produced by the AI from the video interview. An example would be to modify Illinois’s statute by adding the following language: “An employer may not share applicant videos,” or data developed from the video, “except with persons whose expertise or technology is necessary in order to evaluate an applicant’s fitness for a position.”214 Similarly the extracted data should be deleted when the applicant requests the video to be deleted. The applicant should not have to specify that they also want the data to be deleted as this expects the applicants to have a deep understanding of how AI video interviews function. When the employer receives this request, the applicant’s video and data should be deleted and the AI vendor and developers should be instructed to delete both the video and data as well.

However, there can be an exception if the employer receives consent for the extracted data to be used for other purposes besides evaluating applicant fit for the role. The legislature should include in this exception that the employer is required to specifically request for the data to be used for other purposes. These other purposes, whether research, candidate tracking, or improving the AI algorithm, should be explicitly listed in the request. In the case of the consent to the data being used for other purposes, then the employer could share the data with which ever people or companies whose technology is required to fulfill those other purposes. Additionally, if the applicant requested the video interview be deleted, the extracted data would be allowed to remain and be used for the other consented purpose.

Algorithmic Discriminatory Bias Protection

Alongside a data protection provision, lawmakers should add in a bias protection provision to the statutes. To protect job applicants and to clear up any future questions of liability regarding bias, the legislature should add a provision requiring both the employer and AI vendor to be able to be held liable if the AI algorithm has a discriminatory effect on hiring. Having both the employer and AI vendor be liable for discriminatory algorithms guarantees both parties will take responsibility in verifying that the algorithms do not develop any discrimination. Liability promotes accountability on both parties. Employers need to be aware not to give data to the vendors that is only based on past hires if their current and previous employees were not diverse. The AI vendors need to be responsible if a bias develops within the algorithm to eliminate it.

Another avenue legislation could take is requiring periodic reviews of the algorithms by the vendors and of the positive candidates by the employers. Similar to Illinois amendment, the employers would send the data to the state agency who would review it for bias.215 Instead of only requiring employers that solely rely on the AI video interview to send the data, all employers would be required to send the data for who past the screening level, though they do not need to include any process past the initial AI video interview selection level.216 By having periodic reviews, it gives the AI programmers a chance to find any discriminatory patterns or deviations from the original algorithmic intent. The employers’ periodic review would require them to look at the candidates that passed the AI video interview screening and seeing if they are diverse candidates or if those candidates appear to be screened out. This vigilance on both sides of the AI product will safeguard against any discrimination in the initial algorithm or from developing based on the data it receives from new candidates. One of the benefits of AI is removing human bias from the interview process.217 To protect AI from developing its own bias, the best tool is human oversight.218

Additional Protection for Applicants with Disabilities

As discussed in Section II.B.2, AI is not able to properly adjust to interviewees with physical or neurological abnormalities that cause the interviewee to act or react in different manners than expected of the general population.219 Trying to account for bias against neurological and physical differences is not the most economical use of AI vendors’ resources.220 Until the technology is further developed to better account for physical and neurological abnormalities, the employers will have the extra burden of providing accommodations for those with disabilities.221 When the employer invites an individual to participate in a AI video interview there should be an option to provide accommodation for these individuals. As discussed in the consent provision, a solution to this issue could be requiring employers to provide an alternative to all applicants who do not consent to the AI video interview. Having an available alternative across the board lets those with disabilities decide whether or not to reveal their disability. The alternative gives applicants with disabilities the choice of whether they would like to disclose their disability or not. While employers cannot discriminate based on disability,222 there is a risk the employer may unconsciously take the disability into account during the hiring decision.223 Additionally, applicants should be informed of AI’s difficulty in properly accounting for disabilities in its algorithm when assigning traits to an applicant. Understanding AI’s limitation will allow applicants with disabilities to choose whether to try the AI interview or ask for an accommodation. With the alternate option, applicants with disabilities have the power to decide instead of being forced to disclose.

Why Legislature Should Not Leave AI Video Interview
Regulation to the Market

The AI vendors may argue there is no need for legislation as algorithmic bias is not a new problem in AI.224 AI vendors are aware of possible issues with AI and attempt to account for the issues in the products. Issues of algorithmic bias are well known to AI developers and most AI video interview companies, such as HireVue, have taken steps to try and combat any potential discriminatory bias in their algorithms.225 HireVue went as far as removing the facial recognition aspect of the video interview to alleviate concerns from the public.226

While the AI vendors already work on combatting bias, their concern is with producing the best possible product for their customers, the employers. The AI vendor’s primary concern is not to protect applicants, but rather to give employers the best algorithm for the employer’s hiring needs.227 Legislatures need to step in to be the voice of the job applicants and to protect them in the hiring process, a process where applicants do not generally have much power.228 Having guidelines such as required periodic reviews of algorithms will also protect AI vendors and the employers against possible litigation. If they are taking the proper steps to stop algorithmic bias, the court is less likely to find a discriminatory hiring process.229

In addition, it is beneficial for the legislature to establish what is expected of employers when utilizing AI video interviews in the company’s hiring process. The provisions outlined above would establish that the employer needs to be transparent, receive consent from the applicant, protect the applicant’s data and privacy rights, and use the technology in a way that attempts to combat bias rather than instill hidden biases. By knowing what the legislature expects in the practice of AI video interview analysis, the employer may confidently use it to search for close or distant candidates that will fit their hiring needs. Legislatures need to regulate AI video interviews to protect applicants during the interview process and to protect employers and vendors from litigation after the interview.

Conclusion

AI continues to grow and integrate into different markets. State legislatures need to keep up with the new technology and protect their citizens. Job applicants need to be protected against the possible discrimination and privacy violations in AI video interview algorithms. The legislation needs to make clear to employers and AI vendors that the government will encourage new developments but not at the expense of job applicants. With AI video interview legislation, the state legislatures would be making an important first step into catching up on regulating ever-developing AI.

  1. [1]. See, e.g., Sooraj Shah, Why Hasn’t AI Changed the World Yet?, BBC (Mar. 3, 2020), https://
    www.bbc.com/news/business-51632840 [https://perma.cc/89JZ-4PSQ] (“[A]n entirely new drug molecule ‘invented’ by AI will soon be used in human trials to treat patients who have obsessive-compulsive disorder ... .”); Stephanie Burns, Four Quick Facts About How AI is Changing the World, Forbes (Mar. 19, 2020, 4:00 AM), https://www.forbes.com/sites/stephanieburns/
    2020/03/19/four-quick-facts-about-how-ai-is-changing-the-world [https://perma.cc/6NHC-TWUX] (“[R]obot-assisted surgery – in other words, having robots in the operating room contributing to the operation – can provide real time data and information.”).

  2. [2]. See, e.g., Blade Runner (Warner Bros. 1982); The Terminator (Orion Pictures 1984); The Matrix (Warner Bros.1999).

  3. [3]. AI has three phases: Artificial Narrow Intelligence (“ANI”), Artificial General Intelligence (“AGI”), and Artificial Super Intelligence (“ASI”). ANI is typically designed to do one specific task. AGI is “defined as a machine’s ability to perform any task that a human can. . . . AGI would rival and potentially equal our intelligence . . . .” Donovan Alexander, 15 Amazing and Weird Technologies That’ll Change the World in the Next Few Decades, Interesting Eng’g (Sept. 26, 2020), https://interestingengineering.com/15-amazing-and-weird-technologies-thatll-change-the-world-in-the-next-few-decades. Once AGI is achieved, “[e]ventually, it would upgrade itself learning from us and its experiences at an exponential rate, eventually leading to an Artificial Super Intelligence, the moment where it vastly surpasses human capabilities.” Id.

    However, current AI is only at the ANI level and there is considerably doubt about whether AGI will be developed. See Ragnar Fjelland, Why General Artificial Intelligence Will Not be Realized, Humans. & Soc. Scis. Commc’ns (June 17, 2020), https://www.nature.com/articles/s41599-020-0494-4 [https://perma.cc/PC5V-QL48] (asserting AGI will not be realized due to the partly tacit nature of human intelligence and the development required).

  4. [4]. See, e.g., supra note 1 (describing innovations in the medical industry); see infra Part II (describing integration of AI in employment); Ron Schmelzer, AI Makes A Splash in Advertising, Forbes (June 18, 2020, 10:36 PM), https://www.forbes.com/sites/cognitiveworld/2020/06/
    18/ai-makes-a-splash-in-advertising [https://perma.cc/3RFL-NDGY] (describing how AI is used in advertising to identify customers, determine ad spending and placement, and “hyperpersonalize” the ads to the target customers).

  5. [5]. See infra Part II.

  6. [6]. See infra notes 20–22, 26–2929 and accompanying text.

  7. [7]. See infra notes 30–35 and accompanying text.

  8. [8]. See infra Section II.B.1.

  9. [9]. See infra Section II.B.3.

  10. [10]. See infra note 186–191.

  11. [11]. See infra Section II.A.

  12. [12]. Infra Section II.B.2.

  13. [13]. Infra Section II.B.3.

  14. [14]. Aaron J. Burstein, Employers Beware: The Illinois Artificial Intelligence Video Interview Act is Now in Effect, Kelly Drye: AD L. Access (Jan. 15, 2020), https://www.adlawaccess.com/2020/
    01/articles/employers-beware-the-illinois-artificial-intelligence-video-interview-act-is-now-in-effect [https://perma.cc/83V2-4BLW] (explaining that the Illinois Artificial Intelligence Video Interview Act was “the first state law regulating the use of AI for employee interviews”).

  15. [15]. Daron Acemoglu & Pascual Restrepo, Artificial Intelligence, Automation and Work 1 (Nat’l Bureau of Econ. Rsch., Working Paper No. 24196, 2018), https://www.nber.org/papers/w24196 [https://perma.cc/5U9F-6UC7].

  16. [16]. Techjury.net, Infographic: How AI Is Being Deployed Across Industries, Robotics Bus. Rev. (Apr. 5, 2019), https://www.roboticsbusinessreview.com/ai/infographic-how-ai-is-being-deployed-across-industries [https://perma.cc/AXJ3-G5V5].

  17. [17]. Id.

  18. [18]. See Artificial Intelligence with Human Values for Sustainable Development, UNESCO, https://
    en.unesco.org/artificial-intelligence [https://perma.cc/Z336-28GR] (providing multiple resources to look into what UNESCO believes should be the principles and ethics for the use and development of AI).

  19. [19]. See Jacob Douglas, These American Workers are the Most Afraid of A.I. Taking Their Jobs, CNBC (Dec. 11, 2019, 9:48 AM), https://www.cnbc.com/2019/11/07/these-american-workers-are-the-most-afraid-of-ai-taking-their-jobs.html [https://perma.cc/E2SX-LSVK] (“Only a little over one-quarter (27%) of all workers say they are worried that the job they have now will be eliminated within the next five years as a result of new technology, robots or artificial intelligence ... .”).

  20. [20]. For examples of AI programs for employee training, see Solutions: Zoomi Delivers Immediate Improvements in Your Training Investment, Zoomi, https://zoomi.ai/solutions [https://
    perma.cc/K543-D62T]; Dom Barnard, Get Instant Feedback on Your Presentations with Virtual Reality, Virtual Speech (Dec. 1, 2017), https://virtualspeech.com/blog/analysing-speech-in-vr-with-feedback [https://perma.cc/5VWY-ABJH]; Pluralsight Iris, Pluralsight, https://www.plural
    sight.com/product/iris [https://perma.cc/KV4R-YZN8].

  21. [21]. See Padraig Belton, How Does it Feel to be Watched at Work All the Time?, BBC News (Apr. 12, 2019), https://www.bbc.com/news/business-47879798 [https://perma.cc/WL49-JGFG] (discussing the use of tools including AI to monitor employees emails, computer usage, and movements around the office).

  22. [22]. See Bernard Marr, Robotic Process Automation and Artificial Intelligence in HR and Business Support – It’s Coming, Forbes (June 6, 2018, 12:28 AM), https://www.forbes.com/sites/bernard
    marr/2018/06/06/robotic-process-automation-and-artificial-intelligence-in-hr-and-business-support-its-coming/?sh=24d6ec475d7d [https://perma.cc/QGE6-74X9] (using HR as an example of how AI can complete the repetitive tasks to free up employees for the more skilled and interpersonal aspects of the role).

  23. [23]. See generally Peter Cappelli, Your Approach to Hiring is All Wrong: Outsourcing and Algorithms Won’t Get You the People You Need, Harv. Bus. Rev. (May 2019), https://hbr.org/2019/05/
    recruiting [https://perma.cc/Z2EN-2Q3D] (explaining the costs of the hiring process and importance of effectively finding good candidates).

  24. [24]. Id. (outlining indicators employers use to assess a candidate, such as past performance).

  25. [25]. Id. (explaining how the lengthy lists of qualifications and time to apply and interview hinder applications).

  26. [26]. Ameya Deshmukh, AI Resume Screening vs Recruiting Chatbots vs Conversational AI, Mya (Mar. 1, 2020), https://www.mya.com/blog/ai-resume-screening-avoiding-common-pitfalls [https://
    perma.cc/KDT5-U37N] (discussing advantages and disadvantages of different types of resume screening AI).

  27. [27]. Rebecca Heilweil, Artificial Intelligence Will Help Determine if You Get Your Next Job, VOX Recode (Dec. 12, 2019, 8:00 A.M.), https://www.vox.com/recode/2019/12/12/20993665/
    artificial-intelligence-ai-job-screen [https://perma.cc/EZZ4-FTHD] (discusses how different AI recruitment platforms such as ZipRecruiter keeps tracking candidates for similar positions).

  28. [28]. See, e.g., Pymetrics, https://www.pymetrics.ai [https://perma.cc/857J-MSC4]; Suited, https://
    www.wellsuited.com [https://perma.cc/VY3R-5ASM] (a platform used to evaluate candidates for banking and law firms).

  29. [29]. See, e.g., HireVue, https://www.hirevue.com [https://perma.cc/R2VR-9ACL].

  30. [30]. David J. Garraux, From the Jetsons to Reality, or Almost: What Employers Need to Know About Robots and AI in the Workplace, Nat’l L. Rev.. (Sept. 16, 2019), https://www.natlawreview.com/
    article/jetsons-to-reality-or-almost-what-employers-need-to-know-about-robots-and-ai [https://
    perma.cc/N7KN-CGQ4].

  31. [31]. Id. “According to a December 2016 article in the Harvard Business Review, AI is being used by businesses to screen out up to 70 percent of job applicants without any of the candidates having interacted with a human being.” Id.

  32. [32]. Id. For example, “Unilever has reported a 16 percent increase in new-hire diversity.” Id.

  33. [33]. Id.

  34. [34]. Matthew Jedreski, Jeffrey S. Bosley & K.C. Halm, Illinois Becomes First State to Regulate Employers’ Use of Artificial Intelligence to Evaluate Video Interviews 1 (2019), https://www.dwt.com/-/media/files/publications/2019/11/111219jedreskibosleyhal
    mil-regulates-ai-in-video-i.pdf [https://perma.cc/P6NX-TRSF].

  35. [35]. Garraux, supra note 30.

  36. [36]. B.J. Copeland, Artificial Intelligence, Britannica [hereinafter Copeland, AI], https://
    www.britannica.com/technology/artificial-intelligence [https://perma.cc/U9US-3QL5].

  37. [37]. Id.

  38. [38]. B.J. Copeland, Artificial Intelligence: Alan Turing and the Beginning of AI, Britannica, https://www.britannica.com/technology/artificial-intelligence [https://perma.cc/U9US-3QL5].

  39. [39]. See Copeland, AI, supra note 36.

  40. [40]. Ameet V. Joshi, Machine Learning and Artificial Intelligence 4 (2020).

  41. [41]. Algorithm, Merriam-Webster, https://www.merriam-webster.com/dictionary/algorithm [https://perma.cc/6LBL-ABJ2].

  42. [42]. Mehryar Mohri, Afshin Rostamizadeh & Ameet Talwalkar, Foundations of Machine Learning 1 (2d ed. 2018).

  43. [43]. See Hung-Yue Suen, Kuo-En Hung & Chien-Liang Lin, TensorFlow-Based Automatic Personality Recognition Used in Asynchronous Video Interviews, 7 IEEE Access 61018, 61018 (2019) [hereinafter Suen et al., TensorFlow] (explaining how most employers use personality assessments to determine fit for the role).

  44. [44]. Id.

  45. [45]. Id.

  46. [46]. Id. at 61020.

  47. [47]. Hung-Yue Suen, Kuo-En Hung, & Chien-Liang Lin, Intelligent Video Interview Agent Used to Predict Communication Skill and Perceived Personality Traits, 10 Hum.-centric Computing & Info. Scis., Jan. 8 2020, at 1, 3–4 [hereinafter Suen et al., Intelligent Video]; see Kelly Cahill Timmons, Pre-Employment Personality Tests, Algorithmic Bias, and the Americans with Disabilities Act, 125 Penn St. L. Rev. 389, 400–04 (2021).

  48. [48]. See Drew Harwell, A Face-Scanning Algorithm Increasingly Decides Whether You Deserve the Job, Wash. Post (Nov. 6, 2019), https://www.washingtonpost.com/technology/2019/10/22/ai-hiring-face-scanning-algorithm-increasingly-decides-whether-you-deserve-job [https://perma.cc/JB6H
    -XQ9G] (explaining how HireVue uses AI to determine job qualification).

  49. [49]. Suen et al., Intelligent Video, supra note 47, at 3.

  50. [50]. Id.

  51. [51]. Id.

  52. [52]. Id.

  53. [53]. Suen et al., TensorFlow, supra note 43, at 61019. AI using the OCEAN model looks for the following in regard to the “big five” traits:

    Openness: the degree to which an individual is imaginative and creative.

    Conscientiousness: the degree to which an individual is organized, thorough, and thoughtful.

    Extraversion: the extent to which an individual is talkative, energetic, and assertive.

    Agreeableness: the degree to which an individual is sympathetic, kind, and affectionate.

    Neuroticism: reflects the tension, moodiness, and anxiety an individual may feel.

    Id.

  54. [54]. See Harwell, supra note 48.

  55. [55]. Id.

  56. [56]. Id.

  57. [57]. See generally id. (describing how the AI analyzes someone’s response to see how well they would deal with an angry customer).

  58. [58]. Andrea Murad, The Computers Rejecting Your Job Application, BBC News (Feb. 8, 2021), https://www.bbc.com/news/business-55932977 [https://perma.cc/HS4J-GQ8N].

  59. [59]. HireVue Services, HireVue, https://www.hirevue.com/solutions/services [https://perma.cc/ZU9X-NWYS] (“[F]ull platform customization based on your business objectives.”).

  60. [60]. See The Top 12 Best Video Interviewing Platforms - April 2021, Select Software Revs., https://
    www.selectsoftwarereviews.com/buyer-guide/best-video-interview-software [https://perma.cc/
    5TR6-Y3NT] (reviewing different AI video interview platforms including VidCruiter, MyInterview, and SparkHire).

  61. [61]. “HireVue was named among the top 10 ‘Most Promising Companies’ in America by Forbes and is a five-time winner of ‘Top HR Product of the Year’ by HR Executive Magazine.” About: You’re in Good Company, HireVue, https://www.hirevue.com/about [https://perma.cc/
    73TJ-SME3]; see also Richard Feloni, I Tried the Software that Uses AI to Scan Job Applicants for Companies like Goldman Sachs and Unilever Before Meeting Them — and it’s Not as Creepy as it Sounds, Bus. Insider (Aug. 23, 2017, 11:00 AM), https://www.businessinsider.com/hirevue-ai-powered-job-interview-platform-2017-8 [https://perma.cc/ARE7-NUY8].

  62. [62]. Enterprise Video Interview Software, HireVue [hereinafter Enterprise Video Interview Software, HireVue], https://www.hirevue.com/platform/online-video-interviewing-software [https://perma.cc/
    8E3L-T8SS].

  63. [63]. Why HireVue?, HireVue, https://www.hirevue.com/why-hirevue [https://perma.cc/
    NR7Q-JGFF].

  64. [64]. Feloni, supra note 61.

  65. [65]. Frequently Asked Questions, HireVue [hereinafter FAQs, HireVue], https://www.hire
    vue.com/candidates/faq [https://perma.cc/28WC-BPBA].

  66. [66]. Feloni, supra note 61.

  67. [67]. Id.

  68. [68]. Id.

  69. [69]. Id.

  70. [70]. O’Kelly E. McWilliams, III & Jennifer R. Budoff, Artificial Intelligence in the Employment Relationship: Friend or Foe?, Nat’l L. Rev. (Aug. 21, 2019), https://www.natlawreview.com/article/
    artificial-intelligence-employment-relationship-friend-or-foe [https://perma.cc/ML8D-A34Y].

  71. [71]. Jenny R. Yang, Three Ways AI Can Discriminate in Hiring and Three Ways Forward, Urb. Inst. (Feb. 12, 2020), https://www.urban.org/urban-wire/three-ways-ai-can-discriminate-hiring-and-three-ways-forward [https://perma.cc/V923-UVMV].

  72. [72]. Id.

  73. [73]. Alexandra George, Thwarting Bias in AI Systems, Carnegie Mellon Univ. Coll. of Eng’g, https://engineering.cmu.edu/news-events/news/2018/12/11-datta-proxies.html [https://
    perma.cc/U9UC-BZN8].

  74. [74]. Id.

  75. [75]. Id.

  76. [76]. Yang, supra note 71.

  77. [77]. McWilliams, III & Budoff, supra note 70.

  78. [78]. Id.

  79. [79]. Id.

  80. [80]. Joseph O’Keefe, Danielle Moss & Tony Martinez, AI Regulation and Risks to Employers 2 (2019), https://prfirmpwwwcdn0001.azureedge.net/prfirmstgacctpwwwcdncont0
    001/uploads/3562d2b44291641560ccfaec272fdc50.pdf [https://perma.cc/PA6Y-6RCA].

  81. [81]. Id.

  82. [82]. Id.

  83. [83]. Id.

  84. [84]. Id.

  85. [85]. Id.

  86. [86]. Id.

  87. [87]. George, supra note 73.

  88. [88]. Suen et al., Intelligent Video, supra note 47, at 2.

  89. [89]. See, e.g., Lauren Rhue, Racial Influence on Automated Perceptions of Emotions 1 (Dec. 17, 2018) (unpublished manuscript), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=
    3281765 [https://perma.cc/V7L9-MRCT ] (studying the ways facial recognition AI categorizes different emotions on people based on race); Lou Safra, Coralie Chevallier, Julie Grèzes & Nicolas Baumard, Tracking Historical Changes in Trustworthiness Using Machine Learning Analyses of Facial Cues in Paintings, 11 Nature Commc’ns, Sept. 22, 2020, at 1, 2 (examining honesty through the ages by using AI).

  90. [90]. Rhue, supra note 89, at 6.

  91. [91]. Safra et al., supra note 89, at 2–3. This case study is an example of human bias being put into the algorithm. James Felton, “Trustworthiness” Study Is Basically Phrenology, Annoying Scientists, Historians, Just About Everyone, IFL Sci. (Sep. 29, 2020, 3:32 PM), https://www.iflscience.com/
    technology/trustworthiness-study-is-basically-phrenology-annoying-scientists-historians-just-about-everyone [https://perma.cc/TLH5-8S4P] (“The algorithm was trained on avatars and tested for validity on databases of faces, as rated by human participants.”).

  92. [92]. Alan Burdick, The A.I. “Gaydar” Study and the Real Dangers of Big Data, New Yorker (Sept. 15, 2017), https://www.newyorker.com/news/daily-comment/the-ai-gaydar-study-and-the-real-dangers-of-big-data [https://perma.cc/Z8WQ-QCVS].

  93. [93]. Yilun Wang & Michal Kosinski, Deep Neural Networks are More Accurate than Humans at Detecting Sexual Orientation from Facial Images, 114 J. Personality & Soc. Psych. 246, 246 (2018).

  94. [94]. Bostock v. Clayton Cnty., 140 S. Ct. 1731, 1754 (2020).

  95. [95]See generally Timmons, supra note 47 (discussing how AI personality test do not properly account for disabilities in evaluating participants); Alex Engler, For Some Employment Algorithms, Disability Discrimination by Default, Brookings: Techtank (Oct. 31, 2019), https://www.brookings.edu/blog/techtank/2019/10/31/for-some-employment-algorithms-disability-discrimination-by-default [https://perma.cc/PG23-WAJ7] (explaining how algorithms are not developed to account for disabilities).

  96. [96]. Engler, supra note 95.

  97. [97]. Id.

  98. [98]. Id.

  99. [99]. Supra notes 53–57 and accompanying text.

  100. [100]. This point was emphasized by the rights group Electronic Privacy Information Center (“EPIC”) in their official complaint filed with the FTC, noting that eye tracking can be indicative of other neurological conditions such as Parkinson’s, Alzheimer’s, and depression. Complaint at 7, In re HireVue, Inc. (F.T.C. filed Nov. 6, 2019), https://epic.org/privacy/ftc/hirevue/EPIC_
    FTC_HireVue_Complaint.pdf. [https://perma.cc/8S2P-MQ2H] (citing Ian Taylor Logan, Comment, For Sale: Window to the Soul Eye Tracking as the Impetus for Federal Biometric Data Protection, 123 Penn. State L. Rev. 779, 783–85 (2019)).

  101. [101]. Id. (citing Corinne Green & Kun Guo, Factors Contributing to Individual Differences in Facial Expression Categorization, 32 Cognition & Emotion 37, 38 (2018), (“[Autism] patients have a tendency to avoid looking at other people’s eyes, with a preference instead towards the mouth ... .”)).

  102. [102]. Engler, supra note 95 (“The problem that people with disabilities face through this kind of AI is, even if they have a strong set of positive qualities for certain jobs, the AI is unlikely to highlight those features and could generate low scores for those individuals.”).

  103. [103]. Id. (“To account for this problem, AI training data would have to include many people with diverse disabilities. Since each job type has a distinct model, this would have to be true for many different models (for context, HireVue has over 200 models). While it is possible AI software could include such a range of different individuals, it would take a tremendous effort.”); see also Timmons, supra note 47, at 444, 450 (discussing how personality tests are typically less expensive than traditional hiring methods).

  104. [104]. Timmons, supra note 47, at 444.

  105. [105]. Id.

  106. [106]. FAQs, HireVue, supra note 65.

  107. [107]. Id.

  108. [108]. Id.

  109. [109]. Id.

  110. [110]. An applicant with a disability may not wish to disclose to the potential employer their disability if it does not affect their job performance but may affect their AI video interview performance as there is a known bias against those with disabilities. There is a preconceived notion for many employers that those with disabilities are “problems in the workforce—less productive and more difficult employees to have on board.” Carrie Griffin Basas, Back Rooms, Board Rooms—Reasonable Accommodation and Resistance Under the ADA, 29 Berkeley J. Emp. & Lab. L. 59, 63 (2008). Additionally of note, according to the American Community Survey the employment rate of those with any disability in 2018 between ages 21-64 was estimated to 37.8 percent while for the same group but those without any disability employment rate was estimated to 80.0 percent. W. Erickson, C. Lee, & S. von Schrader, Cornell Univ., Yang-Tan Inst. on Emp. & Disability, 2018 Disability Status Report: United States 8 (2020), https://www.disability
    statistics.org/StatusReports/2018-PDF/2018-StatusReport_US.pdf [https://perma.cc/9MWR-JUE7].

  111. [111]. Cameron F. Kerry, Protecting Privacy in an AI-Driven World, Brookings (Feb. 10, 2020), https://www.brookings.edu/research/protecting-privacy-in-an-ai-driven-world [https://perma.cc/
    8MMJ-YZAM].

  112. [112]. Id.

  113. [113]. Id.

  114. [114]. See id. (“As artificial intelligence evolves, it magnifies the ability to use personal information in ways that can intrude on privacy interests by raising analysis of personal information to new levels of power and speed.”).

  115. [115]. Id.

  116. [116]. See Schmelzer, supra note 4.

  117. [117]. See Kerry, supra note 111 (discussing recommendations for legislation to protect the tech users right to their personal information); see generally Richard A. Bales & Katherine V.W. Stone, The Invisible Web at Work: Artificial Intelligence and Electronic Surveillance in the Workplace, 41 Berkeley J. Emp. & Lab. L. 1 (2020) (explaining how employee data protection statutes were created before AI’s massive data collections and do not require employee consent to data collection).

  118. [118]. Supra Section II.A.2.

  119. [119]. See generally Kaori Ishii, Comparative Legal Study on Privacy and Personal Data Protection

  120. [120]. Id.

  121. [121]. Id.; see Bales & Stone supra note 117, at 33–34.

  122. [122]. See generally Jedreski et al., supra note 34 (discussing the lack of guidance on how to handle the data gathered from the video interview).

  123. [123]. Supra Section II.B.3.

  124. [124]. Burstein, supra note 14.

  125. [125]. In April 2019, a bill focused on AI regulation was introduced to Congress, titled The Algorithmic Accountability Act. The bill aims to grant the Federal Trade Commission (“FTC”) authority to create and enforce regulations to monitor AI algorithms for “accuracy, fairness, bias, discrimination, privacy, and security.” Algorithmic Accountability Act of 2019, H.R. 2231, 116th Cong. § 2 (2019). While the bill was passed down to the Subcommittee on Consumer Protection and Commerce from the Committee on Energy and Commerce, there has not been any further movement on the bill. All Information (Except Text) for H.R.2231 - Algorithmic Accountability Act of 2019, Congress.Gov, https://www.congress.gov/bill/116th-congress/house-bill/2231/all-info#:~:text=Introduced%20in%20House%20(04%2F10%2F2019)&text=This%20bill%20requires%20specified%20commercial,artificial%20intelligence%20or%20machine%20learning [https://perma.cc/2P3T-DJXJ]. The identical bill, S.1108, was introduced to the Senate and referred to the Senate committee on Commerce, Science, and Transportation.

  126. [126]. New Jersey and Washington have introduced bills, but they have not been enacted. O’Keefe Et al. supra note 80, at 4. As well, New York created a “state commission to study and investigate how to regulate artificial intelligence, robotics and automation.” H.R. A01746C, 2019–2020 State Assemb., Reg. Sess. (N.Y. 2019), https://nyassembly.gov/leg/?default_fld=&leg_video=&bn=A01746&term=2019&Summary=Y&Actions=Y&Committee%26nbspVotes=Y&Floor%26nbspVotes=Y&Memo=Y&Text=Y&LFIN=Y [https://perma.cc/SN8U-GAGZ]. The commission automatically expired on December 31, 2020.

  127. [127]. Burstein, supra note 14.

  128. [128]. 2016 O.J. (L119) 679.

  129. [129]. General Data Protection Regulation (GDPR), GDPR.EU, https://gdpr.eu/tag/gdpr [https://
    perma.cc/AYP3-YCYC].

  130. [130]. Id.

  131. [131]. Commission Regulation 2016/679, art. 5, 2016 O.J. (L119) 35–36 (EU). The article specifies the data shall be “processed lawfully,” “collected for specified, explicit and legitimate purposes,” “limited to what is necessary,” “accurate,” and “processed in a manner that ensures appropriate security.” Id.

  132. [132]. Ben Wolford, What is the GDPR, the EU’s New Data Protection Law?, GDPR EU, https://
    gdpr.eu/what-is-gdpr [https://perma.cc/HP7A-VWPF].

  133. [133]. Id.

  134. [134]. Joseph E. Mutschelknaus, Top Five Data Privacy Issues that Artificial Intelligence and Machine Learning Startups Need to Know, insideBIGDATA (July 23, 2020), https://insidebigdata.com/
    2020/07/23/top-five-data-privacy-issues-that-artificial-intelligence-and-machine-learning-startups-need-to-know [https://perma.cc/Q3LQ-XWCE]. However, Washington has introduced a range of bills that would regulate data privacy. See Christopher W. Savage, Washington State Takes Center Stage for Privacy Legislation, Lexology (Jan. 28, 2020), https://www.lexology.com/library/detail
    .aspx?g=160db604-5a54-4516-953b-2024b4ef85e2 [https://perma.cc/XBP5-5YCU] (discussing the different data privacy legislation introduced in Washington’s 2020 session).

  135. [135]. Cal. Civ. Code § 1798.100 (West 2020).

  136. [136]. Id. § 1798.100–1798.125.

  137. [137]. Mutschelknaus, supra note 134.

  138. [138]. See supra text accompanying note 127.

  139. [139]. See 820 Ill. Comp. Stat. 42/1 (2021) (noting the statute became effective on January 1, 2020).

  140. [140]. Id. §5.

  141. [141]. Jedreski et al., supra note 34, at 1.

  142. [142]. Id.

  143. [143]. 820 Ill. Comp. Stat. 42/10.

  144. [144]. Id.

  145. [145]. Id. §15.

  146. [146]. Id.

  147. [147]. Jedreski et al., supra note 34, at 2.

  148. [148]. FAQs, HireVue, supra note 65 (requiring the candidate to ask the employer to have the video deleted).

  149. [149]. H.B. 0053, 102d Gen. Assemb., Reg. Sess. (Ill. 2021).

  150. [150]. Id.

  151. [151]. Id.

  152. [152]. Id.

  153. [153]. Id.

  154. [154]. Md. Code Ann., Lab. & Empl. § 3-717 (West 2021).

  155. [155]. Id.

  156. [156]. Id.

  157. [157]. Before the COVID-19 Pandemic, “[m]ore than 26 million Americans—about 16% of the total workforce— ... work[ed] remotely at least part of the time.” Zara Abrams, The Future of Remote Work, Am. Psych. Ass’n (Oct. 1, 2019), https://www.apa.org/monitor/2019/10/cover-remote-work [https://perma.cc/JD4D-WYUG]. Due to the COVID-19 pandemic, “tens of millions of American workers” were forced to work from home. Derek Thompson, The Workforce is About to Change Dramatically, Atlantic, (Aug. 6, 2020), https://www.theatlantic.com/ideas/
    archive/2020/08/just-small-shift-remote-work-could-change-everything/614980 [https://perma.cc/
    2V5V-B4W4]. After the pandemic, “one in six workers is projected to continue working from home or co-working at least two days a week.” Id.

  158. [158]. Virtual Vocations, a remote jobs board, reported that August 2020, compared to the year prior had, “a 67% increase in the total number of available remote job openings.” Kimberly Back, September 2020 Remote Jobs Report, Virtual Vocations Telecommuting News, (Sept. 14, 2020), https://www.virtualvocations.com/blog/telecommuting-news/september-2020-remote-jobs-report [https://perma.cc/8Q3K-XV7M].

  159. [159]. Jedreski et al., supra note 34, at 1.

  160. [160]. 820 Ill. Comp. Stat. 42/5 (2021).

  161. [161]. Id.

  162. [162]. Id.

  163. [163]. AI integration into video interviews is an invention of the last decade. “HireVue launched its AI assessment service in 2014 as an add-on to its video-interview software
    ... .” Harwell, supra note 48.

  164. [164]. Employers may use similar video interview technology to conduct asynchronous interviews without the AI evaluation. Some of the same companies that offer the AI video interviews supply it as an optional additional service to purchase beyond the asynchronous interview technology. HireVue Hiring Assessment Software, HireVue, https://www.hirevue.com/
    platform/assessment-software [https://perma.cc/D5AY-JFQQ] (“Combine the interview and assessment into a single unified experience, in less than 20 minutes.”). As employers may use either asynchronous interviews with or without the AI assessment, the employer needs to be clear about which technology is being used.

  165. [165]. 820 Ill. Comp. Stat. 42/5.

  166. [166]. Id.

  167. [167]. Id.

  168. [168]. Id.

  169. [169]. Id.

  170. [170]. Id.

  171. [171]. Id. § 10.

  172. [172]. Id. § 15.

  173. [173]. Id. § 5.

  174. [174]. Id.

  175. [175]. Jedreski et al., supra note 34, at 1–2.

  176. [176]. Id.

  177. [177]. Supra notes 96–103 and accompanying text.

  178. [178]. Jedreski et al., supra note 34, at 2.

  179. [179]. Id.

  180. [180]. 820 Ill. Comp. Stat. 42/10 (2021).

  181. [181]. See Jedreski et al., supra note 34, at 1–2 (discussing the extracted data being used for other purposes such as evaluating the interview questions or for functionality of the AI system); Bales & Stone, supra note 117, at 33–34.

  182. [182]. Jedreski et al., supra note 34, at 2.

  183. [183]. Id. at 1–2.

  184. [184]. Id. at 2.

  185. [185]. Id.

  186. [186]. See, e.g., virtual meetings (GoToMeeting, Zoom, Skype), remote employee monitoring software (Time Doctor, Roadmap, Tick), project management programs (Workfront, Monday, Basecamp).

  187. [187]. “There was a time when remote work as we know it wasn’t even a possibility, because the technology didn’t exist. If your colleagues and business partners wanted to get in touch with you when you were out of the office, they couldn’t email, text or direct-message you.” Sean Peek, Communication Technology and Inclusion Will Shape the Future of Remote Work, Bus. News Daily (Mar. 18, 2020), https://www.businessnewsdaily.com/8156-future-of-remote-work.html [https://perma.cc/
    LUY5-4EMQ]. “[T]echnology ha[s] advanced to the point where some businesses thrive with completely remote teams.” Id. U.S. employees who work virtually, “[b]etween 2005 and 2015,
    ... increased by 115%.” Abrams, supra note 157.

  188. [188]. See Dylan Matthews, Covid-19 Taught Companies How to Run Remote Offices. That Could Transform Developing Countries, Vox (Sept. 21, 2020, 2:00 PM), https://www.vox.com/future-perfect/21445830/work-from-home-covid-19-offshoring [https://perma.cc/2G54-PXA2] (predicting “tele-migration” increasing “in the near future as employers in rich Western developed countries begin to export tasks to less expensive workers in the developing world, exploiting technologies”). Companies have already begun using distant employees for service roles such as call centers. Id.

  189. [189]. See Thompson, supra note 157 (discussing how the pandemic caused many jobs to become virtual with many companies including Google, Facebook, and Twitter planning on virtual workers being a long-term plan and possibly permanent).

  190. [190]. Abrams, supra note 157 (“Many workers view telecommuting as a job perk, with more than half seeking the arrangement as a way to improve work-life balance.”).

  191. [191]. See Harwell, supra note 48 (“[M]ore than 700 companies have used [HireVue AI video interviews] for nearly 12 million interviews worldwide.”).

  192. [192]. See, e.g., Complaint at 1, In re HireVue, Inc. (F.T.C. filed Nov. 6, 2019), https://epic
    .org/privacy/ftc/hirevue/EPIC_FTC_HireVue_Complaint.pdf [https://perma.cc/EM2R-3XMW].

  193. [193]. Patricia Barnes, EPIC Asks Federal Trade Commission to Regulate Use of Artificial Intelligence in Pre-Employment Screenings, Forbes (Feb. 3, 2020, 7:03 PM), https://www.forbes.com/sites/
    patriciagbarnes/2020/02/03/group-asks-federal-trade-commission-to-regulate-use-of-artificial-intelligence-in-pre-employment-screenings/?sh=312c529f2b54 [https://perma.cc/7ZMC-F4E5].

  194. [194]. In re HireVue, EPIC.org, https://epic.org/privacy/ftc/hirevue [https://perma.cc/
    GLB9-W973].

  195. [195]. Id.

  196. [196]. Id.

  197. [197]. See supra Section III.A.

  198. [198]. 820 Ill. Comp. Stat. 42/5 (2021).

  199. [199]. Id.

  200. [200]. Most AI video interviews are asynchronous and allow the interviewee to do the 20 to 30-minute interview at whichever time works best for the interviewees schedule. However, some companies do provide for both asynchronous interviews and live interviews. Enterprise Video Interview Software, HireVue, supra note 62.

  201. [201]. 820 Ill. Comp. Stat. 42/5.

  202. [202]. Id.

  203. [203]. See supra Section III.A.2.

  204. [204]. 820 Ill. Comp. Stat. 42/5.

  205. [205]. Federal law holds employers liable in violation of Title VII “only when they ‘fail or refuse to hire,’ ... someone because of a statutorily protected characteristic like sex.” Bostock v. Clayton Cnty. 140 S. Ct. 1731, 1740 (2020) (quoting 42 U.S.C. § 2000e-2(a)(1) (2020)). As Scalia notes in the dissent, “Even idiosyncratic criteria are permitted; if an employer thinks that Scorpios make bad employees, the employer can refuse to hire Scorpios.” Id. at 1762 (J. Scalia, dissenting); see also Farver v. Mcarthy, 931 F.3d 808, 812 (8th Cir. 2019) (“Employers may consider many experiences and skill sets when making a hiring decision.”).

  206. [206]. 820 Ill. Comp. Stat. 42/10.

  207. [207]. Id.

  208. [208]. Id.

  209. [209]. Id.§ 15.

  210. [210]. Id.

  211. [211]. Id.

  212. [212]. Id.

  213. [213]. FAQs, HireVue, supra note 65.

  214. [214]. 820 Ill. Comp. Stat. 42/10.

  215. [215]. H.B. 0053, 102d Gen. Assemb., Reg. Sess. (Ill. 2021).

  216. [216]. Id.

  217. [217]. See McWilliams, III & Budoff, supra note 70 (examining the role AI algorithms can play in removing human bias from the hiring process). “Unlike human decisions, decisions made by AI could in principle (and increasingly in practice) be opened up, examined, and interrogated.” Jake Silberg & James Manyika, Tackling Bias in Artificial Intelligence (and in Humans), McKinsey & Co.: McKinsey Glob. Inst. (June 6, 2019), https://www.mckinsey.com/featured-insights/artificial-intelligence/tackling-bias-in-artificial-intelligence-and-in-humans# [https://perma.cc/W9JW-3SCE]. “AI can reduce humans’ subjective interpretation of data, because machine learning algorithms learn to consider only the variables that improve their predictive accuracy, based on the training data used. In addition, some evidence shows that algorithms can improve decision making, causing it to become fairer in the process.” Id.

  218. [218]. See Silberg & Manyika, supra note 217 (discussing the roles humans need to take in evaluating the AI decision and accounting for fairness).

  219. [219]. Supra notes 95–102 and accompanying text (explaining the difficulties of accounting for differences in disabilities in AI video analysis).

  220. [220]. Supra notes 103–05 and accompanying text (discussing the expenses and work associated with accounting for disabilities in AI video analysis).

  221. [221]. FAQs, HireVue, supra note 65 and accompanying text (“Most employers today are ready to support neurodiverse as well as candidates with disabilities if you prefer to take an interview or assessment another way... . You can ... request accommodations from your recruiter or the hiring team at that company.”).

  222. [222]. Under Title I of the Americans with Disabilities Act “[n]o covered entity shall discriminate against a qualified individual on the basis of disability in regard to job application procedures, the hiring, advancement, or discharge of employees, employee compensation, job training, and other terms, conditions, and privileges of employment.” 42 U.S.C. § 12112(a) (2020).

  223. [223]. See Shelly Maciujec, Unconscious Bias Towards People with Disabilities in the Workplace, Ability Mag., https://abilitymagazine.com/unconscious-bias-pwds-workplace [https://perma.cc/
    7ATJ-36PW] (discussing the high rate of unconscious bias against people with disabilities and lower employment rates).

  224. [224]. James Manyika, Jake Silberg & Brittany Presten, What Do We Do About the Biases in AI?, Harv. Bus. Rev. (Oct. 25, 2019), https://hbr.org/2019/10/what-do-we-do-about-the-biases-in-ai [https://perma.cc/KX38-WECX] (discussing an AI algorithm used in 1988 for medical school admissions was “biased against women and those with non-European names”).

  225. [225]. As an example, in HireVue’s FAQ asking about how HireVue deals with algorithmic bias:

    As for algorithmic bias, HireVue has pioneered many new methods for preventing bias in our assessment tests (algorithms). The main way we do this is by making sure that the assessments focus only on the skills and competencies that research shows is important to success in the job you’re applying for. Assessments never look at age, race, gender, or even things on your resume such as GPA or the college you attended. Only skills and competencies are measured.

    FAQs, HireVue, supra note 65.

  226. [226]. Jeremy Kahn, HireVue Drops Facial Monitoring amid A.I. Algorithm Audit, Fortune (Jan. 19, 2021, 11:01 AM), https://fortune.com/2021/01/19/hirevue-drops-facial-monitoring-amid-a-i-algorithm-audit/ [https://perma.cc/2EP5-8535]. However, HireVue still uses AI to evaluate candidates based on word choice, tone, etc. See Kiki Leutner, Josh Liff, Lindsey Zuloaga, & Nathan Mondragon, HireVue White Paper, HireVue’s Assessment Science 9 (2021), https://
    webapi.hirevue.com/wp-content/uploads/2021/03/HireVue-Assessment-Science-whitepaper-2021.pdf [https://perma.cc/8HRZ-ZYFT].

  227. [227]. HireVue’s “About” page focuses on how by using their products, HireVue “clients lead their industries with 50% faster growth, 29% less turnover, and 13% more top performers.” About, HireVue, https://www.hirevue.com/about [https://perma.cc/N4CB-5FEA]. However, HireVue does state the company “has always been committed to good science that creates a level playing field for all candidates.” Our Science, HireVue, https://www.hirevue.com/our-science [https://
    perma.cc/VA2H-C86S].

  228. [228]. See Charles J. Muhl, The Employment-at-Will Doctrine: Three Major Exceptions, Monthly Lab. Rev., Jan. 2001, at 3, 3, https://www.bls.gov/opub/mlr/2001/01/art1full.pdf [https://perma.cc/
    E2G2-VWL6] (“Rather than seeing the relationship as being on equal footing, courts and legislatures slowly began to recognize that employers frequently have structural and economic advantages when negotiating with potential or current employees.”).

  229. [229]. Supra note 205 and accompanying text.

*

J.D. Candidate, The University of Iowa College of Law 2022; B.A., University of Minnesota, 2015.