ROBOT RECRUITERS: HOW EMPLOYERS & GOVERNMENTS MUST CONFRONT THE DISCRIMINATORY EFFECTS OF AI HIRING.

• Author: Marlowe, Alexandra N.

1. Introduction

   "The real question is, when will we draft an artificial intelligence bill of rights? What will this consist of? And who will get to decide that?" (1) Twenty years ago these questions would have never been asked, let alone seriously considered, but as a result of rapidly advancing technology, artificial intelligence ("AI") has permeated everyday life through a multiplicity of technologies, such as ridesharing applications, voice-to-text, Google Maps, and social media. (2) Despite its prevalence and influence within society, AI is one of the most poorly understood concepts, with the general public and many top business leaders lacking a detailed understanding of how the technology operates. (3) The confusion is understandable given that AI is often depicted as a rogue robot in Hollywood movies, and even compared to a "demon" by Elon Musk. (4) While the comparison that Musk makes is distressing, it was an attempt to draw attention to the desperate need for regulation in the ever-expanding field of AI technology. (5)

   AI offers many advantages to industries, however, the advantages must not completely blind society to the dangers lurking in the shadows. (6) Research continuously indicates that AI algorithms replicate and amplify bias, despite the perpetuated claim that they are objective and have the potential to eradicate human bias. (7) In spite of these major flaws, many employers have implemented AI technology within their hiring process. (8) If this trend continues, employers must acknowledge the dangerous pitfalls associated with its use, and take steps to mitigate the known dangers or risk contributing to the perpetuation of discrimination in the workplace. (9) Illinois took a significant step to address some of the potential risks presented by AI software by passing and enacting the Illinois Artificial Intelligence Video Interview Act ("Illinois Act"). (10) The Illinois Act focuses solely on privacy protections for individuals and missed a momentous opportunity to address pressing bias and discrimination concerns that accompany the use of AI software. (11) While privacy rights are vital, laws specifically addressing both privacy and discrimination will become crucial as more employers begin implementing AI software in their hiring processes. (12) There is no time like the present to enact meaningful regulations and guidelines to ensure that Americans are protected from hiring bias and discrimination resulting from AI. (13)

   There is a critical need for anti-discrimination regulations as they relate to the use of AI software in hiring processes, a concern hedged by the Illinois Act. The public has a general distrust of AI technology, and is wary of its increasing presence in society. While there are anti-discrimination laws in place, they do not provide adequate protections against the many flaws associated with employing AI software in the hiring process. This Note aims to demonstrate that AI software perpetuates many of society's inherent biases. States and the federal government should take steps to protect Americans from discrimination by enacting anti-discrimination regulations specifically related to AI in the hiring process and employers utilizing the AI technology within their businesses must take responsibility and accountability for its use.

2. History

   1. Defining Artificial Intelligence Technology

      The field of AI is seldom understood, despite existing since 1956. (14) This incomprehension is evidenced by the lack of a universally accepted definition. (15) The field is most often characterized as a branch of computer science. (16) Beyond that characterization, AI has a multiplicity of definitions, such as the "study of ideas to bring into being machines that respond to stimulation consistent with traditional responses from humans, given the human capacity for contemplation, judgment and intention." (17) Moreover, the machines "[p]roduced by human skill and labor... should conduct themselves in agreement with life, spirit and sensitivity, though in reality, they are imitations." (18) AI has also been depicted as machines that can both act and make decisions for themselves, similar to the way in which a human or an animal makes choices. (19) The machines are intentional in their decisions, and can adapt their behavior based upon their own learning and reasoning capacities. (20)

      Within the broad scope of AI, there are many subsets, such as reactive machines, limited memory, theory of mind, and self-awareness. (21) The heavily discussed subset called machine learning is responsible for many of the known AI breakthroughs. (22) Machine learning employs algorithms to locate patterns within data. (23) It allows the machine to learn from data and improve its decision-making skills over time. (24) This technology is the force behind platforms like YouTube, Spotify, and Netflix, as well as the source of what users see on their Facebook and Twitter feeds. (25) The goal of services like Netflix and Facebook is to gather large amounts of data about the user in order to make predictions about what they might want next. (26) In order for machine learning to operate, three components are typically required: datasets, features, and algorithms. (27) Machine learning employs algorithms trained to find patterns and features within large amounts of data. (28) After determining the distinguishing patterns and features, it can make decisions and predictions on new data. (29)

      Machine learning can be further subdivided into what is known as "deep learning." (30) The technology, inspired by the structure of the human brain, utilizes deep learning algorithms and multi-layered neural networks. (31) Deep learning models utilize large amounts of data that pass through multiple layers of calculations, each applying weights and biases, to continuously adjust, enhance, and improve outcomes. (32) Deep learning expands machine learning by further amplifying the machine's ability to identify patterns, by relying on an abundance of "layers of simple computational nodes." (33) It has been said that deep learning is "machine learning on steroids." (34)

      Machine learning and deep learning have greatly advanced their capabilities, however, each is still imperfect and can host a wide variety of defects, including bias. (35) It is critical to understand that while AI has significant technological capabilities, it is still created by humans, who are prone to flaws. (36) As a result, bias can creep into the process at the very early stages, such as when a computer scientist is first designing a model, as well as when the data is being collected. (37) For instance, the models created are likely to be riddled with the biases of white men, given that computer scientists are overwhelming white and male. (38) Discriminatory outcomes are often produced even when there is no discriminatory intent. (39) History demonstrates that fixing these flaws can be arduous because processes often do not plainly show bias. (40) Moreover, these models are not necessarily designed with a thoughtful eye toward bias, or with a process in place that can detect potential bias. (41) These discriminatory outcomes are prevalent across many industries, but are most evident in criminal procedure and employment contexts. (42) As is seen in other settings where bias and discrimination exist, such as housing discrimination, finding a solution to the problem can be challenging. (43)

   2. Perceptions of Artificial Intelligence

      AI technology pervades society, playing an active role in our development, whether or not individuals are aware of its presence. (44) The technology has spawned the creation of natural language tools, automated financial investing, virtual travel booking, proactive healthcare management, and disease mapping. (45) It has the potential to enhance efficiency, reduce costs for businesses, expand creativity, curtail energy costs, and lessen the number of mundane tasks for employees to complete. (46) Yet, despite the gambit of technological advancements and its capability to streamline business practices, there is a deep lack of confidence in AI. (47)

      The public's lack of confidence and wariness is illustrated by the responses to a survey conducted by the Center for Governance of AI, Future of Humanity Institute at the University of Oxford. (48) The survey looked at 2,000 responses from American adults in 2018. (49) In their responses, Americans expressed mixed support for the development of AI, with 41% responding that they "somewhat" or "strongly" support the development of AI, and 22% say they "somewhat" or "strongly" opposed its development. (50) In addition, 82% of Americans believed that robots and/or AI "should be carefully managed," demonstrating the public's vigilant attitudes toward AI. (51) While this survey indicated that Americans are wary of AI, there is evidence of cautious optimism as well. (52) In another survey, 51% of Americans indicated that they support continuing AI research, while only 31% oppose the research. (53) These responses seem to indicate that Americans are open to AI's growing prevalence and its impact on their futures. (54)

      It is no mystery why this public perception persists, as big players in the field, such as Bill Gates and Elon Musk, have expressed their apprehension. (55) Elon Musk went as far as saying that AI technology could be our "biggest existential threat," a vexing statement given his knowledge and expertise in this subject. (56) Numerous other AI leaders have expressed fears of potential job displacement and their worry about privacy issues raised by the technology. (57) These perceptions by experts, who should have the deepest understanding of the technology's capabilities, underly and further enhance anxieties surrounding AI and its potential impact across industries and society at large. (58)

      Despite the mixed perception of AI, there is a business case to be made given that the...