Professionalism, Mobley v. Workday and the Ethics of Algorithmic Hiring

In the modern job market, the first “person” to read a résumé is rarely a human being. It is increasingly common for algorithmic systems to act as the initial gatekeeper for employment, screening thousands of applicants instantly. While these tools offer efficiency to employers, they introduce complex ethical questions regarding bias, transparency, and accountability. This chapter examines these issues through Mobley v. Workday, Inc., a lawsuit filed in 2023. By analyzing allegations that Workday’s AI screening tools produced discriminatory outcomes based on race, age, and disability, we can derive generalizable lessons about professional responsibility in automated hiring. Mobley v. Workday remains ongoing; this chapter reflects developments through May 2025 and may be updated as the case evolves.

Introduction

Algorithmic screening is widespread

For decades, the standard application process involved a human recruiter physically reviewing documents. Today, the scale of digital applications necessitates the use of Applicant Tracking Systems (ATS). These platforms help employers collect, sort, and rank candidates. According to industry analysis by Jobscan, 97.8% of Fortune 500 companies use ATS software to manage their hiring pipelines.[1] Within these systems, AI or machine learning components are often used to score candidates on their “fit” for a role, effectively deciding who proceeds to a human interview and who is rejected instantly.

The professionalism question

The central ethical tension revolves around the delegation of decision-making. When an employer (client) delegates screening to a software vendor (creator) using complex algorithms (tool), responsibility lines blur. If the tool discriminates, who is ethically responsible? The engineers who built the model, the vendor who sold it as “unbiased,” or the employer who deployed it without auditing the results?

Misconceptions about automated rejection

It is important to clarify that not all automated rejections are “AI decisions.” Many ATS rejections are based on human-configured “knockout questions” (e.g., “Do you have a valid work visa?”). If an applicant answers “no,” they are disqualified by a static rule. However, the ethical stakes rise when systems move beyond static rules to algorithmic scoring, where software infers quality based on historical patterns, potentially serving as a “black box” neither applicant nor employer fully understands.[2]

The algorithmic hiring ecosystem

From resume to recommendation

The hiring ecosystem is a “funnel.” Candidates submit applications, and the ATS parses the text, organizing unstructured data (like a PDF résumé) into structured fields (Skills, Education, Experience). In advanced systems, AI models analyze this data to rank candidates or assign a “match score.” This recommendation engine is socio-technical: relying on code, but calibrated by human goals. Recruiters may set the system to only show candidates with a “5-star match,” effectively rendering lower-ranked candidates invisible.

The drive for efficiency

Employers adopt these tools primarily for speed and scale. A single job posting at a major tech company may receive thousands of applications. Reviewing these manually is practically impossible for a standard HR team. Vendors market these tools as providing consistency and objectivity, arguing that algorithms remove human biases (such as favoring an applicant who went to the same university as the recruiter).

Ethical failure modes

However, efficiency can come at the cost of fairness. The primary failure mode is the replication of historical bias. If a system is trained on data from a company that has historically hired mostly young, white men, the algorithm may learn to downgrade résumés containing markers of other demographics. Researchers have warned that variables like zip codes, gaps in employment, or graduation dates can act as proxies for race or age.[3] Because these candidates are filtered out before reaching a human, discrimination becomes invisible; the employer sees a “qualified” pool and may never realize whom they missed.

Workday as a hiring intermediary

The role of Workday

Workday is a leading provider of enterprise cloud applications for finance and human resources. While offering many services, their talent acquisition suite is ubiquitous in large-scale hiring. For many job seekers, “applying for a job” means creating an account on a Workday-hosted portal. The platform handles the entire application lifecycle, from intake to the offer letter.[4]

AI and “skills intelligence”

Workday markets its tools as utilizing responsible AI to help companies match talent to opportunities. They emphasize features like “skills intelligence,” which infers a candidate’s capabilities even if not explicitly listed, and intelligent ranking to surface “best-fit” candidates. The vendor promotes these features to reduce bias by focusing on skills rather than pedigree. However, the efficacy of these claims depends on how specific models are built and trained.[4]

The configuration dilemma

A key aspect of Workday’s legal defense is the distinction between the tool and the configuration. As a multi-tenant cloud provider, Workday's core software is the same, but each customer (employer) configures it differently. One employer might turn off AI ranking, while another leans on it heavily. This variance complicates professional accountability. If a candidate is rejected, was it Workday’s algorithm, or the employer's specific criteria? This ambiguity makes auditing difficult and is a central theme in Mobley v. Workday.

Case study: Mobley v. Workday

The allegations

Derek Mobley, a Black man over 40 suffering from anxiety and depression, filed a class action against Workday in February 2023. Mobley alleged he applied to over 100 positions at companies using Workday’s screening tools since 2017. Despite holding degrees in finance and network systems administration, he was rejected from every position—often immediately or at hours suggesting automated decision-making.[5] Mobley did not sue the 100 individual employers. Instead, he sued Workday, alleging the vendor’s algorithm was systematically biased based on race, age, and disability. He claimed the screening tools acted as a common agent, creating a “pattern and practice” of discrimination.

Workday’s defense

Workday moved to dismiss, arguing a strict separation of duties. Their defense relied on the legal definition of an “employer” or “employment agency.” Workday argued they are merely a software vendor (an information processor) and that actual hiring decisions are made by the companies purchasing their software. They argued they cannot be held liable for discrimination because they do not employ the applicants, nor do they act as a traditional recruitment agency.[6]

Procedural history and rulings

The case has seen significant developments reshaping the liability landscape for AI vendors: July 2024 Ruling: The U.S. District Court dismissed Mobley’s claim that Workday is an “employment agency.” However, the court allowed the case to proceed on the theory that Workday could be an “agent” of the employers. The judge ruled it plausible that Workday’s software performs a delegated hiring function (screening). Crucially, the court allowed “disparate impact” claims, acknowledging that facially neutral algorithms can violate anti-discrimination laws.[6] May 2025 Developments: The court granted conditional certification for a nationwide collective action regarding Age Discrimination in Employment Act (ADEA) claims. This allowed rejected applicants aged 40 and older to join the lawsuit, suggesting the court took seriously the possibility of a “unified policy” of discrimination embedded in the software.[7]

Professional ethics lessons

The Mobley case serves as a powerful case study for computer scientists, data engineers, and HR professionals. Regardless of the final verdict, the case highlights three major areas of professional responsibility.

1. Disparate impact and data provenance

In professional ethics, “intent” matters less than “outcome.” Disparate impact refers to practices nondiscriminatory in intent but discriminatory in operation. Professionals building hiring algorithms must recognize that historical data is almost always biased. If an algorithm is trained to replicate “successful” past employees, and those employees were predominantly of one demographic, the algorithm will use proxy variables to maintain that status quo. For example, an algorithm might not explicitly filter by “race,” but might penalize zip codes or university names that correlate with race. Lesson: Professionals have a duty to audit training data for representation and test models for disparate impact across protected classes before deployment.

2. The danger of automation bias

A major ethical risk is automation bias, the tendency to over-rely on automated suggestions. Even if a vendor like Workday claims their tool provides “recommendations,” time-strapped recruiters often treat them as final. If the AI ranks a candidate as a “bad fit,” the recruiter rarely overrides it. Lesson: Labeling a system as “advisory” does not absolve creators of responsibility for how it is used. Systems should be designed with “friction” that encourages human review of rejections, rather than enabling frictionless, mass rejections.[8]

3. Alignment with the ACM Code of Ethics

These issues are addressed by the ACM Code of Ethics. Principle 1.2 states “Avoid Harm,” which includes the negative consequences of data systems that result in discrimination. Principle 1.4 requires professionals to “Be fair and take action not to discriminate.” In Mobley, declaring “we are just the vendor” is not a sufficient ethical defense. If a professional builds a tool they know (or should know) can filter out protected groups based on flawed proxies, they fail to uphold the standard of fairness required by the profession.[9]

4. A duty of care framework

The industry is shifting toward a “Shared Responsibility” model, similar to cloud security. For Vendors: The duty is Transparency and Testing. Vendors have ethical reasons to provide audit-relevant transparency while protecting trade secrets. They should document how the model scores candidates and provide evidence of bias audits. For Employers: The duty is Monitoring. Mobley highlighted the difficulty of untangling which employer enabled which feature. Employers must actively monitor their hiring funnels for skewed demographics, rather than assuming the “black box” works correctly. For Engineers: The duty is Accessibility. Systems must accommodate non-standard applicants. For example, gamified assessments or timed responses may unfairly screen out qualified neurodivergent candidates.[10]

Governance baseline: beyond compliance

Regulation is slowly catching up to technology. New York City’s Local Law 144, enforced July 5, 2023, requires that automated employment decision tools (AEDT) undergo annual bias audits and candidates be notified. However, for the ethical professional, compliance is a floor, not a ceiling. Mobley illustrates that complying with current statutes might not be enough to prevent harm. Even if a tool passes an audit, it may still produce unfair outcomes. True professionalism requires a commitment to ongoing fairness, transparency, and accountability, proactively identifying and remediating harms even when not explicitly forced by a regulator.

Conclusion

The case of Mobley v. Workday is an early, high-profile test of legal frameworks surrounding AI in hiring. It challenges the industry defense that software vendors are merely platform providers with no stake in how their tools are used. The court’s willingness to entertain the theory that a software vendor can be an “agent” of discrimination suggests shared liability between those who build the code and those who use it. For professionals, the lessons are clear: Algorithms are not neutral. They encode the biases of their creators and their data. Delegation is not absolution. You cannot outsource ethical obligations to a vendor or client. Design for auditability. Systems impacting livelihoods must explain their decisions. As AI reshapes the labor market, ethical computer scientists must prioritize fairness as highly as efficiency, ensuring the “future of work” is accessible to everyone. Because the litigation is still underway, the legal boundaries described here should be read as emerging, and future editions may incorporate additional rulings clarifying vendor and employer liability.

References

• ↑ "2025 Applicant Tracking System (ATS) Usage Report: Key Shifts and Strategies for Job Seekers". Jobscan. 2025-07-14. Retrieved 2025-12-13.
• ↑ Paradis, Tim (2025-12-07). "AI isn't likely throwing out your résumé. A human is". Business Insider. Retrieved 2025-12-13.
• ↑ Datta, Anupam (2017-07-25). "Proxy Non-Discrimination in Data-Driven Systems". arXiv:1707.08120.
• ↑ a b "Talent Acquisition and Recruiting Software". Workday. Retrieved 2025-12-13.
• ↑ "Millions of resumes never make it past the bots: one man is trying to find out why". Northeastern University College of Social Sciences and Humanities. 2025-06-23. Retrieved 2025-12-13.
• ↑ a b "Mobley v. Workday, Inc., Order Granting in Part and Denying in Part Motion to Dismiss" (PDF). 2024-07-12. Retrieved 2025-12-13.
• ↑ "Mobley v. Workday, Inc., Order Granting Conditional Certification" (PDF). 2025-05-16. Retrieved 2025-12-13.
• ↑ Skitka, L. J. (1999). "Does automation bias decision-making?". International Journal of Human-Computer Studies. 51 (5): 991–1006.
• ↑ "ACM Code of Ethics and Professional Conduct". Association for Computing Machinery. 2018. Retrieved 2025-12-13.
• ↑ "The Americans with Disabilities Act and the Use of Software, Algorithms, and Artificial Intelligence to Assess Job Applicants and Employees". U.S. Equal Employment Opportunity Commission. 2022-05-12. Retrieved 2025-12-13.

External links

• EEOC brief in Mobley v. Workday
• NYC DCWP page on AEDTs and Local Law 144

Where this page came from

This page was imported from Wikibooks. From “Professionalism” on Wikibooks, by its contributors, under CC BY-SA 4.0. Changed here: set as a page; navigation and edit links left out; each image under its own licence, credited in its caption.

Nobody has written it yet — it is the source material at a new address, which is why search engines are asked to skip it and why no one earns from it. It is up for grabs: take it on, and it is yours to rewrite and to earn from.