Project – Algorithmic Recruitment and Perceived Fairness in Employee Selection: A Study of Selected Recruitment Firms in Lagos State

CHAPTER ONE

INTRODUCTION

1.1 Background to the Study

The recruitment and selection of employees are among the most important functions of human resource management because the quality of people recruited into an organisation can influence productivity, innovation, service quality, employee commitment and long-term organisational performance. Traditionally, recruitment decisions have relied heavily on human judgement, including the assessment of curriculum vitae, interviews, references, tests and other selection procedures. However, the rapid development of digital technologies has increasingly transformed recruitment practices. Recruitment organisations now use applicant-tracking systems, automated screening tools, machine-learning models, online assessments and artificial intelligence (AI) to identify, evaluate and rank potential candidates. This development has given rise to what is commonly described as algorithmic recruitment.

Algorithmic recruitment refers to the use of computational procedures, automated systems and algorithmic models to support or perform activities involved in sourcing, screening, ranking, assessing and selecting job applicants. These systems may analyse information contained in applicants’ CVs, application forms, online profiles, assessment results and other data to determine whether candidates meet specified requirements. Köchling and Wehner (2020) explain that algorithmic decision-making has increasingly entered human resource management and can potentially improve efficiency while simultaneously creating concerns about discrimination and fairness.

The growing adoption of algorithmic recruitment is partly driven by the increasing volume of job applications received by recruitment organisations. Human recruiters may have to evaluate hundreds or thousands of applications for particular positions. Manually reviewing every application can be time-consuming, expensive and susceptible to inconsistent judgement. Automated systems can process large amounts of information rapidly and identify candidates according to predetermined criteria. Consequently, algorithmic recruitment is increasingly viewed as a means of improving recruitment efficiency and managing large applicant pools.

Despite these potential benefits, algorithmic recruitment raises important questions concerning fairness. Recruitment decisions have significant consequences for applicants because they can determine access to employment opportunities, income and career development. If an algorithm systematically disadvantages certain groups of applicants, the consequences can be substantial. The concern becomes more significant when applicants do not understand how automated systems evaluate them or when organisations cannot adequately explain why an applicant was rejected.

Fairness in employee selection refers broadly to the extent to which selection procedures are considered just, equitable, consistent and free from inappropriate discrimination. Gilliland (1993) established an influential framework for understanding applicant reactions to selection procedures and argued that perceptions of procedural justice can influence applicants’ reactions to organisations. Candidates may evaluate not only the outcome of a selection process but also the fairness of the procedures used to reach that outcome.

Perceived fairness is particularly important because applicants may respond negatively to recruitment procedures they consider unfair, even when the final decision is objectively defensible. Hausknecht, Day and Thomas (2004), in their meta-analysis of applicant reactions to selection procedures, found that applicants’ perceptions of selection procedures are related to important outcomes such as organisational attractiveness, recommendation intentions and job acceptance. This suggests that fairness in recruitment has implications extending beyond individual selection decisions.

Algorithmic recruitment creates a new dimension of this fairness problem. Human recruiters can potentially explain decisions, respond to questions and consider contextual information about applicants. An algorithm, by contrast, may produce a recommendation or ranking based on a complex combination of variables that applicants and even recruiters may not fully understand. Burrell (2016) identifies opacity as a central problem in machine-learning systems because some algorithms are difficult to interpret, particularly when their internal operations are not transparent to users.

The issue of algorithmic opacity has important implications for recruitment. If an applicant is rejected by an automated system, the applicant may not know whether the decision resulted from qualifications, work experience, keywords, educational background, assessment performance or another factor. When applicants cannot understand the basis of a decision, they may perceive the recruitment process as unfair. Consequently, transparency and explainability have become important issues in discussions about algorithmic decision-making.

Another concern is algorithmic bias. Algorithms do not necessarily eliminate human bias simply because they are automated. Algorithms are developed using data, rules and assumptions generated by human beings. If historical recruitment data reflect existing inequalities or discriminatory practices, an algorithm trained on such data may reproduce or even amplify those patterns. Barocas and Selbst (2016) demonstrate that data-driven decision-making can produce discriminatory outcomes even when systems do not explicitly use protected characteristics such as race or gender.

This concern was illustrated by Raghavan et al. (2020), who examined claims surrounding algorithmic hiring and found that the development and deployment of hiring algorithms involve important questions about validation, bias, fairness and accountability. Their analysis suggests that organisations should not assume that automated hiring systems are automatically objective simply because they are based on computational processes.

The problem of algorithmic bias is particularly important in employee selection because seemingly neutral variables may function as proxies for socially sensitive characteristics. For example, variables associated with educational institutions, geographic location, employment history or patterns of online behaviour may indirectly correlate with gender, socioeconomic background or other characteristics. Consequently, an algorithm can potentially produce unequal outcomes without explicitly considering a protected characteristic.

Gender bias is one area that has attracted considerable attention. Dastin (2018) reported that Amazon abandoned an experimental recruitment algorithm after it was found to disadvantage women applicants because the system had learned patterns from historical recruitment data that reflected male dominance in the technology workforce. Although the specific system was not necessarily representative of all algorithmic recruitment technologies, the case demonstrated how historical data can influence automated selection outcomes.

Algorithmic recruitment may therefore produce what can be described as a black-box problem. Candidates may know that an automated system is involved in the selection process but may have limited knowledge of how it works. This lack of transparency can affect perceptions of procedural justice. Applicants may consider a selection process less fair when they believe that important decisions are being made by a system they cannot understand or challenge.

At the same time, algorithmic recruitment may also be perceived as fairer than human recruitment under certain conditions. Human recruiters are susceptible to cognitive biases, stereotypes, mood effects, similarity bias and other subjective influences. An algorithm that applies consistent criteria to all candidates may potentially reduce some forms of human inconsistency. Köchling and Wehner (2020) note that algorithmic decision-making in human resource management can create opportunities for more consistent decision-making while also generating new forms of discrimination and bias.

This creates an important paradox. The use of algorithms in recruitment may be introduced partly to make selection more efficient, consistent and objective, yet applicants may perceive automated recruitment as unfair if the process is opaque, impersonal or discriminatory. Thus, the mere presence of an algorithm does not determine whether applicants will perceive a selection process as fair. Perceptions are likely to depend on factors such as transparency, consistency, explainability, opportunities for human review and confidence in the organisation.

The concept of procedural justice provides an important theoretical basis for understanding perceived fairness in recruitment. Leventhal (1980) proposed principles of procedural justice relating to consistency, bias suppression, accuracy, correctability, representativeness and ethicality. These principles are particularly relevant to algorithmic recruitment because applicants may evaluate whether automated procedures are consistently applied, based on accurate information and capable of being corrected when errors occur.

Gilliland’s (1993) model is also important because it highlights the role of selection procedures in shaping applicants’ fairness perceptions. Selection systems can influence applicants’ reactions depending on whether procedures are job-related, consistently administered, provide opportunities to perform, offer explanations and treat candidates with respect. Algorithmic recruitment therefore needs to be considered not merely as a technological tool but as a selection procedure experienced by applicants.

Another important issue is the extent to which applicants trust organisations that use algorithmic recruitment. Trust is closely connected to perceptions of fairness and transparency. If candidates believe that an organisation uses automated systems responsibly, they may accept algorithmic decision-making more readily. Conversely, concerns about surveillance, data collection, privacy or discrimination may reduce trust.

Data privacy represents another major concern. Algorithmic recruitment systems require applicant information to function. Recruitment firms may collect CVs, employment histories, educational qualifications, assessment scores, demographic information and other personal data. Some emerging recruitment technologies may also analyse video interviews, speech, online profiles or other behavioural indicators. The more data a system processes, the greater the importance of responsible data governance and transparency.

The issue of data protection is particularly relevant in Nigeria following the enactment of the Nigeria Data Protection Act 2023. The Act establishes a legal framework for the protection of personal data and imposes obligations relating to the processing of personal information. Algorithmic recruitment firms operating in Nigeria therefore need to consider not only the effectiveness of automated selection technologies but also the rights and expectations of applicants concerning their personal information (Nigeria Data Protection Commission, 2023).

The Nigerian recruitment industry is increasingly influenced by digital transformation. Recruitment firms use online job platforms, applicant-tracking systems, digital assessments, social media recruitment and automated candidate-management systems to improve the efficiency of recruitment processes. Lagos State is particularly relevant because it is Nigeria’s major commercial centre and hosts a large concentration of businesses, recruitment agencies, technology companies and professional-service organisations.

Recruitment firms in Lagos frequently serve clients across sectors such as banking, telecommunications, consulting, technology, manufacturing and professional services. The volume and diversity of recruitment activities create incentives for the use of technology in sourcing and screening candidates. Algorithmic systems can assist recruitment firms in processing large applicant pools and identifying candidates who appear to meet client requirements.

However, the growing use of automated recruitment technologies may create uncertainty concerning how candidates perceive fairness. Applicants may be comfortable with digital application systems but may become concerned when they realise that automated systems are making or influencing decisions about their suitability for employment. This distinction is important because technological convenience does not necessarily imply perceived procedural fairness.

There is also a contextual issue concerning the transfer of findings from developed countries to Nigeria. Much of the existing research on algorithmic hiring has focused on North American and European contexts. The regulatory environment, labour-market conditions, technological infrastructure and applicant experiences in Nigeria may differ considerably. Consequently, there is a need for empirical research examining algorithmic recruitment from the perspective of Nigerian job applicants and recruitment professionals.

Furthermore, the use of algorithms does not necessarily mean that human involvement has disappeared. Recruitment may involve a combination of automated screening and human decision-making. An algorithm may initially rank candidates while recruiters make the final selection. This creates questions about how applicants perceive the interaction between automated recommendations and human judgement. A process may be viewed as more legitimate if candidates know that humans can review or challenge automated decisions.

The question of human oversight is therefore central to algorithmic fairness. If an automated system makes an error, applicants need mechanisms through which the decision can be reviewed or corrected. Without such mechanisms, candidates may perceive the recruitment process as rigid and unfair. This is consistent with Leventhal’s (1980) principle of correctability, which suggests that fair procedures should provide opportunities for decisions to be modified when errors occur.

Another important issue is explainability. Candidates may want to know why they were rejected or shortlisted. An explanation can increase perceptions of procedural justice by helping applicants understand how decisions were reached. However, some machine-learning systems can be difficult to explain because their decisions emerge from complex relationships among numerous variables. Burrell (2016) argues that opacity can arise from technical complexity, secrecy and the inability of users to interpret sophisticated computational systems.

Algorithmic recruitment also raises concerns about the future of human resource professionals. Recruiters may increasingly become responsible for supervising algorithmic systems rather than making every selection decision manually. This requires recruiters to understand not only recruitment principles but also the limitations, biases and data requirements of algorithmic tools. Raghavan et al. (2020) emphasise the importance of examining how hiring algorithms are actually developed, validated and used rather than assuming that technological systems are inherently neutral.

The importance of perceived fairness extends beyond applicants who are rejected. Even successful applicants may form opinions about an organisation based on how they experienced the recruitment process. If candidates believe that an organisation treats applicants fairly, they may develop more favourable attitudes toward the organisation. Hausknecht et al. (2004) show that applicant reactions to selection procedures can influence organisational attractiveness and related outcomes.

Conversely, perceived unfairness can damage an organisation’s reputation. In an environment where applicants share recruitment experiences through social media and professional networks, negative perceptions can spread beyond individual candidates. Recruitment firms therefore have incentives to ensure that algorithmic recruitment systems are not only efficient but also perceived as legitimate and fair.

The issue is particularly significant for recruitment firms because they often act as intermediaries between applicants and client organisations. A candidate’s experience with the recruitment firm may influence perceptions of both the recruitment agency and the organisation offering the job. Consequently, recruitment firms have an important responsibility to ensure that technologies used in selection processes are transparent, valid, reliable and consistent with principles of fairness.

The literature therefore presents a complex picture. Algorithmic recruitment can potentially increase efficiency, consistency and scalability, but it can also create concerns about bias, opacity, explainability, privacy and accountability (Köchling & Wehner, 2020; Raghavan et al., 2020). The consequences of these issues may depend significantly on how applicants perceive the fairness of the selection process.

Against this background, this study examines Algorithmic Recruitment and Perceived Fairness in Employee Selection: A Study of Selected Recruitment Firms in Lagos State. The study focuses on the relationship between the use of algorithmic recruitment and candidates’ perceptions of fairness in employee selection. It is expected to provide empirical evidence concerning whether algorithmic recruitment practices are perceived as fair and the implications of such perceptions for recruitment processes in Lagos State.

1.2 Statement of the Problem

Employee recruitment and selection are critical organisational activities because they determine who gains access to employment opportunities. Traditionally, recruitment decisions have been made primarily by human recruiters using CV reviews, interviews, tests, references and professional judgement. However, the increasing volume of applications and the development of artificial intelligence have encouraged recruitment firms to adopt automated technologies for sourcing, screening, ranking and assessing candidates. Although these technologies may improve speed and efficiency, they create concerns about whether employee selection remains fair from the applicant’s perspective.

The first major problem is algorithmic bias. Algorithms learn from data, and the data used to develop or train recruitment systems may reflect existing social and organisational inequalities. Barocas and Selbst (2016) demonstrate that seemingly neutral data-driven systems can produce discriminatory outcomes even where protected characteristics are not directly included. In recruitment, this means that an automated system may unintentionally favour some categories of candidates over others.

The widely reported Amazon recruitment case illustrates the potential danger. The experimental system reportedly learned from historical recruitment data and developed a preference for patterns associated with male applicants, ultimately leading Amazon to abandon the tool (Dastin, 2018). Although this case does not establish that all recruitment algorithms are biased, it demonstrates the potential for historical data to reproduce existing inequalities.

A second problem is lack of transparency. Applicants may know that an automated system is being used but may not understand the criteria or processes through which candidates are evaluated. Burrell (2016) identifies opacity as a major challenge in machine-learning systems. Where candidates cannot understand how decisions are made, they may perceive the selection process as unfair, particularly when the outcome negatively affects their employment opportunities.

A third problem concerns explainability and accountability. If an applicant is rejected by an algorithm, it may be difficult to determine who is responsible for the decision—the recruitment firm, software provider, recruiter or algorithm itself. Raghavan et al. (2020) highlight the importance of examining the practical realities of algorithmic hiring, including how systems are validated and how organisations address concerns about bias and accountability.

A fourth problem is the potential reduction of human judgement and contextual consideration. Algorithms may evaluate candidates according to predetermined variables and may not fully appreciate circumstances that are difficult to represent numerically. For example, a candidate’s career gap may have a legitimate explanation that an automated screening system does not recognise. Similarly, transferable skills or unconventional career paths may be overlooked if an algorithm relies heavily on historical patterns or keyword matching.

Another problem relates to applicant trust. Candidates may be less willing to trust recruitment decisions when they believe that a computer has made an important decision about their employment without meaningful human involvement. Gilliland (1993) demonstrates that applicant perceptions of procedural justice are influenced by characteristics of selection procedures. If algorithmic recruitment procedures are viewed as impersonal, opaque or lacking opportunities for communication, applicants may develop negative perceptions of the recruitment organisation.

There is also a problem concerning privacy and personal-data processing. Algorithmic recruitment systems require access to applicant information. Recruitment firms may process CVs, educational records, employment histories, assessment results and other personal information. The collection and processing of such data raise questions concerning consent, purpose limitation, transparency and data security. In Nigeria, these issues are particularly relevant under the Nigeria Data Protection Act 2023 (Nigeria Data Protection Commission, 2023).

Another major problem is that technological efficiency may be mistaken for fairness. An algorithm can process applications faster than a human recruiter, but speed does not necessarily mean that the process is equitable. A system can apply the same flawed criteria consistently to every applicant and therefore be consistent without being substantively fair. Köchling and Wehner (2020) consequently caution that algorithmic decision-making in human resource management can simultaneously offer efficiency benefits and create new forms of discrimination.

There is also a potential research gap in the Nigerian context. Existing scholarship has examined algorithmic decision-making, automated hiring and fairness largely within Western and technologically advanced contexts. While studies such as Köchling and Wehner (2020) and Raghavan et al. (2020) provide important insights, there remains limited empirical evidence concerning how applicants and recruitment stakeholders in Nigeria perceive algorithmic employee selection.

The Lagos recruitment environment requires particular attention because the city serves as a major employment and business centre. Recruitment firms in Lagos operate in a labour market characterised by a large pool of job seekers, intense competition for employment and growing adoption of digital technologies. The use of automated recruitment systems within such an environment may have substantial implications for applicants’ access to employment opportunities.

The problem is further complicated by the possibility that applicants may not know whether an algorithm has been used in evaluating their applications. Where recruitment firms do not clearly communicate the role of automated systems, candidates may be unable to assess the fairness of the process or determine whether they have avenues for review. This lack of transparency can potentially undermine confidence in recruitment decisions.

The consequences of perceived unfairness can extend beyond individual applicants. Hausknecht et al. (2004) demonstrate that applicant reactions to selection procedures can influence organisational attractiveness and other outcomes. Therefore, if applicants perceive algorithmic recruitment as unfair, recruitment firms may experience reputational consequences, reduced applicant trust and difficulties attracting high-quality candidates.

Despite these concerns, algorithmic recruitment is likely to continue expanding because of its potential to reduce recruitment costs, improve processing speed and manage large applicant populations. The appropriate response is therefore not necessarily to reject technology but to understand how it can be used in ways that applicants perceive as fair, transparent and accountable.

The central problem addressed by this study is therefore the limited empirical understanding of the relationship between algorithmic recruitment and perceived fairness in employee selection among selected recruitment firms in Lagos State. Specifically, it remains unclear whether applicants perceive algorithm-assisted selection as fair, transparent and unbiased and whether the use of such technologies affects their confidence in recruitment decisions.

This gap makes the study necessary. Examining perceived fairness can provide recruitment firms with evidence about how candidates experience algorithmic selection and whether improvements are required in transparency, human oversight, explanation and data governance.

1.3 Aim of the Study

The main aim of this study is to examine the relationship between algorithmic recruitment and perceived fairness in employee selection among selected recruitment firms in Lagos State.

1.4 Objectives of the Study

The specific objectives are to:

1. examine the extent to which selected recruitment firms in Lagos State use algorithmic technologies in employee recruitment and selection;
2. assess applicants’ perceptions of fairness in algorithm-assisted employee selection;
3. examine the influence of algorithmic transparency on perceived fairness in employee selection;
4. assess the relationship between algorithmic recruitment and applicants’ trust in employee-selection decisions; and
5. determine whether algorithmic recruitment is significantly related to perceived fairness in employee selection.

1.5 Research Questions

The study will answer the following research questions:

1. To what extent do selected recruitment firms in Lagos State use algorithmic technologies in employee recruitment and selection?
2. How do applicants perceive the fairness of algorithm-assisted employee selection?
3. To what extent does algorithmic transparency influence perceived fairness in employee selection?
4. What relationship exists between algorithmic recruitment and applicants’ trust in employee-selection decisions?
5. Is there a significant relationship between algorithmic recruitment and perceived fairness in employee selection?

1.6 Research Hypothesis

The following null hypothesis will be tested at 0.05 level of significance:

H₀: There is no significant relationship between algorithmic recruitment and perceived fairness in employee selection among selected recruitment firms in Lagos State.

1.7 Significance of the Study

The study will be significant to recruitment firms in Lagos State because it will provide evidence concerning applicants’ perceptions of algorithm-assisted recruitment. The findings may assist recruitment firms in determining whether their automated recruitment practices are viewed as fair, transparent and trustworthy.

The study will benefit human resource and recruitment professionals by providing insights into the advantages and potential limitations of algorithmic selection. It may encourage recruiters to combine technological tools with appropriate human judgement and oversight.

The study will be particularly useful to job applicants. The findings may increase awareness of how algorithmic recruitment works and draw attention to issues such as transparency, bias, data privacy and opportunities for human review.

The study will also be relevant to technology developers and providers of recruitment software. Evidence concerning applicant perceptions may encourage the development of systems that incorporate explainability, fairness monitoring, transparency and mechanisms for correcting erroneous decisions.

The study will benefit policymakers and data-protection authorities, particularly in relation to the responsible processing of applicants’ personal information. Nigeria’s data-protection framework creates important obligations for organisations processing personal data, making responsible algorithmic recruitment an increasingly relevant policy issue (Nigeria Data Protection Commission, 2023).

The study will also contribute to academic knowledge by extending research on algorithmic decision-making and procedural fairness into the Nigerian recruitment context. Much existing literature originates from Western settings; therefore, evidence from Lagos State can provide a useful contextual contribution.

Finally, the study will serve as a reference for future researchers interested in artificial intelligence, algorithmic decision-making, human resource management, recruitment, employee selection, procedural justice, workplace technology and data protection.

1.8 Scope of the Study

The study focuses on Algorithmic Recruitment and Perceived Fairness in Employee Selection among Selected Recruitment Firms in Lagos State.

Conceptually, the study covers algorithmic recruitment, automated candidate screening, algorithmic transparency, algorithmic bias, human oversight and perceived fairness in employee selection.

Geographically, the study is restricted to selected recruitment firms operating in Lagos State, Nigeria.

The study focuses primarily on recruitment and selection activities rather than broader applications of artificial intelligence in human resource management such as employee performance appraisal, payroll management or workforce scheduling.

1.9 Operational Definition of Terms

Algorithmic Recruitment: The use of computer-based algorithms, automated systems or artificial intelligence to support or perform recruitment and employee-selection activities.

Algorithmic Selection: The use of computational models to evaluate, rank, filter or recommend job applicants for employment.

Artificial Intelligence (AI): The use of computer systems capable of performing tasks that ordinarily require human cognitive abilities, including learning, pattern recognition, prediction and decision-making.

Applicant: An individual who submits an application or otherwise seeks consideration for a job position.

Perceived Fairness: An applicant’s subjective assessment of whether a recruitment or selection procedure is just, equitable, unbiased and appropriately administered.

Algorithmic Bias: Systematic and potentially unfair differences in the outcomes produced by an algorithm for different groups of applicants.

Algorithmic Transparency: The extent to which information about the operation, criteria and use of an algorithmic recruitment system is made understandable and accessible to relevant stakeholders.

Explainability: The extent to which users can understand the reasons or factors underlying an algorithmic decision or recommendation.

Human Oversight: The involvement of human recruiters or managers in monitoring, reviewing, interpreting or correcting algorithmic recruitment decisions.

Employee Selection: The process through which organisations evaluate applicants and determine which individuals should be offered employment.

Recruitment Firm: An organisation that provides recruitment, staffing, talent-acquisition or employee-selection services to employers.

Procedural Justice: The perceived fairness of the procedures through which decisions are made, including consistency, accuracy, bias suppression and opportunities for correction.

Applicant Trust: The extent to which applicants believe that a recruitment organisation and its selection procedures are reliable, transparent and fair.

Data Privacy: The protection of applicants’ personal information against inappropriate collection, use, disclosure or processing.

1.10 Organisation of the Study

The study is organised into five chapters. Chapter One presents the introduction, background to the study, statement of the problem, aim and objectives, research questions, hypothesis, significance, scope and operational definitions of terms. Chapter Two reviews relevant conceptual, theoretical and empirical literature on algorithmic recruitment and perceived fairness in employee selection. Chapter Three presents the research methodology, including the research design, population, sample size, sampling technique, research instrument, validity, reliability, data-collection procedure and method of data analysis. Chapter Four presents, analyses and interprets the data collected from respondents and tests the stated hypothesis. Chapter Five presents the summary of findings, conclusion and recommendations.

Project – Algorithmic Recruitment and Perceived Fairness in Employee Selection: A Study of Selected Recruitment Firms in Lagos State