AI-Powered Recruitment and the Attitudes of German University Graduates

Table of Contents

List of abbreviations

List of tables

1 Introduction

2 Literature review
2.1 Recruitment Process
2.2 AI-Powered Recruitment
2.3 Candidate Reactions towards AI-Powered Recruitment

3 Methodology

4 Data analysis

5 Conclusion

List of references

Appendix

Abstract

Businesses are becoming increasingly reliant on technology as globalization progresses. New skills and knowledge are vital for organizations to be competitive, making Human Resource Management (HRM) more important now than ever. The application of advanced technology like Artificial Intelligence (AI) has gained significant attention in the recruitment process in recent years. AI offers promising solutions for recruiters to increase cost and efficiency by taking over repetitive tasks, such as screening CVs and conducting job interviews. However, little is known so far about how these solutions affect potential applicants’ attitudes, particularly the attitudes of graduate students from German universities, and whether they are willing to accept and make use of the new opportunities offered by AI. Therefore, the purpose of this thesis is to investigate German graduate students’ attitudes towards AI-powered recruitment. Toward this end, data were collected through a questionnaire that was e-mailed to German university (graduate) students. The results indicate that the type of AI is a crucial factor. Furthermore, factors such as attitude towards technology in general, attitude towards AI, potential demonstration, impartiality, feedback, and concern are significantly related to the participants’ attitudes. Amongst these factors, impartiality appears to be the strongest predictor. It can be concluded that employers must pay considerable attention to their AI-powered recruitment tools if they wish to attract the right job candidates in the future.

Keywords: Recruitment process, AI-powered recruitment, Candidate reactions towards AI-powered recruitment

List of abbreviations

AES Applicant Expectation Survey

AI Artificial Intelligence

HR Human Resources

HRM Human Resource Management

IBM International Business Machines Corporation

SPSS Statistical Package for Social Sciences

List of tables

Table 1. Questionnaire variables, number of items and scales

Table 2. Descriptive statistics – Sample demographic information

Table 3. Reliability Statistics – Cronbach’s Alpha

Table 4. Reliability Statistics – Cronbach’s Alpha - Feedback

Table 5. Ranking: AI application examples in recruiting

Table 6. Mann-Whitney U test: Medians and significance levels

1 Introduction

Progressive globalization, increasing demographic changes, and continuous technological innovation are just a few examples of factors that are changing today’s world of work (Cascio, 2003). Especially due to technological progress, competition no longer takes place only locally, but organizations must compete on a global level as those technologies are making the world smaller (Erixon, 2018). Hence, to remain viable in this business environment, there is an increasing demand for skills, knowledge, and abilities that organizations particularly need from future employees. These developments highlight the importance of being able to attract and recruit the most qualified applicants available on the labour market. Therefore, the ability to understand how potential applicants react to an organization’s recruitment process could be key to an organization’s success.

Traditionally, it has been the responsibility of human recruiters to source potential candidates, review applications, and eventually hire new employees. Due to technological developments, such as the application of Artificial Intelligence (AI), the recruitment process has been augmented by algorithms, which can take over tasks that were previously carried out by human recruiters. As a result, the inclusion of human recruiters in the recruiting and selection process is gradually decreasing (Bondarouk & Brewster, 2016).

The purpose of this research is to investigate German graduate students’ attitudes towards AI-powered recruitment. Graduate students are assumed to be the demographic most likely affected by the technology that utilizes AI because they are about to pursue a professional career and to enter the labour market. A self-constructed questionnaire was used for empirical data collection and distributed to both individuals who had recently graduated and individuals who would soon graduate from a German university-level institution (both groups are considered to be ‘graduate students’ in the context of this study). This was done to quantitatively measure which form of AI in recruitment is the most favourable amongst applicants and which psychological factors must be taken into consideration by organizations when applying such technology. Afterwards, an analysis will be conducted which will seek to investigate whether there are any noticeable peculiarities amongst participants who favour and participants who disfavour the application of AI.

Finally, conclusions will be drawn and implications will be derived on how organizations could productively adapt their selection processes as they consider introducing AI technology for selecting German graduate students. Finally, suggestions for further research will be provided. Overall, it is intended that readers of this study will be able to make well justified judgements on the viability of applying AI within the recruiting and selection process.

2 Literature review

The purpose of the following section is to demonstrate the importance of the recruitment process in respect to organizational goal attainment. Various authors’ views, both positive and negative, as well as practical examples of AI and its usefulness within the recruitment process, are demonstrated. Finally, previous research conducted on applicants’ reactions towards AI-powered recruitment will be surveyed.

2.1 Recruitment Process

Recruitment is an integral part of Human Resource Management (HRM) (Wall & Wood, 2005). Although there is no consensus regarding the definition of the term ‘recruitment’ (e.g., Breaugh et al., 2008; Taylor & Bergmann, 1987), the definition provided by Barber (1998) is the most appropriate one within the context of this study. Recruitment comprises the generation of applicants, the maintenance of applicant status, and the ability to influence applicants’ job choices (Barber, 1998).

Previous research has concluded that the way recruitment activities are performed has a significant impact on applicants’ attitudes and that the resulting effects can be both positive and negative (Rynes, Heneman, & Schwab, 1980). One study by Schreurs et al. (2008) designed the Applicant Expectation Survey (AES), a general framework for measuring applicants’ reactions towards selection procedures within Human Resources (HR). The average age of participants in their study was 21-22 (Schreurs et al., 2008). According to the German Federal Statistical Office, the average age of German university-level students was 23.4 years during the winter term of 2018/2019 (Destatis, 2020), which indicates the (partial) applicability of the AES in the present thesis. Schreurs et al. (2008) found that five variables (respect, chance to demonstrate potential, difficulty of faking, unbiased assessment, and feedback) were positively correlated with applicants’ attitudes towards selection procedures. This indicates that applicants’ perceptions of an organization, good or bad, can lead to either an acceptance or a rejection of a job offer, and thus impact the probability of an organization recruiting top candidates. Employee selection is commonly conducted by traditional methods, such as interviewing applicants. Hence, one of the most important aspects to consider when selecting an applicant from a pool of candidates is that every applicant should have an equal opportunity to be selected (Stoilkovska et al. 2015). This may be one of the reasons why organizations have gradually transitioned to more non-traditional methods (Elearn, 2009).

2.2 AI-Powered Recruitment

Technological innovation has remodelled recruitment and application practices (Derous & de Fruyt, 2016). The most recent disruption in HRM has been the introduction of self-learning computing systems that are able to influence or even take over human decision-making (Herbst et al., 2017). Such systems gain more attention every passing year and can be classified as AI-powered technologies. John McCarthy, who is credited for creating the term ‘Artificial Intelligence’, defined it as the science of making intelligent machines. Moreover, he has described it as an autonomous computer-enabled system with the ability to process information and provide outcomes like human beings by means of learning and making decisions (McCarthy, 1998). Consequently, AI is present if, by asking the same question to both a person and an AI, one is not able to determine any difference between the quality of the answers. A more recent definition of AI describes it as an intelligent agent that can take a certain course of action to increase the success of reaching predetermined objectives (Oana, Cosmin, & Valentin, 2017).

Some might question the demand for new recruitment technologies given that there are already established technologies such as E-Recruiting (Sylva & Mol, 2009). However, recent literature states that AI-powered technology is superior to humans in terms of both efficiency and effectiveness, especially in the initial stages of the recruitment process (Kaplan & Haenlein, 2019). This does not mean that AI replaces the human factor in the recruitment process, but rather takes over routine tasks that were previously conducted by human recruiters (Upadhyay & Khandelwal, 2018). Such tasks include the scanning of applicants’ CVs, answering applicants’ questions, and conducting interviews (O’Donovan, 2019).

There are various examples of AI-powered systems in recruiting that already exist today. The spectrum ranges from tools that help to respond to questions from potential applicants to tools that support the pre- and final selection of applicants. Furthermore, different forms of AI can be distinguished, such as Augmented Intelligence and Autonomous Intelligence. These forms can be characterized by several basic functions, including the ability to analyse and interpret large amounts of unstructured data and thereby identify differences and patterns, provide knowledge and findings, and make predictions about the future (Castro & New, 2016).

One existing Augmented Intelligence application is IBM’s chatbot, Watson . Chatbots allow for quicker reactions to applications and applicants’ requests compared to if they were processed by a human recruiter (Davenport & Ronanki, 2018). This can reduce recruitment costs (Dickson & Nusair, 2010) and decrease the number of frustrated applicants that drop out of the application process due to a long waiting period (Dahm & Dregger, 2018). Furthermore, IBM’s Watson can support recruiters by providing them with information, which otherwise could not have been obtained if there is an unusually high number of applicants. The algorithm analyses the data of applicants and compares them with previously created historic business data. If, for instance, the analysis concludes that successful HR managers require a high level of social competence, the algorithm focuses on applicants with exactly this skill set. Eventually, the system selects potential candidates by creating individual scores for the target performance indicator (e.g., social competence) based on the available documents (IBM, n.d.). This implies that the system analyses all incoming applicants’ data with the same level of attention, which could lead to an increase in equity (Dahm & Dregger, 2018). Furthermore, AI systems can be taught to bypass the analysis of candidates’ names, gender, and age, which are the primary sources of recruiter bias (Upadhyay & Khandelwal, 2018).

In contrast, Autonomous Intelligence is characterized as the highest level of AI since it operates mostly independently, and human monitoring is only occasionally required. Companies like HireVue provide solutions, such as AI-powered pre-assessments, video interviewing, and interview scheduling. HireVue markets itself as an assessment and video interview software developer that aims at assisting organizations in making better predictions, decisions, and hires (HireVue, 2019). HireVue intends to combine interviews with predictive analytics to speed up decision-making while eliminating unconscious human bias. This is supported by the claim that perfect rational decision-making is not possible for human beings (Simon, 1968). HireVue compares the applicants’ characteristics with the ones of high-performing employees of an organization by analysing keywords, facial expressions, and voice patterns. Eventually, the system selects and suggests the best applicants to human recruiters (HireVue, 2019). According to a case study available on HireVue’s website, the global hotel chain Hilton Worldwide Holdings managed to reduce the length of their recruitment process from 42 days to only 5 days due to the use of AI-powered video interviews (HireVue, 2017).

Although IBM and HireVue already provide marketable AI solutions, some practical examples shed light on the downsides of AI-powered recruitment technology. For example, algorithms are usually based on decisions and results from the past. Since those past decisions and results usually originate from humans, there remains a potential risk that the algorithms learn and apply human prejudice (Langer, Baum & König, 2018). This is supported by Scherer (2017) who claims that AI in recruitment is prone to be biased as the data put into the system will be subjective. This hypothetical scenario became real for the e-commerce giant Amazon in 2018 when the algorithm started to replicate human hiring bias. Amazon developed their own AI-powered selection tool for filling software developer positions, which analysed each applicant’s CV and determined an individual score. The data base consisted of applications from the past ten years. As most software developers were male, the algorithm considered the attribute “gender” as important and, consequently, automatically favoured men over women. As previously mentioned, AI can be programmed in a way to bypass the gender attribute. In fact, gender was not listed as an evaluative criterion in Amazon’s algorithm. Nonetheless, it downgraded applications including the word ‘women’s’ (e.g., ‘women’s college’) and ultimately disadvantaged female applicants. Eventually, the built-in bias was discovered and the project was taken down (Dastin, 2018).

2.3 Candidate Reactions towards AI-Powered Recruitment

AI-powered recruitment is a recent innovation in the personnel selection process (Mejia & Torres, 2018). When reviewing the current state of research, it has been found that the investigation of applicants’ reactions towards AI-powered recruitment is lagging behind the current practice of it (as previously outlined). One reason for this might be that the importance of applicants’ selection expectations has only relatively recently begun to attract the attention of scholars (Bell et al., 2006). However, Barber (1998) has already pointed out the relevance of looking at the recruitment process from an applicant’s perspective to understand organizational recruitment holistically. In general, applicants’ expectations of the selection process can be defined as their beliefs about the characteristics of the forthcoming selection procedure (Schreurs et al., 2008). Furthermore, applicants’ perceptions can affect their attraction to an organization (Schinkel et al., 2013). Dietvorst et al. (2016) found that people might respond differently to decisions made by algorithms as opposed to decisions made by people. If applicants feel that AI invades their privacy, which could result in a violation of data protection laws (Scherer, 2017), these negative feelings might affect their overall motivation to apply for jobs (Van Esch et al., 2019). Nevertheless, the latter might also be true for a data protection law violation committed by a human recruiter.

The literature review revealed two previously conducted studies, which account for much of the attention given to the present research topic. Both studies were conducted in Germany and dealt with the topic of AI-powered recruitment and its effect on potential applicants. One empirical study conducted by Dahm & Dregger (2018) addressed questions regarding the attitudes young applicants have towards AI-powered recruitment. It investigated the roles of applicants’ psychological variables, such as trust in the system, appreciation, and joy, as well as the type of artificial intelligence used (Augmented Intelligence or Autonomous Intelligence). The target of the study, however, was to find out how applicants’ acceptance of AI-powered tools can be promoted. A survey was conducted, which comprised 238 young professionals who assessed various AI-powered recruiting tools using a questionnaire. The study found that AI-powered recruitment was perceived differently by the participants. In particular, the type of AI and psychological variables have a significant influence on participants’ acceptance of it. For instance, if confronted with the situation of Autonomous Intelligence steering the recruitment process by itself, 76.00% of the participants would prefer a human recruiter. Whereas, if confronted with a situation where Augmented Intelligence was used, only 41.10% of the participants claimed to prefer a human recruiter. When assessing potential measures that might positively influence participants’ acceptance, 70.60% of participants endorsed the availability of a human recruiter when being assessed by an AI. For the present study, Dahm’s and Dregger’s (2018) research indicates that even with a young target group, the introduction of AI in personnel recruitment is not easy.

In 2018, another study about AI-powered recruitment and its effect on potential applicants was published by Viasto (2018), a German software developer. Like in the study done by Dahm and Dregger (2018), Viasto focused on how applicants perceive AI-powered recruitment and what companies can do to influence applicants’ attitudes towards it. Data were collected from 1,008 participants aged between 18 and 69 years via an online questionnaire. One of the key findings was that 17.50% of all participants characterized their perspective on the use of AI in recruiting as ‘highly sceptical’. However, scepticism increases the higher the age of the participants. While only 10.00%-11.00% of 18-29-year-olds express themselves as ‘highly sceptical’, it is 31.60% for those 60 years of age and older. This indicates that relatively high agreement with AI-powered recruitment may be expected from German university (graduate) students in the present study. Another key finding was that participants’ attitudes depend on the type of AI used, as mentioned previously. Participants spoke out in favour of technology that decreases discrimination against applicants and increases objectivity. The majority (59.50%) feel that the best solution is for both AI and human recruiters to be involved in the recruitment process, with the human recruiter making the final decision (Viasto, 2018).

Based on the preceding literature review, it is evident that AI-powered solutions are an emerging trend within the field of recruitment and selection. Various examples provide insights into how AI is already being used by organisations and demonstrate its future potential. However, its future relevance might be significantly determined by job seekers’ attitudes, as indicated by previous research. This aspect will be further investigated in this thesis, using an exploratory approach.

3 Methodology

A survey was conducted to investigate German graduate students’ attitudes towards AI-powered recruitment.

The present study intends to gain insights into the attitudes of German graduate students towards AI-powered recruitment and to exploratively investigate potential differences and/ or similarities amongst those students. There are two reasons for using a survey for this thesis instead of another research methodology, such as a case study. First, a survey allows for the gathering of a large amount of information, which can be used for statistical inference or reveal unsuspected results (Swetnam, 2000). Second, this research is being carried out during the early period of the global COVID-19 pandemic, which comes along with restrictions, such as contact bans and curfews. Although a case study “could be a study of a particular practice, for example recruitment and selection processes” (Quinlan, 2011, p. 182), it is usually attached to one locality.

The study population consists of German graduate students. In the context of this study, a graduate student is any German student who has either already obtained an academic degree (bachelor’s/master’s) or is currently working towards obtaining an academic degree from one of the 424 university-level institutions in Germany (Destatis, 2020). The latter qualification is important because German bachelor’s programs usually last between 6-8 semesters (3-4 years) and, therefore, undergraduate students will soon be directly engaged in a permanent employment search. More precisely, the population of this study consists of any (predefined) student who intends to apply for a job and pursue a professional career after earning his or her degree. In the winter term of 2019/2020, according to the German Federal Statistical Office, approximately 2.5 million German students were enrolled in German university-level institutions (Destatis, 2020). However, it cannot be assumed that every German (graduate) student is part of the study population, as some students may not pursue a professional career or get involved in an HR recruitment and selection process after their studies.

The sampling method in this study follows a non-probability sampling approach combining the two sampling techniques of convenience sampling and snowball sampling. In the initial stage of the process, a convenience sampling technique is applied with a focus on engaging potential participants who are the easiest to include. Therefore, members of the researcher’s personal network of German (under-)graduate students at different locations in Germany were approached directly. To exploit the full potential of the researcher’s network, a snowball sampling technique was applied in the second stage of the process. Each initial-stage participant was asked to introduce and engage another potential participant in this study. This enables the researcher to potentially reach a more diverse and geographically scattered sample (Saunders et al., 2016). Furthermore, participation was incentivised by a (voluntary) post-participation lottery which included three, €50-gift cards.

Data was collected during a 13-day period with the help of an online questionnaire (Appendix A.1), conducted through the online survey tool surveyhero. This seemed to be the most appropriate method to collect data from a geographically scattered population and to reach an adequate number of participants (Fowler, 2009). The original questionnaire was conducted in German, and the results were then translated into English. Participation was voluntary as well as anonymous, unless respondents enrolled in the lottery with their personal contact details. The questionnaire was developed based on seven different variables, with each variable containing several items (Table 1).

Abbildung in dieser Leseprobe nicht enthalten

Table 1. Questionnaire variables, number of items and scales

The first section consisted of seven questions about respondent demographics, as well as one question about whether the respondents considered themselves confident in their abilities during interviews. In the second section, the variables and their respective items were modified from the previous studies of Dahm and Dregger (2018) and Viasto (2018), regarding one variable (attitude towards AI in recruiting), and the study of Schreurs et al. (2008), regarding three variables (potential demonstration, impartiality, and feedback). Responses on all the variables were measured on 5-point Likert Scales (ranging from ‘1’= do not agree at all to ‘5’ = totally agree), on dichotomous scales (yes/no), and with open-ended questions. Furthermore, three additional variables (general attitude towards technology, general attitude towards AI, and concern) were created specifically for this study. These three variables are intended to evaluate participants’ general beliefs and basic assumptions that might influence their specific attitudes towards AI-powered recruitment. At the end of the questionnaire, participants were asked an open-ended question regarding their perceptions of the subject in order to provide the opportunity for them to share their personal opinions and feelings.

The first three variables (general attitude towards technology, general attitude towards AI, and attitude towards AI in recruiting) evaluate how applicants perceive new technologies, especially AI-powered technologies both in a general and HR context. With respect to AI being used in recruiting, participants were asked to rank three different scenarios. The three scenarios were automatically displayed in a randomized order for each participant to avoid any potential bias. In each scenario the degree to which an AI takes responsibility for certain tasks is different. The three scenarios are briefly described in the following:

Scenario 1: Applicant communicates with an AI-powered chatbot that can answer applicants’ questions, make appointments, and suggest job offers.

Scenario 2: An AI-powered tool pre-assesses applicants and creates an individual suitability score based on the application documents. A human recruiter makes the final decision, taking the generated scores into consideration.

Scenario 3: An AI-powered tool autonomously carries out an interview. A personality profile based on the applicant’s gestures, facial expressions, posture, tone, and content is created and compared to existing employees of the organization to determine the candidate’s suitability. The AI makes the final decision.

The data gathered were quantitatively analysed with the statistics software package SPSS from IBM. First, a descriptive analysis was carried out to describe the data gathered. The descriptive statistics used in this research include frequencies, percentages, modes, medians, and means of the sample. For instance, in the case of the three scenarios above, frequencies of the candidates’ choices will be used as indicators of which type of AI (scenario) is preferred to the others.

The four variables, potential demonstratio n, impartiality, feedback, and concern will be analysed using the Mann-Whitney U test. For this purpose, the sample will be divided into three independent groups according to participants’ attitudes towards the usefulness of AI in recruiting (question 19; Appendix A.1). As a matter of simplicity, participants who chose either ‘4’ (slightly agree) or ‘5’ (totally agree) on the 5-point Likert Scale are considered ‘supporters’. Participants who chose either ‘2’ (slightly disagree) or ‘1’ (strongly disagree) are considered ‘opponents’. The third group, participants who chose ‘3’, are not considered in the analysis as it represents participants who neither agree nor disagree with AI’s usefulness in recruiting. The Mann-Whitney U test is used to determine if there are differences between the two independent groups of supporters and opponents in respect to each of the four dependent variables - potential demonstration, impartiality, feedback, and concern. The main reason why the Mann-Whitney U test was chosen over other statistical tests that investigate differences between two groups (e.g., the independent-samples t-test) is because the dependent variables were measured on an ordinal scale, and the independent-samples t-test requires a continuous dependent variable.

In terms of reliability, the Cronbach’s Alpha test will be conducted to measure the internal reliability of the variables of g eneral attitude towards technology, potential demonstration, impartiality, feedback, and concern. What constitutes a good level of internal reliability differs depending on which source one refers to. For this purpose, a minimum Cronbach’s Alpha value of 0.600 is taken into consideration as suggested by Hair (1998). This test is generally applied to determine the extent to which the items of one variable measure the same underlying dimension.

4 Data analysis

The online questionnaire gathered data from a total of 456 different respondents. However, 50 respondents dropped out before completion and a further 36 respondents were not in line with the inclusion criteria. These respondents needed to be excluded from the sample, which resulted in a final sample size of n=370 individuals.

Abbildung in dieser Leseprobe nicht enthalten

Table 2. Descriptive statistics – Sample demographic information

As presented in Table 2, the sample consists of more female (54.59%) than male respondents (45.41%). The average age of participants is 23.06 years, which is slightly below the average age of 23.40 years for German university-level institution students (Destatis, 2020). The sample includes participants who are studying or have studied business sciences (56.14%), engineering (27.84%), and other study programmes (7.03%). The mean number of work experience (including internships) is 1.90 years. However, 57.03% of participants claim to have less than 1.50 years of work experience, including 15.41% of participants without any prior experience. Participants were asked to assess their general level of self-confidence during application and interview scenarios, which resulted in a mean score of 3.97 on a 5-point Likert Scale. This indicates that participants slightly agree that they are confident with their abilities during an application process or interview.

Before proceeding with more complex statistical analysis regarding the internal consistency of each of the five variables - general attitude towards technology, potential demonstration, impartiality, feedback, and concern - Cronbach’s Alpha is used to determine the extent to which the items of one variable measure the same underlying dimension.

For the variables general attitude towards AI, as well as attitude towards AI in recruiting, Cronbach’s Alpha is not applicable because both variables include different scales of measurement (Table 1) and, therefore, reliability can only be assumed. The Cronbach’s Alpha values for each of the five variables are presented in the following table:

Abbildung in dieser Leseprobe nicht enthalten

Table 3. Reliability Statistics – Cronbach’s Alpha

The highest Cronbach’s Alpha value was found for potential demonstration (=.879), whereas the lowest value is present for feedback (=.436). Higher values for Cronbach’s Alpha are considered better, which shows a high level of internal consistency for the respective scale. Four of the variables (general attitude towards technology, potential demonstration, impartiality, and concern) received a value above the target value of .600 (Hair, 1998) and even above .700, which indicates a high level of internal consistency (Kline, 2005).

Abbildung in dieser Leseprobe nicht enthalten

Table 4. Reliability Statistics – Cronbach’s Alpha - Feedback

As the value of the variable feedback (.436) is below the minimum target value of 0.600, further investigation was required. The ‘Cronbach’s Alpha if item deleted’-values show how the calculated value of .436 would change if each question were removed from the scale (Table 4). In the case of feedback, removing any question would not lead to the minimum target value of .600. The ‘Corrected item – total correlation’-values represent the Pearson correlation coefficient between each specific item and the sum of all the other items. If the items were measuring the same underlying dimension, these correlation coefficients would be expected to be relatively high. However, according to Cohen (1988) a Pearson coefficient value between 0.1-0.3 is an indication that the respective item might not measure the same construct and, therefore, removal could be taken into consideration. As questions 25d and 25e resulted in a relatively low coefficient value, it was decided to remove them from further analysis. Once removed, the new Cronbach’s Alpha value for the variable feedback increased from 0.436 to 0.639, which is above the minimum target value of .600. Finally, it can be concluded that all variables are reliable.

The first variable of the questionnaire is meant to convey a general impression of participants’ attitudes towards technology. Four different statements were measured on a 5-point Likert Scale regarding participants’ (1) confidence in using new technologies in their daily lives, (2) openness to new technologies, (3) interest in testing new technologies, and (4) perception of whether new technologies make their lives more efficient. The statements received mean scores of 4.28, 4.26, 4.19, and 3.93, respectively. These values suggest that the participants have a positive general attitude towards new technologies.

The second variable was used to gain insights into participants’ general attitudes towards AI. When asked whether they thought AI generally will make tasks easier in the future, a mean score of 4.16 was achieved, which indicates slight agreement with the statement. Of the participants, 150 (40.50%) claimed to have at least once dealt with AI in their lives. Out of these 150 participants, 53.30% felt partially or completely confident during their experience with AI, whereas only 12.30% did not feel confident. Examples of their encounters with AI include chatbots, smart assistants like ‘Amazon Alexa’, and personalized ads on social media.

The third variable investigates participants’ attitudes towards AI-powered recruitment. Participants evaluated the usefulness of AI-powered recruitment with a mean score of 3.14, with the distribution of the data slightly skewed to the left (Appendix B.1). This means that they might either slightly agree or neither agree nor disagree. However, a mean score of 2.80 was reached when assessing the personal benefit of AI in recruiting from an applicant perspective. The data distribution is marginally skewed to the right (Appendix B.2), indicating that participants might slightly disagree with AI being useful for applicants.

[...]