Figure - uploaded by Sladjana Nørskov
Content may be subject to copyright.
Source publication
This research examines the perceived fairness of two types of job interviews: robot-mediated and face-to-face interviews. The robot-mediated interview tests the concept of a fair proxy in the shape of a teleoperated social robot. In Study 1, a mini-public (n=53) revealed four factors that influence fairness perceptions of the robot-mediated intervi...
Contexts in source publication
Context 1
... seen in Table 5, statistically significant effects of time were only observed for the preference for being interviewed by a robot. Here, participants rated their preference for being interviewed by a robot significantly higher at T4 compared to T1 (t(47)=-2.83, ...Context 2
... participants rated their preference for being interviewed by a robot significantly higher at T4 compared to T1 (t(47)=-2.83, p=.009), while there were no statistically significant differences between any of the other time points or on any of the other questionnaire items (questionnaire items 2, 3, or 4; at any time T1, T2, T3, T4; see Table 5). ...Citations
... However, it is worth noting that other studies found no significant difference in the perceived fairness (Suen et al., 2019) as well as trustworthiness (Ötting & Maier, 2018) between AI and human decision-making. Moreover, applicants consider robots to be fairer than humans in decision-making and thus prefer robot-mediated interviews (Nørskov et al., 2022). In fact, applicants' perceptions of and preferences for AI decisionmaking can be influenced by their relative advantages, which we define as the perception that applicants have about their competitiveness and standing compared to other applicants in the pool. ...
Artificial intelligence (AI) has greatly enhanced the effectiveness of human resource management (HRM). Previous studies have primarily focused on the benefits that AI brings to the HRM process. Despite increasing attention to the applicant’s perspective, conflicting findings persist regarding how applicants respond to AI-driven decision-making processes. Drawing upon prospect theory, we conducted a survey in campus recruitment and four online experiments about HRM decisions to determine if applicants hold a lay belief that an AI agent is more consistent than a human in making decisions, and furthermore, if applicants’ relative advantage strongly affects their preference for AI decision-making. Specifically, we found that applicants who are at an advantage prefer AI decision-making more than those at a disadvantage because they desire a highly consistent agent to maintain their existing advantages. However, their preference for AI decreases when decision-making is not transparent. Our findings not only help to clarify when and why applicants like or dislike AI decision makers but also provide advice for HRM departments to better employ AI decision-making systems.
... However, this process may be necessarily subjective and some bias from the researcher may come into play. According to Nørskov et al. (2022), there is nothing wrong with biases stemming from well-meaning criteria in that they can sometimes disadvantageously exclude relevant information while advantageously including other types of studies. For instance, applying filters that narrow down the articles to only those from peer-reviewed journals means that good information might be filtered out alongside the unpublished studies that could be useful in answering the research question (Relevo and Balshem, 2011). ...
This research explore how the implementation of mindfulness and stress management can affect the behaviour and performance of students, in light of the emergent mental health issues affecting learners at school. Based on secondary data, it examines existing and secondary data sources and related research and literature in various educational settings. In the research, it is clearly illustrated how academic demands, peer relations, and personal development issues cause stress and reduce academic outcomes and behaviours among students. Mindfulness programs, defined by promoting the direct experience of the current moment and completion of emotional self-regulation, appear as effective tools for boosting cognition, diminishing behavioural concerns, and optimising quality of life. Research shows that there are modest and meaningful changes in educational outcomes like CGPA scores, test scores and some sort of positive behavioural changes like decrease in misconducts within classroom and cases of student discipline. Additionally, such programs contribute to maintaining high emotional stability, relieved stress levels, and life satisfaction. However, the study notes deficiencies in the evaluation of the nature and extent of variability in program effect and implementation, moderated by such factors as socioeconomic status. Both discuss the timeliness of culturally sensitive treatments and the importance of methodologically sound, controlled, and powered research designs that support causal claims. They also present the issue of ethical issues and discussing how future research needs to incorporate informed consent and other cultural appropriateness in its framework whilst following careful methodological processes and guidelines. Finally, the study points out attention and stress reduction programs as suitable models for promoting personal development therefore urging education stakeholders to incorporate into the curricula for improved student performance.
... The study examined the attributions people make when encountering deceptive behaviour and discussed the practical implications of these early impressions in contexts like hiring decisions, interpersonal relationships, and legal proceedings. In 2020, Nørskov [12] and colleagues investigated the dynamics of human-robot interaction in job interviews, shedding light on the ethical concerns and psychological implications associated with robotic interviewers. Finally, in 2021, Aslan et al. [13] and Song et al. [14] continued to advance the field by combining models for improved accuracy and exploring self-supervised learning for automated personality trait recognition based on facial dynamics, highlighting practical applications in human-computer interaction and affective computing. ...
In the evolving landscape of smart cities, employment strategies have been steering towards a more personalized approach, aiming to enhance job satisfaction and boost economic efficiency. This paper explores an advanced solution by integrating multimodal deep learning to create a hyper-personalized job matching system based on individual personality traits. We employed the First Impressions V2 dataset, a comprehensive collection encompassing various data modalities suitable for extracting personality insights. Among various architectures tested, the fusion of XceptionResNet with BERT emerged as the most promising, delivering unparalleled results. The combined model achieved an accuracy of 92.12%, an R2 score of 54.49%, a mean squared error of 0.0098, and a root mean squared error of 0.0992. These empirical findings demonstrate the effectiveness of the XceptionResNet + BERT in mapping personality traits, paving the way for an innovative, and efficient approach to job matching in urban environments. This work has the potential to revolutionize recruitment strategies in smart cities, ensuring placements that are not only skill-aligned but also personality-congruent, optimizing both individual satisfaction and organizational productivity. A set of theoretical case studies in technology, banking, healthcare, and retail sectors within smart cities illustrate how the model could optimize both individual satisfaction and organizational productivity.
... The research conducted in this area is diverse and encompasses various aspects of human cognition and behavior (Hallion & Ruscio, 2011;Krieger, 1995;Kube et al., 2024). Cognitive biases in job interviews and employee selection, as well as biases in courts, have been studied for decades (Darley & Gross, 1983;Fiarman, 2016;Koriat et al., 1980;Nørskov et al., 2022;Oberai & Anand, 2018;Roehling et al., 1999;Roulin et al., 2023). Many studies have shown that humans think and act in relation to other people in irrational ways. ...
The study of cognitive biases in job interviews has garnered significant attention due to its far-reaching implications for the economy and society. However, little research has focused on the biases exhibited by expert psychologists serving on psychology specialization examination committees. As such, this study has conducted a comprehensive examination of biases within the specialization exam in Israel. One additional objective of the research is to assess the levels of distress experienced by examinees following the examination. Questionnaires were administered to 418 psychologists participating in the clinical and educational psychology specialization exams. The findings unveiled several noteworthy outcomes. Firstly, several biases were identified, including ethnic stereotypes, biases stemming from cognitive load, and more. Secondly, examinees who presented a cognitive-behavioral therapy (CBT) case experienced a higher failure rate. Thirdly, a positive association was found between exam failure and personal distress and this effect was stronger for educational examinees compared to clinical examinees. The most intriguing discovery was that all biases, without exception, occurred among clinical psychologists, whereas educational psychologists displayed no biases. This outcome contrasted with initial expectations. Consequently, the present study aims to expand the existing knowledge about psychological biases and stereotypes by elucidate the reasons behind this discrepancy between the two disciplines while considering the advantages and disadvantages associated with a sense of "expertise" in the realm of adult diagnostics.
... Moreover, candidates often perceive robot-led interviews as less fair (Noble et al., 2021;Nørskov et al., 2020), particularly with lower perceived fairness in job relevance and two-way communication (Muralidhar et al., 2020;Noble et al., 2021). On the other hand, Nørskov et al. (2022) showed the opposite results, with respondents preferring robots to humans; similar results were reported by Min et al. (2018). In conclusion, previous studies on RMJI have provided mixed evidence. ...
Chatbot-mediated job interviews raise questions about applicant perception of justice, which in turn affects organisational attractiveness. This study uses experimental data to assess applicants' perceptions of procedural, interactional, and interpersonal justice, and their relationship to organisational attractiveness in three scenarios: chatbot, human and unspecified interviewer. The research aims to differentiate the fairness perception effect between robots and human interviewers using an identical written job interview preselection procedure. The results show that the perception of justice significantly impacts organisational attractiveness. Organisations that comply with rules of procedural and interactional justice can implement chatbots without compromising organisational attractiveness. Although the results are mixed, we do not find evidence to suggest that revealing information about chatbots endangers the organisation's attractiveness relative to that of the human recruiter. Younger candidates with previous experience in chatbot interviews even preferred a chatbot to a human recruiter.
... Moreover, candidates often perceive robot-led interviews as less fair (Noble et al., 2021;Nørskov et al., 2020), particularly with lower perceived fairness in job relevance and two-way communication (Muralidhar et al., 2020;Noble et al., 2021). On the other hand, Nørskov et al. (2022) showed the opposite results, with respondents preferring robots to humans; similar results were reported by Min et al. (2018). In conclusion, previous studies on RMJI have provided mixed evidence. ...
Chatbot-mediated job interviews, raise questions about applicant
perception of justice, which in turn affects organisational attractiveness. This study uses experimental data to assess applicants’ perceptions of procedural, interactional, and interpersonal justice, and their relationship to organisational attractiveness in three scenarios: chatbot, human and unspecified interviewer. The research aims to differentiate the fairness perception effect between robots and human interviewers using an identical written job interview preselection procedure. The results show that the perception of justice significantly impacts
organisational attractiveness. Organisations that comply with rules of procedural and interactional justice can implement chatbots without compromising organisational attractiveness. Although the results are mixed, we do not find evidence to suggest that revealing information about chatbots endangers the organisation’s attractiveness relative to that of the human recruiter. Younger candidates with previous experience in chatbot interviews even preferred a chatbot to a human recruiter.
... Previous research on AI-AVI has primarily focused on its technical aspects (Su et al., 2021), as well as its social factors on various impacts, such as interviewer ratings (Suen et al., 2019a), interviewee perception (Nørskov et al., 2022), trust (Suen and Hung, 2023), discrimination (Köchling and Wehner, 2020), ethics (Hunkenschroer and Luetge, 2022), and bias detection (Wall and Schellman, 2021), in the interdisciplinary field of technology and social behaviors. However, there is still insufficient understanding of how job candidates present themselves differently across AI interfaces in AI-AVI. ...
... Therefore, it is crucial for both scholars and practitioners to determine whether job candidates exhibit different IMs during AI video interviews since these behaviors can significantly affect hiring decisions and introduce biases (Langer et al., 2019a), thereby altering employers' selection validity (Basch et al., 2020). As a result, studies on human-AI interactions have called for research on job candidates' self-presentation (Jalagat and Aquino, 2022;Treem et al., 2020) and digital personnel selection interviews (Basch et al., 2020;Nørskov et al., 2022;Roulin et al., 2021). ...
... In essence, our rigorous empirical approach not only deepens the comprehension of human-AI interplay but also enriches the discourse on IM strategies within the AI-AVI context, extending the narrative on usercentric AI paradigms (e.g., Kim and Im, 2023;Nørskov et al., 2022;Langer et al., 2019b). ...
Automated video interviews powered by artificial intelligence (AI) are increasingly being adopted by employers to screen job candidates, despite concerns regarding the humanity and transparency of AI. Accordingly, researchers and practitioners advocate overcoming these concerns by refining AI interfaces in terms of tangibility, immediacy, and transparency. However, AI video interviews featuring different interfaces may impact interviewees' tendencies to engage in impression management behaviors (IMs), which can either improve or impair personnel selection validity. This study addressed calls for research to investigate the issues mentioned above by conducting a field study to explore the ways in which AI and AI interfaces affect candidates' IMs in asynchronous video interviews (AVIs). We developed three AVI interfaces and measured real job applicants' self-reported IMs across four experimental treatments. We found that different AI interfaces could increase or decrease candidates' honest IMs and deceptive IMs in different ways. An exploratory analysis also found that candidates' interview anxiety could be mitigated by an AI interface.
... Today's working environment is transforming into a hybrid world where humans and technical agents (e.g., AI, robots) are expected to work collaboratively [9,22,30]. For example, humans and robots can collaborate in personnel selection [14,20,26] or manufacturing processes [2,25]. Moreover, AI support teachers during exams [18], journalists during research [1,4,29] or physicians with diagnostics [27,31]. ...
In a world where humans and technical agents (e.g., robots, AI) work collaboratively, processes of social inclusion and exclusion in human-agent interaction (HAI) gain importance. However, the current focus of social exclusion in HAI is too narrowminded and neglects many forms of social exclusion (e.g., averted eye gazes, microaggressions, hurtful laughter). To change this, the effects of different types of social exclusion will be explored in a series of experiments against the background of William's need-threat model [34]. By doing so, we will test the transferability of the model, build a HAI-specific taxonomy, and derive prevention strategies. We look forward to interdisciplinary discussions about this topic and hope to receive valuable feedback and inspiration for the presented PhD project which has just started a few months ago.
... In trying to resolve this problem, social robots are being tested for their applicability in mediating job interviews to ensure objectivity and increase applicants' fairness perceptions Nørskov et al., 2022). ...
This report investigates current practice at the interface between the social (human) decision-making process and digital technology. It provides an overview of current learning across various practice contexts, through a unified lens which allows us to assimilate the learning and guide further research.
By cross analysing 13 cases, we identified themes about how to promote the:
• Adoption of the integration of the digital technology in decision-making, in terms of enablers and barriers
• Assessment of the integration of the digital technology in decision-making in terms of effectiveness and appropriateness
• Adaptation of the human systems to integrate digital technologies in decision- making, including inward and outward adaptation
... Another question is, why is healthcare identifiable as a standalone topic in our topic model, and one that has become central to the topic network, while areas such as workplace or commerce are not as clearly identifiable in our data? Future research should continue to investigate the negativity that may arise towards the virtual agent and hinder widespread adoption, whether it arises from algorithmic aversion (Mahmud et al., 2022) or bias concerns (Nørskov et al., 2022;Rabassa et al., 2022). ...
... Furthermore, existing research on social media (Arsenyan and Mirowska, 2021; Mirowska and Arsenyan, 2023), entertainment (Vitek and Peer, 2020), cyber-physical systems (Kumar and Lee, 2022), traffic (Hu et al., 2015), and workplace applications (Nørskov et al., 2022) are absent from our results. We speculate that these disparate bodies of literature are too small to appear as stand-alone topics within our corpus, particularly as our corpus covers almost 30 years of research, while applications of virtual agents in the aforementioned areas are relatively recent phenomena. ...
... This contextual question becomes particularly important when considering the interaction medium. Presenting a virtual versus human agent on video (Lawson et al., 2021), or embedding a virtual agent into an artefact versus operating the artefact by a human agent (Nørskov et al., 2022) would have different implications for the questions of acceptance, persuasion, and general reactions. This "environment", as we have labelled it in our framework, should therefore be part of the research design, whether it be cyber-physical or purely digital. ...
Virtual agent research has evolved into a substantial body of work, albeit one with a fragmented structure and overlapping, and at times inconsistent, definitions and results. The current paper presents a computational literature review of 1865 academic journal publications and conference proceedings from 1995 to 2022 using Latent Dirichlet Allocation to understand the publication trends in the field, its intellectual structure, and how topics within virtual agent research have evolved and relate to each other. Our results point to a model of 16 topics as best representing the current state of the research landscape. We present descriptions of these topics, as well as topic dynamics and networks, in order to provide a clear picture of the current state of the field. We then organise these topics into a Human-Virtual Agent Coexistence Framework, identifying current trends and opportunities for future research.
