Enhancing Inclusivity in Interviewing: Harnessing Intelligent Digital Avatars for Bias Mitigation

Abstract

Interviewing is a critical process in various fields such as human resource management, psychology, and social sciences, serving as a fundamental tool for evaluation, selection, and research. However, this process is inherently susceptible to human bias, which can adversely affect the fairness and validity of outcomes, leading to discrimination and a lack of diversity. Addressing these biases is imperative to ensure fair and equitable practices. In this chapter, we introduce an innovative application designed to address this vulnerability allowing interviewers to practice with digital humans, powered by generative artificial intelligence, and learn how to mitigate such biases. These avatars embody a substantial representation of diversity and are designed with advanced social intelligence capabilities, enabling them to engage in realistic and dynamic interactions, present a range of topics engagingly, and simulate human-like emotional and mood responses. The avatars challenge interviewers to navigate complex, multifaceted interactions, thus honing their abilities to identify and overcome personal biases. Practicing conversations with digital humans accelerates learning from experience without the risks associated with learning in the field. After each interaction, the system provides immediate feedback, fostering self-awareness and performance improvement. The integration of intelligent digital avatars into the interviewing process holds great promise for enhancing inclusivity and reducing bias. By leveraging the power of artificial intelligence and extended reality, it is possible to create a more equitable and effective interview process beneficial for both interviewers and candidates. As this technology continues to develop, it has the potential to significantly transform the landscape of interviewing across a wide range of fields.

1. Introduction

The employment of interviews is a fundamental method in disciplines such as human resource management, psychology, and the social sciences, and is critical for the collection of qualitative data, evaluation of candidates, and comprehension of human behaviors [1]. However, biases, both conscious and unconscious, can compromise the integrity of traditional interview methods. These biases may manifest in the way candidates are chosen, how their answers are interpreted, and during the overall assessment process, which can result in discriminatory practices and outcomes that are not inclusive [2, 3, 4].

In an innovative step forward, the integration of intelligent digital humans into training programs is being considered. These virtual beings, driven by generative artificial intelligence, are capable of emulating a vast array of human interactions. They offer a uniform interviewing experience, ensuring that evaluations are based on the substance of responses, free from irrelevant influences. These avatars serve as instruments for educating real individuals to identify and mitigate their biases, conversational errors, and unfounded presumptions. They are equipped to mimic different interview situations, displaying an assortment of reactions and behaviors, hence allowing trainees to partake in genuine interactions that can reveal their biases. Avatars can be customized to demonstrate specific traits or elicit particular responses that confront the interviewer's ingrained beliefs or trigger unconscious biases. The aim is to foster a secure and controlled educational setting where one can refine interviewing techniques without the worry of tangible repercussions [5, 6].

Immediate feedback is provided by these avatars, highlighting instances of bias or poor judgment. For instance, if a trainee inadvertently relies on the avatar's looks or speech to make assumptions, the avatar can respond to this directly or through subsequent analytical and feedback sessions. This instantaneous feedback is essential for educational purposes and assists trainees in acknowledging and amending their biases as they happen.

Moreover, the data garnered from these simulations can be leveraged to discern common bias patterns and errors among different trainees. This information can be scrutinized to create specialized educational components that tackle prevalent concerns. By using machine and deep learning techniques, the avatars can evolve in response to a trainee’s progress, presenting increasingly intricate scenarios as the trainee's capability to navigate biases enhances.

Through this method, trainees can confront and comprehend their biases in a manageable environment, cultivating a more conscious and contemplative interviewing stance. This practice may translate into more balanced evaluations of interviewees in actual scenarios, encouraging impartiality and inclusivity [7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23].

In conclusion, intelligent digital avatars act as a progressive educational tool focused on improving the self-awareness and analytical capabilities of the interviewer. This novel approach has significant potential to refine educational practices related to interviewing techniques, possibly leading to fairer results in human resource management, psychology, and social sciences.

As technology progresses, the role of intelligent digital avatars in educational programs is poised to fundamentally transform how interviewing abilities are taught and executed, advancing toward more just and inclusive methods in various fields.

2. Recruitment interviews and bias mitigation

Interview selection processes stand at the forefront of our pedagogical initiatives utilizing avatars, due to their essential role in sculpting an organization's workforce, the very foundation of any enterprise. These selection procedures are a vital component of any organization's hiring strategy, crucial for both inclusiveness and a myriad of other factors. They function as the conduit for attracting talent, shaping the caliber and heterogeneity of the workforce that propels the organization's innovation and expansion. Recruitment interviews scrutinize more than just technical capabilities; they appraise a candidate’s compatibility with the company culture, their latent potential, and their capacity to enhance the team and organization. In light of their substantial influence on the organization’s trajectory and the risk of biases affecting the outcomes, it is imperative to administer recruitment interviews with the utmost fairness and efficiency. Thus, our emphasis on recruitment interviews is designed to refine these pivotal exchanges, ensuring they remain unbiased and uphold the principles of diversity and inclusion.

Biases in recruitment interviews can manifest in multiple forms [24, 25, 26, 27, 28, 29]. For instance, the issue of bias in these interviews may involve interviewers making decisions predicated on extraneous elements such as a candidate's age, gender, ethnicity, physical attributes, and other personal traits, instead of their professional qualifications and expertise. Interviewers may succumb to biases, like confirmation bias and the halo effect, potentially distorting their assessment of candidates and concurrently leading to a workforce that is less varied and inclusive. Biases may stem from subconscious prejudices, stereotypes, and personal predilections. They might also stem from entrenched notions concerning gender, race, and age, resulting in discriminatory hiring practices. An interviewer, for instance, may show preference for applicants with similar academic histories, culminating in a less diverse workforce. They may favor candidates who resemble themselves or those who have previously succeeded due to affinity bias.

Organizations can adopt various strategies to mitigate biases in recruitment interviews. One tactic is to employ structured interviews, which require posing the same queries to each candidate in an identical sequence. This helps guarantee that all candidates are assessed based on the same standards and that interviewers' judgments are not swayed by irrelevant elements.

Incorporating diversity into interview panels can help avert hiring based on collective biases. Leveraging technology can further diminish bias in recruitment. For instance, using artificial intelligence to sift through resumes and job applications to pinpoint the most competent candidates, or utilizing online tests to gauge candidates' skills and competencies in a uniform and impartial manner.

Blind screening is an additional method to enhance the fairness of the hiring process. This technique eliminates unnecessary details from the candidate's resume, such as their name, age, ethnicity, and gender. This practice helps curtail the recruiter’s subconscious biases, ensuring candidates are evaluated purely on their qualifications and expertise.

Another novel strategy, developed in collaboration with Eni, a global corporation with a workforce exceeding 80,000 across more than 80 nations, focuses on training interviewers to heighten their bias awareness and take steps to reduce it. In our case, this involves delivering education and training on diversity and inclusion and motivating interviewers to pursue feedback and perspectives from peers and other stakeholders. An educational route crafted by Barbara Bertagni from e-REAL Labs, following the guidance offered by Martin Eppler and associates [30], revolves around a mnemonic: D.E.B.I.A.S. (Don't Easily Believe in Automated Suggestions). This philosophy centers on a critical inquiry: What constitutes the essence of debiasing? It embodies a prudent skepticism toward our immediate reactions, intuitions, and presumptions, as well as against precipitate advice from others, including AI systems. In this context, debiasing signifies a methodical effort to lessen the influence of biases (departures from rational judgment) on individuals, teams, and organizations. It aids in recognizing or preventing frequent cognitive shortcuts, inequitable preferences, skewed analyses, and poor decision-making (Figure 1).

Figure 1.

A map of the most frequent and impactful biases, part of the D.E.B.I.A.S. program designed by Martin Eppler and colleagues from the University of St. Gallen, Switzerland, and the Nuremberg Institute for Market Decisions, Germany.

3. An innovative educational project on bias-free recruitment interviews

In collaboration with Eni, we have developed a platform designed to elevate social interaction by facilitating the exchange of insights, knowledge, and ideas. This platform, which is one component of the e-REAL Online learning ecosystem, provides an interface reminiscent of popular social media platforms, making it inherently intuitive for users to contribute content in various forms, including text, images, and videos. Moreover, learners are empowered with the ability to manage the visibility of their content using categories and tags, providing a semblance of control and personalization (Figure 2).

Figure 2.

The home page of the e-REAL Online platform about bias-free recruiting interviews.

What sets e-REAL Online apart is its underlying objective to foster a culture of motivated learning. This is achieved through a bespoke gamification framework, which introduces a competitive edge to the learning process. Colleagues progress through leaderboards and rankings that reflect their level of achievement, incentivizing participation. Active engagement is not only encouraged but also recognized across the enterprise, creating a visible acknowledgment of learners' efforts.

The platform's educational components are strategically designed to target the essence of bias-free recruitment practices. These include:

• Simulated recruiting interviews that utilize a diverse array of conversational avatars acting as job candidates. These avatars are designed to replicate a multitude of interviewing scenarios, challenging the learners to confront and overcome inherent biases in their approach.

• Comprehensive tests and self-assessment tools are available for learners to gauge their unconscious biases, providing insights into areas that require attention and development.

• Structured evaluation exercises aim to cultivate a bias-free mindset, enabling learners to apply objective and equitable assessment criteria in their decision-making processes.

• Interactive riddles and quizzes are integrated to subtly confront and challenge implicit biases, fostering critical thinking and self-awareness among learners.

At the heart of this solution lies the innovative use of conversational digital humans. These AI-powered entities are sophisticated enough to provide a robust challenge to learners, pushing them to enhance their skills and competencies in bias-free interviewing. By engaging with these digital humans, learners can experience real-time feedback and dynamic interactions that mirror the complexities of human conversations and relationships. Through this immersive learning environment, our solution aspires to redefine the educational landscape of recruitment by equipping professionals with the tools and knowledge necessary to conduct fair, unbiased interviews. This forward-thinking approach promises to not only transform individual perspectives but also to instill a broader cultural shift toward more inclusive recruitment practices (Figure 3).

Figure 3.

Representative digital humans by e-REAL Labs.

4. The e-real intelligent digital humans and their feedback

In recent times, digital humans have experienced a significant transformation, propelled by advancements in generative AI and machine learning, especially when contrasted with the initial embodied conversational agents from a few years prior [31, 32]. At e-REAL Innovation Labs, in association with prestigious global research bodies, we have been at the forefront of developing a new generation of socially intelligent avatars. These avatars are not only proficient in presenting various subjects dynamically but are also skilled in emotionally resonant interactions, effectively conveying emotions and sentiments.

Designed to showcase unique characters, interactive dynamics, and expressive emotions, the avatars from e-REAL offer an enhanced user experience with both consistency and depth. Their interactive learning capabilities lead to a continually improving skill set in conversation. In the role of simulated candidates during recruitment interviews, their articulated responses, vocal subtleties, and expressive facial features are meticulously constructed to test and train individuals in the role of recruiters. The overarching goal is to foster a cadre of recruitment experts adept at conducting interviews with an acute consciousness of bias, thereby promoting equity and diversity in recruitment choices. This progressive technological advancement signifies a transformative moment in the realm of recruitment training, establishing new benchmarks for the functionalities of intelligent digital entities.

The e-REAL Software Development Team has skillfully programmed these digital humans to mimic authentic human behavior, offering interviewers a challenge in verbal, paraverbal, cognitive, and emotional aspects. This complexity provides interviewers with rich opportunities to address potential biases and mistaken preconceptions. These avatars are powered by cutting-edge AI technologies, such as natural language processing, machine learning, and computer vision, allowing for intricate interactions that include asking probing questions, sharing perspectives, and replicating the nuances of real recruitment interviews.

Our efforts have been geared toward tailoring these avatars for detailed role-playing, converting them into credible conversational counterparts. This enables immersive training environments for handling complex conversations, incorporating factors like non-verbal signals, varied communication styles, and inclusivity. After each session, our platform's advanced monitoring system offers immediate feedback, helping learners to identify and enhance their conversational techniques (Figure 4).

Figure 4.

Representative feedback provided in real-time by the e-REAL tracking system.

The advanced monitoring capabilities of the e-REAL platform are engineered to meticulously record the subtleties of every exchange during the recruitment interview simulations. It accomplishes this by assessing various elements of the dialogue that transpires between participants and the digital avatars. Analyzed factors include speech patterns, language selection, and non-verbal indicators such as facial expressions, body movements, and the timing of responses.

Following each mock interview, the system promptly supplies feedback to the participants. This feedback is comprehensive, accentuating both commendable aspects and those warranting enhancement. For example, it may draw attention to unintended biases in how questions are posed or a propensity to overtalk the avatar, which may be indicative of potential real-life candidate interactions. Additionally, the system might appraise the participant's capacity for establishing a connection, regulating the conversation's progression, and upholding a professional manner.

By delivering this instantaneous, in-depth feedback, participants have the opportunity to contemplate their performance while the memory of the interaction is still vivid. This approach promotes a more effective educational experience by empowering learners to pinpoint and adjust specific actions to better their interview techniques. The aim is to cultivate a more attentive, equitable, and proficient interview methodology that is transferable to actual professional settings, thereby enhancing overall recruitment practices.

5. Communication with digital humans

Digital humans are engineered to comprehend natural language and engage in dialogue. Each digital human is distinct, possessing a unique set of abilities and constraints. For instance, utilizing vague terminology or complex jargon could perplex the digital human; providing context for inquiries or requests aids the digital human in grasping the conversation's intent and yielding more pertinent and precise answers. Allowing the digital human to speak without interruption contributes to a smoother dialogue flow.

Despite their design, which increasingly resembles human behavior, digital humans are fundamentally different from human beings. One of the primary advantages of conversational digital humans is that they offer an interaction reminiscent of human engagement without the necessity for a live person. This feature is particularly beneficial in scenarios where live agents are not accessible around the clock, or when handling a large quantity of interactions is essential. Another merit of employing intelligent digital humans is their capacity to enable learners to experience a variety of scenarios and engage with diverse personalities swiftly, thereby enhancing their proficiency and comprehension.

At Eni, digital humans are a crucial element of the training curriculum, which focuses on the uniform and consistent assessment of candidates. Initial educational assessments indicate that by employing digital humans, the training objectives are achieved. Broadly speaking, conversational digital humans, augmented by artificial intelligence, have the capacity to revolutionize how training is administered across businesses and organizations, fostering more captivating and tailored experiences that contribute to corporate advancement.

Continued investigation is essential to understand the role of intelligent digital humans in diminishing biases during recruitment interviews. At e-REAL Labs, we are dedicated to this line of research, as we perceive generative AI to be a promising catalyst for education and training. A demo of a conversational digital human is accessible by scanning the QR code provided below, which also allows for the arrangement of an online meeting (Figures 5 and 6).

Figure 5.

By scanning the QR code, an avatar will appear and will perform a short self-introduction and provide an online calendar allowing you to book a meeting with several intelligent avatars ready to talk with real human beings.

Figure 6.

Representative conversational digital humans.

Regarding the educational program introduced here, we can say that biases in recruitment interviews can lead to a less diverse and inclusive workforce. The educational program co-developed with Eni provides a solution that reduces biases in recruitment interviews. The use of intelligent digital humans can help to improve the fairness of the hiring process, reduce the time and resources required for recruitment interviews, and increase the diversity and inclusivity of the workforce.

6. Conclusions

Interacting with the e-REAL avatars provides a novel path for swift experiential learning, where trainees are immersed in a vibrant, controlled setting perfect for skill enhancement. This innovative training method not only accelerates the learning process but also acts as a safeguard, mitigating the risks typically involved in conventional, on-the-ground training approaches. In this virtual environment, unlike real-life contexts where errors can lead to serious repercussions, learners are able to trial and perfect their interview strategies without the danger of affecting real hiring processes.

Additionally, each engagement with the digital humans concludes with an in-depth feedback session, injecting a rich educational layer into the training procedure. This feedback system plays a critical role in empowering participants by highlighting precise areas in their interviewing methods that require attention. It encourages introspection and the adoption of a growth mentality, which is essential in cultivating a culture of perpetual learning among HR professionals.

The sophistication of these digital humans goes beyond simple role-play. They are designed to authentically emulate human reactions, presenting interviewers with complex challenges. Through both verbal and non-verbal communication, they elicit intricate responses that reflect the dynamics of actual interviews. Furthermore, these virtual counterparts incorporate both cognitive and emotional complexity into the learning scenario. Capable of displaying a range of emotions and temperaments, they test interviewers on their ability to manage the subtleties of human interactions. This deep, immersive practice acts as a reflective tool, highlighting the dangers and weaknesses that may arise when biases and misconceptions infiltrate the interviewing process.

The adoption of these intelligent digital humans is a leap beyond conventional training practices, providing a comprehensive and revolutionary learning experience. It not only expedites the advancement of skills but also nurtures self-insight, instills a dedication to ongoing enhancement, and emphasizes the importance of mindset in achieving diversity and inclusivity within recruitment protocols.