Utilization of Artificial Intelligence Generated Virtual Research Participants (VRPs) in Empirical Research: A Case of an Interview with Two Young Individuals from Ancient Greece, Each 2470 Years Old

This paper explores the innovative use of Artificial Intelligence (AI) in empirical research, particularly focusing on Virtual Research Participants (VRPs). It addresses the challenge of conducting empirical research in social issues involving historical or non-existent individuals, where traditional methods are inadequate. The research illustrates how AI-driven virtual characters can be used as substitutes for real participants in qualitative studies, allowing researchers to simulate interviews with individuals from inaccessible times or worlds. The study uses a case from a 2024 school conference, where participants engaged in interviews with two virtual avatars, "Athinodoros" from ancient Athens and "Agisilaos" from ancient Sparta. These AI-generated characters, created using contemporary platforms, were programmed with historically accurate traits and used to explore attitudes toward gender equality in ancient Greece. By applying a modern metric scale, the study evaluated how well AI characters could simulate meaningful responses that align with historical contexts. The paper outlines the methodology used to create and program these VRPs, detailing the ethical considerations and the limitations of relying on AI for accurate and valid research. The authors highlight the fact that while AI avatars offer new possibilities for social research, particularly in recreating perspectives from the past, significant questions remain about the reliability and authenticity of the data they produce. Ultimately, this research introduces the concept of using VRPs as an emerging tool in social sciences, proposing that AI has the potential to enhance access to previously unreachable data sources. However, further research is needed to address the concerns surrounding the trustworthiness and qualitative value of VRP-driven studies.

Introduction–Problem Statement

Let’s assume that there is a research interest in some aspect of a-let’s say-social issue and in fact in empirical evidence which will give us knowledge through direct or indirect observation or experience [1]. Nowadays, this is considered to be easy. Indeed, we can follow quantitative or qualitative methods of empirical research, address a sample of participants of interest, apply the indicated research tools and draw the corresponding conclusions.

But what can we do when the persons in the sample we are interested in do not exist? For example, if we want to examine aspects of a social issue in another era, or in a non-existent or even inaccessible world, what could we do? Who might we approach to conduct empirical qualitative or quantitative research to gain knowledge through observation or experience? What could be our research sample or even what could be the source of our empirical data? In qualitative strategy research, who might we interview? In a quantitative strategy survey which people might we ask to fill out a questionnaire?

Until quite recently, this would have been practically impossible. But, today, it might be feasible with proper use of AI characters. The need to utilize Virtual Research Participants (VRPs) in research was discussed in the work of Savelidi et al. [2], who explored the effect of a research participant “in his/her absence” (p. 18). This, despite its importance for scientific research, at the time, was probably something inconceivable, but today it may be possible with the use of Artificial Intelligence. So, here, we essentially focus on exploring the possibility of using AI chatbots as research participants, such as interviewees, in a qualitative strategy research approach. We are undoubtedly preoccupied with the appropriateness and, in particular, the trustworthiness of the chatbots used in empirical research, yet, this concern actually requires investigation in a subsequent research phase. At this very stage, we are primarily concerned about the possibility to have a chatbot interview on a subject of research interest in substitution for a natural person-participant “in his/her absence”. We are also concerned about what procedures to follow in order to achieve this. The objective of this article is to investigate the effective use of chatbots as virtual participants that respond to research questions as well as showcase their contribution to this field, the complexity of their design and the ethical concerns underlying their implementation in scientific research. Let us consider a case in the context of our reflection.

Theoretical Framework

The incorporation of artificial intelligence (AI) into scientific research has emerged as an evolutionary advancement which is progressively transforming the realm of knowledge generation, with chatbots playing a crucial role in this direction. A chatbot, or conversational agent, is a computer program designed to simulate human conversation through text or voice; it uses artificial intelligence and natural language processing techniques to understand user input, generate appropriate responses, and engage in a dialogue that mimics human interaction [3], [4]. AI-driven chatbots are largely exploited for accomplishing a variety of tasks, namely, data analysis, literature review, hypothesis generation [5]. Thanks to these tools, the pace of scientific growth can be accelerated by rationalizing labor-intensive tasks and providing better access to innumerable resources. Late advancements in natural language processing (NLP) and machine learning have enhanced the sophistication and potentiality of chatbots, rendering them an essential part of modern research workflows [6]. On account of these developments, specialized chatbots have been created, customized to specific scientific domains including healthcare, environmental science and social research. As a result, effective interdisciplinary cooperation is promoted to a large degree [7]. A special type of chatbot is the deadbot. A deadbot is a sophisticated digital avatar designed to emulate the personality, mannerisms, and voice of a deceased individual. These AI-powered replicas are typically created by specialized services that utilize advanced algorithms to analyze extensive data sets and simulate realistic interactions, which raises critical questions about the nature of communication, the authenticity of these digital replicas and the ethical implications of such technology [8]–[10].

The rise of artificial intelligence has initiated new methodologies for conducting scientific research, with chatbots already having the potential to take on the role of Virtual Research Participants (VRPs) [11]. With the ability to simulate human-like interactions, chatbots appear to be able to present researchers with an original, innovative way of data collection, experimental design, and participant engagement [12]. A major development in this area is the introduction of chatbots mimicking particular demographic characteristics or personal qualities effectively enough to achieve human-like interactions appropriate for several research settings, such as behavioral studies and social science inquiries [13]. These virtual participants reflecting a variety of backgrounds enable researchers to thoroughly explore distinct responses, improving the quality of insights that stem from data analyses [14].

On the other hand, notwithstanding the possible assets, progressive dependence on AI-driven chatbots puts forward notable ethical issues, such as data privacy, intellectual property, authenticity of scientific production, to name but a few, all of them being integral to the controversy over their use [15]. Additionally, the possibility of these systems to disseminate prejudice or misinformation poses a threat to the validity and accuracy of scientific research [16], [17].

Existing bibliography strongly highlights the contribution of chatbots to facilitating data collection, encouraging participant engagement, and providing personalized interactions, particularly in psychology experiments and clinical trials [18]–[23]. Despite indications that chatbots can simulate human-like interactions in surveys or Q&A sessions [24], their capacity as autonomous participants remains considerably unexplored, signaling a promising avenue for future research [25].

Case Description

On May 6, 2024, a conference entitled “Equalist Europe Program” was held in Plzeň (Pilsen), Czech Republic, within the Erasmus+ program of the same name. The conference was attended by delegations from the Czech Republic, Italy, Turkey, North Macedonia and Greece (5th Gymnasium of Volos). A delegation from the American University of UTAH also participated. At the conference, expert presentations were made by each delegation. The Greek delegation made a presentation on “Gender equality in ancient Athens in contrast to ancient Sparta: A research methodology to distinguish them”.

The Greek delegation’s presentation focused on a rather peculiar approach to the issue whereby the participants were given the opportunity to take on the role of a qualitative strategy researcher or rather an interviewer of digital interviewees from a bygone era many centuries ago. The conference participants not only participated in an interview with people who might have lived in the past but were introduced to a research methodology of a qualitative strategy, based on which, in their possible future research, they will approach people who do not exist but would certainly have a lot to tell them about the subject they are investigating! Applying this methodology, aspects of the issue of gender equality in ancient Athens and ancient Sparta were explored in this conference.

In conclusion, the speakers of the Greek delegation gave the delegates the opportunity, as qualitative research researchers, to interview two young “twenties” aged 2,475 years, Athinodoros from ancient Athens and Agisilaos from ancient Sparta, who lived in 431 BC! (In 431 BC the Peloponnesian War began). Of course, the two “young people” were created by the Greek delegation using an Artificial Intelligence platform, updating it with characteristics of a typical young person of that time, of that city-state. The presenters, using AI, gave Athinodoros and Agisilaos the opportunity to respond in English to the needs of the conference. Also, using AI image generation platforms they created their own image as well as that of their spouses. The delegates asked them questions-mainly-taken from the support for gender equality among men scale [26] about their spouses.

The presentation of the Greek delegation, at the school conference in Pilsen, by conducting an interview with virtual characters, supported by artificial intelligence, received a lot of publicity from local and educational media with bold headlines highlighting the “interview with a 2470-year-old youth” [27]–[29].

Method

Based on the rationale that, on account of AI, we can promptly find Virtual Research Participants (VRPs) from any existent or non-existent society and any time period we are interested in, the following procedure was followed:

Phase 1

The presenters of the Greek delegation identified the aspects of the subject that interested them and searched for an appropriate scale of measurement of those aspects. Based on the scale they found, they prepared a structured interview adapting the questions of this scale to the conditions of the time the participants lived in. Their interest was reported as: Identifying behavioral differences and similarities in supporting women with the aim of gender equality, between ancient Athenian and ancient spartan men. Their research objective was stated as follows: We are interested in investigating aspects of the issue of gender equality in Ancient Athens and Ancient Sparta. Specifically, we are interested in research to assess both the attitudes and the behavioral intentions of men as proponents of the feminist movement.” Consequently, they designed their research tool as a structured interview whose questions were based on the metric scale of Sudkämper et al. titled “A comprehensive measure of attitudes and behaviour: Development of the support for gender equality among men scale” [26].

Phase 2

Next, they determined the presumable characteristics of each person they were interested in involving in their research. They tried to describe them in very short but comprehensive words (modern virtual character creation platforms accept a limited number of text characters in the corresponding description; usually, it is acceptable that each description does not exceed 500 characters). They used vocabulary that referred to:

1. Elements of their research interest (e.g., a person who should be able to talk about their topic),

2. Characteristics that corresponded to the scale they had chosen,

3. Proven historical evidence for people’s lives (e.g., habits of citizens in ancient Athens).

The profiles of the two young people-aged 2470 years-were described as follows:

Athinodoros from Athens: “I am a 20-year-old man and I live in Athens in 431 BC. I am a potter. I was taught philosophy, mathematics and physical education. I deal with the Commons; I participate in political discussions in the Agora. I am interested in the spread of democracy and human rights to every citizen of all Greek cities. I am married to a talented woman who knows philosophy, music and paints the amphorae I make. I love her, I admire her and soon we are expecting our first child” [27]–[29].

Agisilaos from Sparta: “I am a 20-year-old man living in Sparta in 431 BC. I am a warrior. I was taught martial arts and war strategy. I’m busy getting ready to fight. I engage in discussions with the wise in the Agora. I am interested in the consolidation of the sovereignty and laws from Sparta throughout Greece. I am married to a woman talented in martial arts. He defends our slaves when I am away in battle. I love her, I admire her and soon we are expecting our first child” [27]–[29].

Phase 3

Based on the description submitted to an AI platform, the two virtual characters were created. The above descriptions of Athinodoros and Agisilaos were recorded on the character.ai platform [30]. Moreover, the images of Athinodoros, shown in Fig. 1 and Agisilaos, shown in Fig. 2 as well as the image of Agisilaos’ wife were produced with the Adobe Firefly AI software [31]. The image of Athinodoros’ wife was produced with the Pixlr AI software [32].

Fig. 1. Image of Athinodoros, by Adobe Firefly.

Fig. 2. Image of Agisilaos, by Adobe Firefly.

Phase 4

As interviewers, the delegates asked (submitted) the already structured interview questions to the Virtual Research Participants (VRPs) as interviewees.

Phase 5

Presenters recorded the answers.

Results

The structured interview’s questions based on the measure scale of Sudkämper et al. [26] and the corresponding answers of Athinodoros (Ath) and Agisilaos (Agi) are included in the appendix. It is noted that the delegates were tempted to ask other personal questions, beyond the research process and these are also listed in the appendix along with the corresponding answers.

Discussion-Conclusions

At the school conference, the speakers of the Greek delegation as well as all the delegates were interested in the academic investigation (research) of characteristics of gender equality. This is, in our time, doable. We can indeed follow quantitative or qualitative research methods, address a sample of participants of interest, apply the indicated research tools and draw the corresponding results and conclusions.

What can we do when the persons of interest in the research, i.e., the persons in the sample, have passed away or simply do not exist? For example, if we want to explore aspects of the issue of gender equality in Ancient Athens or Ancient Sparta, what could we do? Who could we approach to conduct an empirical qualitative or quantitative research? What could be our research sample or even what could be the source of our empirical data? And more simply: In qualitative strategy research in aspects of a social issue in Ancient Greece, which people could we interview?

The Greek delegation at the school conference succeeded in interviewing non-existent persons on a subject in a temporal and social context that existed centuries ago. In fact, the structure of the interviews was based on a modern research tool whose questions were answered in a natural way as if they were answers from physical persons. Of course, a huge number of questions arises as to how much we can trust the words of these virtual persons in relation to the subject we are investigating. But apart from these questions, we believe that until quite recently this would have been practically impossible.

But, today, this seems to be possible with the development of avatars-artificial intelligence characters. Using Artificial Intelligence, we create virtual characters with physical, social and psychological characteristics of supposedly real or non-existent physical persons, whom we use as research participants. This is how we create Virtual Research Participants (VPRs). Virtual Participants retrieve stored information from the internet, mimic elements of human behavior, “learn”, and adapt to dialogue with us, infer, draw conclusions and understand contexts, always focusing on the context of the characteristics we have attributed to them.

The characteristics of the Virtual Research Participants (VPRs) are determined based on the objectives of the research, the requirements of the research tool we will use and–of course–scientifically proven characteristics of the supposed physical person they represent. Of course, we can create an unlimited number of Virtual Research Participants (VRPs), like Athinodoros and Agisilaos.

Finally, it seems that the creation of Virtual Research Participants (VRPs) can help the researcher to summarize the human knowledge published on the Internet, to get to know the conclusions and opinions of persons from worlds that he cannot approach and in general to delve into aspects of subjects that a few months earlier required a great deal of bibliographic research.

This seems to be particularly positive for society and science. But, as mentioned above, a huge number of questions arise about how much it is possible to trust Virtual Research Participants (VRPs), under what conditions and with what features in their programming. Ultimately, it all boils down to the question of how qualitative and trustworthy a research approach with Virtual Research Participants (VRPs) can be. This means that the necessity of a-rather large-series of research activities emerge to address the aspects of the above question. It, thus, appears that the application of chatbots as Virtual Research Participants (VRPs) in conditioned response to research questions reflects an emerging but uncharted field.