Fiction has been serving up tales about humans interacting with different kinds of AI for over half a century now. But is science today equipped with the know-how that can turn these exploits into reality? A study including a NEOMA researcher, Agata Mirowska, has analysed the evolution of research into virtual agents over the last 30 years.

“But please remember: this is only a work of fiction. The truth, as always, will be far stranger”, pronounced Arthur C. Clarke in the foreword to 2001: A Space Odyssey. In 1968, Clarke’s novel and Stanley Kubrick’s eponymous film introduced the HAL 9000 computer to the general public for the first time. This virtual crew member of the Discovery One spaceship was powered by AI so that it could easily take part in conversations. In 2002, the movie S1m0ne told the story of a director who is obliged to replace the star of his feature film with a virtual actress. Ultimately, reality has turned out to be closer to the latter scenario: in 2023, film studios in the US are aiming to replace actors with AI-generated images. Has technology finally caught up with fiction? The study, by NEOMA researcher Dr. Agata Mirowska and her colleagues, analysed developments in scientific research into virtual agents between 1995 and 2022. What did they discover during this “close encounter with the virtual kind”? And what will the virtual agents of the future look like?

Research ramped up by the rise of AI

Although fiction has been preparing us up for the idea of interacting with virtual agents for over 50 years, the researchers found there was no commonly-agreed on definition of the term, leading them to define it as: “A digital entity equipped with artificial cognitive capabilities that can interact with human users at the same time as performing tasks for them with a degree of autonomy”. Semi-autonomous beings have already become part of our lives little by little: witness the various types of chatbot, virtual influencers and home assistants that are by our side on a daily basis. In practice, they’re used as much for surveillance as rapid decision-making.

The researchers observed that this step-up is directly linked to the progress made by AI and smart device technology, which has sparked an explosion in research since 2014. Over time, however, virtual agents have moved beyond the bounds of computer science and engineering research, which centred on design issues. Nowadays the focus is in connection with areas of application and integration into society. The most topical research fields in the current scientific literature are health – especially neuroscience and psychology – and education. Studies into the normalisation of virtual agents have always gone hand-in-hand with innovation. In the future, these digital programmes are likely to have a greater impact on our lifestyles owing to their heightened autonomy. These developments, however, will still rely on discoveries derived from AI research. 

What does the future hold for virtual agents?

The study spotlights the emerging concept of the co-existence of humans and virtual agents. It stresses that we need to understand the desirable and undesirable characteristics that will affect how we live our lives and work with these smart algorithms. Research so far has zeroed in on the positive aspects. However, the democratisation of these agents across different sectors of society means that we will have less and less choice about how we interact with them. This explains why the researchers underline the importance of analysing what could turn humans off of using the agents and leave a negative impression. For instance, tackling the idea of intimacy in our interactions with personal-care virtual agents or the promotion of social values. In more general terms, the role of these entities in society needs to be determined.

Furthermore, the credibility and acceptability of these programmes also raise questions. Virtual agents may facilitate the general public’s adoption of AI – or its rejection of it. Future research should continue exploring the ethics and responsibility of algorithms and the trust humans have in these digital beings. This co-existence should be studied in a range of different user contexts and exchange environments, which are both digital and cyber-physical. The researchers argue that virtual agents will steadily evolve into AI-equipped social robots with an increasingly sophisticated physical form.

Find out more

Jbid Arsenyan, Agata Mirowska, Anke Piepenbrink. Close encounters with the virtual kind: Defining a human-virtual agent coexistence framework. Technological Forecasting and Social Change, Volume 193, 2023. https://doi.org/10.1016/j.techfore.2023.122644