Human-AI Interaction: History and Design

Authors

  • Elisabet Delgado Mas University of Vienna

Abstract

Human-AI Interaction is a multidisciplinary field of study concerned with the design of interactive artificial intelligence for human use and the interfaces that mediate it. As a research topic, it is situated at the intersection between applied computer science, design, linguistics, media studies, psychology and other behavioural sciences, given that the quality of these interactions depends on multiple factors.

Since the beginnings of what would later become known as artificial intelligence, humans have tried to interact with these systems. Pioneering examples range from the psychotherapeutic chatbot Eliza (MIT), designed for a terminal screen back in 1966, to state-of-the-art, hyper-realistic affective robots such as Ameca (Engineered Arts), neurorobotic agents with a biologically inspired model of emotions like BabyX (Soul Machines) [1] and artificial companions running on sophisticated neural network machine learning models for open-ended conversations (Replika). As AI advances, models of simulated humans are becoming increasingly realistic [1]. Some studies seem to claim greater realism is likely to improve the user experience, and others alert that having a human likeness continuum paired with atypical features can relate to eeriness [2]. Human-computer interaction (HCI) literature has not fully explored this area and guidance on conducting research on digital AI-driven experiences is needed [3].

This exploratory research will first make an overview of the history, evolution, tendencies and state of the art of human-AI interaction to gather the perspectives and approaches that will enable a deeper understanding of its factors, and how the socio-cultural context of each period and interindividual differences contribute to the user experience of human-like interactive AI systems.

Secondly, we will see how this can be applied in the design of those systems and what the best practises, principles and heuristics to design these interactions currently are, and if they require adaptation to different cultures and individuals.

In summary, the desired outcome of this theoretical work is to provide an informed basis for facilitating the design of favourable experiences during human-AI interactions with AI systems that have human-resembling responses, which will become increasingly common in our societies in the forthcoming years.

References

[1] A. Knott, M. Sagar, and M. Takac, ‘The ethics of interaction with neurorobotic agents: a case study with BabyX’, AI Ethics, vol. 2, no. 1, pp. 115–128, Feb. 2022, doi: 10.1007/s43681-021-00076-x.

[2] T. J. Burleigh, J. R. Schoenherr, and G. L. Lacroix, ‘Does the uncanny valley exist? An empirical test of the relationship between eeriness and the human likeness of digitally created faces’, Computers in Human Behavior, vol. 29, no. 3, pp. 759–771, May 2013, doi: 10.1016/j.chb.2012.11.021.

[3] C. Kliman-Silver, O. Siy, K. Awadalla, A. Lentz, G. Convertino, and E. Churchill, ‘Adapting User Experience Research Methods for AI-Driven Experiences’, in Extended Abstracts of the 2020 CHI Conference on Human Factors in Computing Systems, New York, NY, USA, Apr. 2020, pp. 1–8. doi: 10.1145/3334480.3375231.

Published

2022-06-23