Speakers
Description
This chapter explores the meaning of agency for artificial intelligence (AI) and raises questions about human-AI interaction, particularly focusing on social chatbots. Using Adam Smith's theory of sympathy as a framework, we explore how the absence of fundamental emotions like sympathy hampers machines' ability to understand and ethically respond to certain situations, thereby questioning their ability to serve as moral agents comparable to humans. Moreover, we examine the practical repercussions on human-machine relations, highlighting how these emotional limitations could undermine the essence and depth of human friendships. We will delve into a discussion on how AI systems shape our social behaviors by introducing new forms of interaction that blur the lines between the real and virtual worlds. Particularly, the experience with so-called AI social chatbots that emulates human relationships of friendship and love may suggest a deeper kind of relationships with them. For instance, chatbot friends are like human friends, but they are not human or simple machines like our coffee maker or an automated car. This generates a phenomenon of anthropomorphization of this AI that alters our view on the original meaning of social engagement, acknowledging it as a non-human agent that significantly influences our behavior and expectations within relationships. Our initial step is to explore the possibility of AI agency and whether any sense of agency exists within them. We argue that there is an emotional lapse (Barbosa & Costa, 2023) in AI systems that poses a challenge as to whether AI can truly develop a sense of agency that meets social engagement requirements. According to Legaspi et al. (2022, p. 40), “the sense of agency (SoA) is the subjective experience that one initiates and controls one’s own actions and through them the external world.” Despite advancements in this field, a significant challenge persists in translating the deeply human and subjective experience of Sense of Agency (SoA) into computational models that can genuinely relate to humans. We seek to determine the practicality of creating AI systems that transcend simple interaction or emulation, truly operating under a sense of agency.
Once this is established, we will discuss the practical implications of these shortcomings in social human-machine interactions, with a particular focus on their influence on friendships. So far, it is important to determine how the absence of these relational social phenomena in human-AI interactions may compromise genuine interpersonal connections due to its distinct nature from what is a genuine human relationship. This represents a break from the notion of sympathetic engagement that Smith presents in The Theory of Moral Sentiments, which may interfere with socio-economic relations among individuals. Contemporary societies have experienced a significant increase in loneliness and social isolation of individuals – expressed in the figure of hikikomori in Japan and South Korea –, clearly indicating changes in social behavior and aspects of social psychology. We will explore how the concept of friendship, traditionally characterized in human-human interaction as a special concern for another person (see Annis, 1987), can be applied to human-AI interactions. We will explore this phenomenon from Smith's perspective, which calls for friendship as "the best and most comfortable of all social enjoyments." (TMS VI.iii.15) For that, it will take in considerations factors such as the asymmetrical nature of human-AI interplay and how it can reinforce human-human social isolation in societies When AI is a key component of an environment, psychological and cognitive alterations such as changes in social dependency and emotional depression can occur. This situation is often exacerbated by the shallow nature of AI interactions, which may lead individuals to favor these over deeper human connections, thereby further reinforcing social isolation. This integration of AI into daily life could indicate the emergence of an unprecedented form of human relationship or interaction.
| Organization | Federal University of Pelotas - Brazil |
|---|
