Can an intelligent personal assistant (IPA) be your friend? Para-friendship development mechanism between IPAs and their users

Highlights

• The mechanism by which users develop virtual friendships with intelligent personal assistants (IPAs) is investigated.

• The two critical dimensions of para-friendship are self-disclosure and social support.

• IPAs that provide a sense of intimacy, understanding, enjoyability, and involvement affect para-friendships.

• Para-friendships strengthen users' intention to continue using IPAs.

• The para-friendship mechanism differs according to the users' social isolation tendency.

Introduction

Artificial intelligence (AI) refers to the creation of human-like intelligence in machines that are programmed to think like humans and imitate intelligent human behavior (Shi, 2011). The applications of AI have been implemented in the marketplace in notable ways (McLean & Osei-Frimpong, 2019). Among the various AI devices available to consumers, the most prevalent is an intelligent personal assistant (IPA) (López, et al, 2017). IPAs refer to automated software applications or platforms that assist people with information searches and decision-making efforts using natural language in either a written or spoken form (Reis, Paulino, Paredes, & Barroso, 2017). Commercial examples include Amazon's Alexa, Apple's Siri, Google's Google Assistant, and Microsoft's Cortana. These conversational IPAs have become increasingly popular because they facilitate human-computer interactions in a natural and intuitive way, making them similar to those of interpersonal interactions by answering questions, following a conversation, and helping users accomplish tasks (European Commission, 2018). The number of people using IPAs in the United States reached 102 million in 2018, which is about a 28% increase from 2017, and this number is expected to exceed 122 million by 2021 (Statista, 2019b).

As the popularity of IPAs grows, much attention has been put on understanding users' interactions with IPAs, particularly concerning the reasons why people acquire and utilize IPAs (Kaplan & Haenlein, 2019; López et al., 2017; McLean & Osei-Frimpong, 2019). Evidently, the key reason people use IPAs is to obtain functional benefits (Forbes, 2018). A survey by Statista (2019a) indicated that 52% of their respondents preferred using an IPA over a website or other applications because they found IPAs more convenient. About 48% of them reported that they use an IPA because it allows them to multi-task or work hands-free. Academic literature also indicates that IPAs' utilitarian benefits are critical factors for users to adopt and utilize them. For example, the ability to access IPAs while moving from place to place (i.e., mobility) and their ability to interoperate with applications made by different vendors (i.e., interoperability) increases users' intention of acquiring IPAs (Yang, Lee, & Zo, 2017). People's perception that using IPAs will enhance their job performance (i.e., perceived usefulness: Nguyen, Ta, & Prybutok, 2018) and are easy to use (i.e., perceived ease-of-use: Moriuchi, 2019) also enhances their attitude and intention to try IPAs.

However, IPAs have advanced technologically, and their higher skills now provide consumers with more than just functional and utilitarian benefits. For example, to make customers continuously use IPAs, companies have developed emotion-detecting AI that uses algorithms that can identify how people are feeling (McStay, 2018). For example, to make Alexa sound more human, Amazon developed neural text-to-speech technology that enables its IPA to mimic the intonations of human speech and human emotions (Schwartz, 2019). With this technology, Alexa can make happy, excited, or sad sounds, and these emotional responses were found to improve overall customer satisfaction by 30% (Schwartz, 2019). While IPAs continue to improve and evolve to the phase of recognizing human emotions, academic literature has been rather passive in understanding whether users indeed sense the emotions emanating from IPAs and perceive the devices as their friends with whom they can share their honest feelings and, in return, feel cared for in times of need. While the literature has focused on uncovering users' utilitarian experiences with IPAs (Guzman, 2018; Hoy, 2018), users’ emotional, relational, and social experiences with IPAs have been relatively underexplored in the literature (McLean & Osei-Frimpong, 2019).

In response, this study addresses this issue by examining the mechanism with which users develop virtual friendships (i.e., para-friendships) with IPAs. In doing so, we place self-disclosure (i.e., users' perception that they can reveal their honest feelings to IPAs) and social support (i.e., the perception that they can turn to IPAs in times of need as if IPAs are their friends) as the two most critical dimensions of para-friendships. Specifically, we investigate whether the emotional traits in IPAs—the extent to which IPAs offer a sense of intimacy, understanding, enjoyability, and involvement—affect users' self-disclosure and social support toward IPAs, all of which lead to para-friendships. We then test whether the extent to which users develop a para-friendship with an IPA affects their intention to continuously use IPAs (i.e., stickiness intention). Given that individuals' personal tendency to feel lonely, friendless, and socially remote from others influences their propensity to make self-disclosure to, and seek social support from, others (Leung, 2002), we also examine whether and how the relations depicted in our model differ by users’ personal social isolation tendency.

We based our conceptual model on CASA, a term coined by Nass and Moon (2000) to indicate the phenomenon by which individuals apply social rules and expectations to computers, even though they are aware of that computers do not have any feelings, intentions, or motivations like humans. The theory proposes that people treat computers as social actors rather than just a medium by assigning human traits (e.g., gender or ethnicity) and characteristics (e.g., reciprocity or dominance) to computers. For example, certain characteristics of computers, such as the interactivity of voice input and output and the filling of social roles traditionally provided by humans, were documented as important cues to providing sufficient bases for users to perceive humanness from computers (Nass & Moon, 2000; Wang, Baker, Wagner, & Wakefield, 2007). The more computers displayed human-like characteristics, the more the users applied interpersonal social norms or social responses to the computers they were interacting with (Nass & Moon, 2000). For instance, when computers asked users to evaluate them, the users tended to evaluate them positively as a means to show their politeness. In face-to-face interactions, people tend to give positive evaluations when another person asks them to evaluate her/himself because they are reluctant to hurt anyone's feelings. Likewise, users applied this social rule to human-computer interactions, although they know that computers do not have any feelings like human do (Nass & Moon, 2000).

Much evidence exists of the CASA paradigm across various forms of human-computer interactions. For instance, in human-robot interactions, the facial expressions and autonomous movements from robots trigger users to treat these robots as humans and show an interpersonal attraction toward them (Lee, Peng, Jin, & Yan, 2006). In particular, when a user interacts with a robot that has a personality that is complementary to his own, the user finds the robot to be more enjoyable, more attractive, and thus socially more present than a robot that does not have a personality similar to his own (Lee et al., 2006). In addition, the mechanism of CASA was applied to examine human-avatar interactions. For instance, Wang et al. (2007) show that the more an avatar is perceived to be polite and interactive by users, the more they feel the avatar as socially present, causing the users to feel greater interest, attention, and patronage toward the online retail store of the avatar.

These studies provide evidence that computers, robots, and avatars serve as social actors if they project human-like cues to their users, and thereby make the users feel they are real and present (i.e., socially present). As providing a sense of social presence is essential in facilitating computers or other communication medium to serve as social actors to users, we discuss the concept of social presence more in detail in the next section.

Social presence refers to the extent to which a communication medium (e.g., television) conveys a sense of psychological connection to its users (Yoo & Alavi, 2001). For example, when users perceive a medium to be intimate, personal, and sociable, thus providing a sense of being in touch, they feel the medium is socially present. As their communication medium expands to the digital realm, the concept of social presence also advances to provide a more nuanced account as to how the feelings of social presence can be formed in an online environment. Specifically, Kumar and Benbasat (2002a) coined the term para-social presence to denote the extent to which a website facilitates a sense of understanding, connection, and involvement with its users by depicting the users’ interactions with the website.

Indeed, the concept of para-social presence is distinct from that of social presence in that they define presence differently. For example, social-presence views presence through traditional dimensions, which include social richness (i.e., the extent to which a medium can reproduce the information sent over it) (Burke & Chidambaram, 1999; Carlson & Davis, 1998; Daft & Lengel, 1986); realism (i.e., the extent to which a medium can produce realistic representations of the things one is interested in) (Lombard & Snyder-Duch, 2001); transportation (i.e., the extent to which a medium can facilitate people to lose themselves in a story) (Green & Brock, 2000); and immersion (i.e., the extent to which a medium immerses people psychologically) (Biocca & Delaney, 1993).

On the other hand, para-social presence bases its conceptualization of presence in line with the CASA paradigm by evaluating the extent to which users perceive social cues from an online medium (i.e., social actor as a medium) or by measuring the extent to which a person or an entity within an online medium develops social relationship with its users (i.e., social actor within medium) (Kumar & Benbasat, 2002a). For example, a para-social presence emphasizes the importance of identifying the emotional and relational mechanism between a website and its users or between an avatar within a website and its users.

When uncovering these relational mechanisms involving an online medium, four relational qualities are particularly essential: (1) intimacy, which is formed when users feel close and attached to any person or entity within a medium (i.e., a media entity); (2) a sense of understanding, derived when users’ intentions, needs, and emotions are well understood by a media entity; (3) enjoyability, created when a media entity is friendly, fun, and pleasant; and (4) involvement, derived when users are engaged in, or having immersive interaction with, a media entity (Kumar & Benbasat, 2002b).

Given that an IPA is a more advanced version of an online media entity because it integrates an AI-powered virtual assistant (e.g., Alexa) into an online medium (e.g., Amazon's Echo), understanding whether IPAs can also serve as social actors to its users through the aforementioned qualities is important. We particularly associate the four para-social presence qualities to examine whether IPAs can serve as virtual friends to their users. In so doing, we additionally review the concept of para-friendship that highlights two relational traits in developing a friendship with an online media entity: self-disclosure and social support.

The concept of para-friendship stems from that of a para-social relationship. The term para-social relationship was coined to describe the phenomenon in which media cultivates virtual relationships between media entities and its viewers, which resemble the relationships in real life (Levy, 1979). For instance, when media users are exposed to Justin Beiber over time who openly shares his activities and opinions on television, the users tend to develop an intimacy with him as if they know him in person.

Until recently, the literature has long treated a para-social relationship as a single, homogeneous relationship that contains elements of empathy, perceived similarity, and physical attraction (Rubin, Perse, & Powell, 1985). However, para-social relationships encompass a wider spectrum of distinct relationships, varying from a mere acquaintance to friendship or even love. Therefore, more recently, researchers have divided para-social relationships into para-friendships and para-loves depending on their nature and intensity (Liebers & Schramm, 2017; Tukachinsky, 2011). Para-friendship refers to a sense of companionship developed between a media entity and its users, in which users desire self-disclosure with, and show support for, a media entity. For instance, people often feel a television character is a real person and are concerned for their well-being just as they are concerned about their own friends. On the other hand, para-love is marked by more intense feelings, in which users feel a physical attraction to a media entity and have passionate thoughts about him or her (Liebers & Schramm, 2017). For example, people can project romantic feelings onto a media character and desire to go out with him or her. In para-social relationships, para-friendship is more important because of the long-standing view that a para-social relationship is a quasi-friendship (Tukachinsky, 2008).

The literature has long emphasized that a friendship is founded on self-disclosure and support (Davis, 2012; Hays, 1984). Likewise, two critical dimensions of a para-friendship also include self-disclosure, the perception that one can reveal his honest feelings and thoughts to others, and social support, the perception that one is cared for and has friends to turn to in times of need (Tukachinsky, 2011). Self-disclosure is important in developing a para-friendship because other noticeable elements that help make friends in real life, such as playing sports together or being physically proximate, cannot be fulfilled in an online environment (Kim & Song, 2016; Tukachinsky, 2011). For instance, online communities provide an environment where people develop para-friendships by sharing common interests and thoughts among the members (Shih, Hsu, & Lee, 2015). Social support is another critical quality for developing a para-friendship because, as technology advances, the availability and the desire for online support increases as well (Oh, Ozkaya, & LaRose, 2014). We thus expect users to develop a para-friendship with an IPA by showing self-disclosure and social support toward it.