Disentangling Two Fundamental Paradigms in Human-Machine Communication Research: Media Equation and Media Evocation

Source: Margot J. van der Goot, Katrin Etzrodt; Human-Machine Communication 6, journal article (open access) · 2023-07-01

In this theoretical paper, we delineate two fundamental paradigms in how scholars conceptualize the nature of machines in human-machine communication (HMC). In addition to the well-known Media Equation paradigm, we distinguish the Media Evocation paradigm. The Media Equation paradigm entails that people respond to machines as if they are humans, whereas the Media Evocation paradigm conceptualizes machines as objects that can evoke reflections about ontological categories. For each paradigm, we present the main propositions, research methodologies, and current challenges. We conclude with theoretical implications on how to integrate the two paradigms, and with a call for mixedmethod research that includes innovative data analyses and that takes ontological classifications into account when explaining social responses to machines.

Introduction. In this theoretical paper, we delineate two fundamental paradigms in how scholars conceptualize the nature of machines in human-machine communication (HMC). In addition to the well-known Media Equation paradigm (e.g., Reeves & Nass, 1996), we distinguish the Media Evocation paradigm (following the concept of evocative objects in Turkle’s influential work of 1984 and 2007). We argue that these two paradigms fundamentally differ in their explanations for why humans respond socially to machines, in their research questions, and in their research methodologies. The key notion in the Media Equation paradigm is that people respond mindlessly to machines as if they are humans. In contrast, the Media Evocation paradigm conceptualizes machines as objects that are betwixt and between former diametrical opposites—such as person versus thing— evoking reflection and negotiation processes about the nature of the object but also about ourselves and human identity. Here, social responses are potentially due to the fact that machines are a kind of social actors—albeit different ones than human social actors. For this succinct summary of the two paradigms we were inspired by the observation that articles referring to the CASA framework—which is the most often employed framework to guide HMC research (e.g., Fortunati & Edwards, 2021; Gambino et al., 2020; Spence, 2019)—use the acronym both for “computers as social actors” as well as for “computers are social actors,” usually without addressing this difference.1 In our interpretation, the main notions of the Media Equation paradigm imply that humans respond to computers or machines as social actors, whereas in the Media Evocation paradigm machines are (some type of) social actors. The interchangeable use of “as” and “are” in the CASA acronym could be interpreted as an indicator of the fact that thus far HMC research lacked an explicit differentiation between the Media Equation and Media Evocation paradigms. The current paper aims to present these two paradigms by showcasing their main propositions, associated research methodologies, and current challenges. Herein, we rely on classical works (particularly Nass et al., 1994; Nass et al., 1993; Reeves & Nass, 1996; Turkle, 2005, orig. 1984), as well as recent theoretical and empirical publications within the HMC field. We argue that these classical works have been visionary in drawing our attention to the huge importance of computers in our lives and to the intriguing observation that humans respond socially to these even though they know that they are not communicating with a human. At the same time, HMC researchers have pointed out that the technical developments that now enable us to interact with AI-enabled communicators such as social robots, chatbots, voice assistants that can talk with us, know our name, distinguish our voice, and learn our preferences make it pressing for the HMC research community to revisit our theorizing and decide how to move forward (e.g., Fortunati & Edwards, 2020, 2021; Fox & Gambino, 2021; Gambino et al., 2020; Guzman, 2018; Guzman & Lewis, 2020; Lombard & Xu, 2021; Spence, 2019). In our conclusions, we note that particularly the Media Equation paradigm gained much momentum in the HMC field, focusing on the notion that people mindlessly respond socially toward machines and typically involving quantitative, experimental research methods. However, current interactions with AI-enabled communicators make it pivotal for HMC scholars to also focus on the Media Evocation paradigm, which emphasizes reflections and negotiations regarding the ontological boundaries concerning (among others) humans and machines. In our conclusions, we also make a plea for more qualitative research, and mixed-method research that includes innovative data analyses and that takes ontological

Related work. Theoretical Notions The first paradigm originates in the book The media equation: How people treat computers, television, and new media like real people and places, in which Reeves and Nass (1996) reported that they found that individuals’ interactions with computers and new media are fundamentally social and natural. They concluded that “media equal real life” and posed that this applies to everyone, applies often, and is highly consequential (p. 5). They emphasized that people have these social responses even though people believe these are not reasonable, and even though they do not think these responses characterize themselves (p. 7). Nass et al. (1993, p. 111) wrote that “[u]sers can be induced to behave as if computers were human, even though users know that the machines do not actually possess ‘selves’ or human motivations,” and Kim and Sundar (2012, p. 241) posited that “[e]verybody knows that a personal computer is not human. [. . .] Yet, we respond to it socially.” Thus, computers are conceptualized from the perspective of what they are not, that is, they are not human. Therefore, in the CASA acronym, the adverb “as” (i.e., computers as social actors) seems most appropriate for the Media Equation paradigm. Within this perspective, the paradoxical situation that users constantly exhibit social responses toward computers, while consciously being aware that this behavior may be inappropriate when exhibited toward nonhuman entities, naturally led to a focus on mindless attribution processes (Fortunati & Edwards, 2021, p. 13). People treat computers in some of the ways they treat humans by mindlessly applying to them the same social scripts they use in human-human interactions (Nass & Moon, 2000). Reeves and Nass (1996, p. 252) wrote that these automatic responses can be initiated with “minimal cues” (p. 253). Machines with limited cues—for instance words as output—induce individuals to employ human-oriented decision rules that they believe are inappropriate for assessing machine behavior (Nass et al., 1993, p. 111). Reeves and Nass’s (1996) explanation for this phenomenon that people are not evolved to 20th-century technology, and that modern media engage old brains (p. 12). That is, people respond to simulations of social actors and natural objects as if they were in fact social and natural: “absent a significant warning that we’ve been fooled, our old brains hold sway and we accept media as real people and places” (p. 12). The notion of “ethopoeia” (a direct HMC Theorizing Taking these Media Equation notions as a starting point, recent theoretical contributions focused on how to conceptualize and study humans’ interactions with AI-enabled communication technologies such as social robots, chatbots, and virtual agents.

Method. In contradiction to the propositions in the Media Equation paradigm, it is not the user’s behavior, but instead, the machine’s state that is paradoxical. Turkle (2005, p. 326) draws on Turner’s (1969) work on liminal objects and Douglas’s (1966) observations about marginality, referring to ambiguous states, or disorientation of individuals or groups (e.g., adolescents): a threshold state in which they find themselves after disengaging from the prevailing social order or pattern. Similar to those individuals or groups, machines possess neither properties of their previous state (e.g., “thing”) nor those of the future one (e.g., “hybrid” or “subject”), inducing reflections and negotiations about these categories and the object itself. Thus, social behavior toward machines is part of a mindful process of reflection, which involves negotiations concerning the nature of the machine, the user, and their relationship. Instead of being fooled by their old brains, users are being inspired by the paradox of the machine to re-think their schemes and paradigms. Hence, computers bring philosophy down to earth, by raising questions about the machine’s “life” and “mind,” and then by extension, making us wonder what is special about our own (Turkle, 2005, p. 2). Computational objects, poised between the world of the animate and inanimate, being at the same time a thing and a subject, alive and not alive, a physical object and an abstract idea, are experienced as both part of the self and of the external world, evoking questions about life, mind, and human identity.

Discussion. We envision that making this differentiation more explicit in HMC research deepens our understanding of why humans react socially toward machines and the consequences this has. Obviously, it is important to continue studying the impact of social cues on social responses, whereas at the same time it is vital to investigate machines in their roles as “philosophical provocateurs” and describe how users negotiate and reflect on their own identities and those of the machines they are interacting with. It is called for to study how the hybrid “betwixt and between” nature of machines makes the boundaries between humans and machines less clear. So far, the Media Equation paradigm and its emphasis on mindless processes (e.g., Nass & Moon, 2000; Reeves & Nass, 1996) gained a lot of momentum in the HMC field. These works enabled HMC researchers to consider machines as serious social actors, although the machines were only conceptualized as as-if actors (i.e., they are not human). Importantly, also in the influential Media Equation publications, authors indicated that the “equation” does not apply to all social interactions with machines. Whereas these authors focused on “social attitudes and behaviors that are controlled by more primitive or automatic processes” and “[r]ules that are used frequently” (Nass & Moon, 2000, p. 99), they did acknowledge the existence of conscious, reflective processes (e.g., Nass & Moon, 2000, p. 99; Reeves & Nass, 1996, p. 9). However, since the Media Equation paradigm does not offer a theoretical framework for exploring these conscious reflections, HMC researchers can in addition turn to the work of Turkle (2005, 2007) and Langer (1992)—authors who were also mentioned by Reeves and Nass, and Nass and Moon. Langer’s (1992) views on mindlessness and mindfulness can provide a bridge to exploring the causes of social responses more holistically. She wrote that “[r]ather than relegating all social interaction to mindless behavioral scripts, I began exploring contextual factors that might shift conscious awareness from minimal structural cues to a more complete awareness of available information” (p. 290). She also emphasized situational factors alongside individual ones. Whereas mindless responses seem invariant, she pointed out that being aware of our behavior, and the factors causing it, makes us more flexible and better able to adapt to new situations (pp. 300–301)—which is very pertinent for our current-day interactions with artificial communicators. Recent work on dual processing will also help HMC researchers to study the interplay between such mindless and mindful processes (e.g., Koban & Banks, 2023; Złotowski et al., 2018). In addition, the inclusion of the Media Evocation paradigm (using Turkle’s 2005, 2007 work) will help to come to a more in-depth understanding of how machines are social actors, by exploring the “betwixt and between” nature of machines, thus moving on from the limited dichotomy of either “thing” or “person” (e.g., Etzrodt & Engesser, 2021; Gunkel, 2022). Recognizing and paying sufficient attention to both of the paradigms will deepen our understanding of how and in which situations Media Equation and Media Evocation processes are at play, thus gaining a more holistic understanding of our social responses to machines.

As said, the Media Equation paradigm gained a lot of momentum, and this paradigm relied almost exclusively on experimental study designs (Gambino et al., 2020). Indeed, a recent analysis of 132 HMC publications in 28 communication journals over the past decade found that quantitative studies accounted for almost half of all studies (48.5%), and that the most used type of data collection was an experiment (40.2%).

Conclusion. In this paper, we presented a distinction between two paradigms that are driving current HMC research. First, the Media Equation paradigm, which—in its seminal works—conceptualized machines as nonhuman beings that “fool” humans into mindless social responses, and that now focuses on how a variety of social cues leads to social responses through both mindless and mindful processes. The empirical studies within this paradigm mostly rely on experimental designs to test these effects. Second, the Media Evocation paradigm, which conceptualizes machines as objects that can evoke reflection or negotiation processes about, e.g., the ontological categories “who” and “what,” since they are “betwixt and between” these categories, and, depending on the situation, culture, or individual inclination are sometimes more one or more the other—resulting in more or less mindful social responses. Within this paradigm—to be able to understand users’ reflections and negotiations—qualitative analyses are more common.