In the current state-of-the-art, social robots are not thought to be bona fide agents or interactants. Yet people sometimes interact with them as if they were real agents. Using the example of the interaction between the English-speaking social robot Smooth and three Danish human speakers, Clark and Fischer (C&F) point out that each speaker engages with Smooth in different ways. One speaker initially treats Smooth as if it were a person but then ignores Smooth, like the other two speakers, as if it were an inanimate object. To explain these differences in interactive behaviors, C&F propose that people represent social robots as interactive depictions of agents, which thus allows them to treat those robots as both an inanimate artifact (base scene) and as an agent (depicted scene).

We argue that in proposing a cognitivist explanation of human interactive capacities, C&F miss out on some important resources for understanding embodied social interactions. Drawing from two strands of ecological psychology, we suggest that even if social robots are interactive depictions, people need not mentally represent them as such. Rather, people can directly engage with the opportunities for action or affordances that such robots/depictions offer to them (Gibson, Reference Gibson1979/2015). Furthermore, the affordances that social robots offer to human interactants are constrained by the sociocultural behavioral setting(s) within which their interactions occur (Barker, Reference Barker1968; Heft, Reference Heft2001). We propose that any account of human–robot interactions ought to take into consideration the affordances of social robots and the sociocultural settings in which those interactions occur.

Consider the case of Smooth. The Seamless huMan–robot interactiOn fOr THe support of elderly people (aka SMOOTH) is an autonomous responsive robot developed for the care of the elderly in a Danish-assisted care facility. As dehydration constitutes a problem among the elderly as people may lose their sense for thirst with increasing age, Smooth was designed to serve water, among other purposes, to the facility's residents and to encourage them to drink more (Fischer et al., Reference Fischer, Seibt, Rodogno, Rasmussen, Weiss, Bodenhagen and Kruger2020). Smooth is not merely a thing or a tool that residents use to satisfy their metabolic needs, rather, it was designed to afford residents the opportunity to engage in interactions not only with it but also with each other. While residents usually retreat to their rooms after mealtimes, the presence of Smooth allows for them to hang around after their meals in the common room to order a drink and to chat with each other. Smooth thus has an intentionally designed positive influence on the social lives of residents (Fischer et al., Reference Fischer, Seibt, Rodogno, Rasmussen, Weiss, Bodenhagen and Kruger2020).

What is it about Smooth that that invites engagement by residents? In its embodied detail, the very design of Smooth – its physical, kinematical, and functional features – elicits certain kinds of interactive behaviors from residents. Social robots like Smooth have a physical, human-like form, unlike digital tools such as Amazon's Alexa that have a disembodied presence, which thus shapes how they are perceived. Smooth has a “penguin-like” shape and has a tray-like surface attached to its back to serve drinks; it also is of a height (and speed) designed for encounters with persons using a wheelchair or a walker (see the Target Article for an image of Smooth). It has “eyes” and “ears,” which indicates that it can see, hear, and be talked to. All these features help contribute to the regulation of Smooth's interactions with residents: Its eye gaze, body orientation, and speed changes, for example, help signal its “intentions” to residents which then allows them to perceive and co-navigate the shared space with Smooth (Fischer et al., Reference Fischer, Seibt, Rodogno, Rasmussen, Weiss, Bodenhagen and Kruger2020). To reiterate, the embodied aspects of social robots invite interactive behaviors from people; we directly perceive the affordances of such robots without the need to mentally represent them (Gibson, Reference Gibson1979/2015).

Crucially, the affordances that Smooth offers to residents are constrained by the larger sociocultural setting within which their interactions occur. Smooth was designed to function in an elderly care facility, and its interactive capabilities need to be understood in the context of a nursing home (Fischer et al., Reference Fischer, Seibt, Rodogno, Rasmussen, Weiss, Bodenhagen and Kruger2020). If Smooth were to be placed in a different sociocultural setting, say at a restaurant or a public park, it may afford different kinds of behaviors to human interactants. At a restaurant, for example, Smooth may be initially welcomed by patrons as it serves them a drink but it might be later ignored as patrons engage with each other, which may explain the behaviors of the Danish speakers in C&F's example. Here, the patrons are not dependent on Smooth for drinking water or for social interactions as in the case of an elderly care facility. In the setting of a care facility, Smooth forms an integral part of the lives of residents and affords behaviors of a different kind from that in a restaurant. Elders come to expect certain kinds of behaviors from Smooth and thus may attribute greater agency to it, and Smooth affords certain kinds of behaviors that temper the expectations of the elderly. Caregivers, on the other hand, may view Smooth as a useful tool to ease their overburdened workload, whereas visitors, say grandchildren, may view Smooth as a plaything (Fischer et al., Reference Fischer, Seibt, Rodogno, Rasmussen, Weiss, Bodenhagen and Kruger2020). The differences in how human interactants engage with social robots can thus be explained by recourse to the larger sociocultural setting within which their interactions occur: Depending on their social roles within the assisted care facility, caregivers, residents, and visitors may engage with Smooth in different ways. The sociocultural setting structures the behaviors of interactants (Barker, Reference Barker1968; Heft, Reference Heft2001); there is no need to posit a complex, three-layered mental representation (of the base scene, the depiction proper, and the depicted agent) to explain how people interact with social robots/depictions.

In conclusion, the embodiment of social robots and the sociocultural settings within which interactions occur between humans and robots play an important role in shaping those interactions. In employing a cognitivist lens, C&F miss out on rich, embodied considerations when attempting to address the social artifact puzzle.