Social robots: Things or agents?

As such, these machines are not only seen as having potential practical purposes, but also as models for human cognition (seen here as “embodied” and “social”).

I do so also because those who interact with robots do. As exemplified by a witnessed instance of a young child who, during her first encounter with a robot, explicitly compared her own arms and head to the robot’s, interactional participants themselves also recognize humanoid features of the robotic technology.

As I followed the design process of social robots in two university laboratories, I saw the efforts put in programming robots’ software, but I also witnessed the attention that went into assembling the visual and tactile appearances of these technologies. Even the meetings with in-house mechanical engineers or the moments when a robot’s readymade parts were delivered to the laboratory abounded in careful visual inspections and judicious touching directed at assessing the robot’s physical architecture.

Since this process of design and construction is meant to respond, at least in part, to the preschool visits, its contingencies do not only concern the work of the roboticists but also the classroom’s interactions between the children and their teachers.

Each of the robots can be mapped on a “project” (Lynch 1985: 53–80) as it participates in organizing laboratory practices within a temporal context. The appearance of the project’s unity and sequential organization is achieved through local production and in situ activities of obtaining funding, responding to grant cycles, writing up results, as well as designing and building physical instantiations of the robotic machine.

The robot is also equipped to display “songs.” When a song (e.g., “Nine Little Monkeys”) is played, the B screen shows cartoon illustrations that accompany the song. On those occasions, the user is asked to look at (but not touch) the screen. During the interaction reported below, the robot does not play any songs, so this feature will not be further discussed in the present text.

Since the F screen is also touch-sensitive, when the user touches it, the eyes move as if following the touching finger. During the interaction reported below, the participants do not touch the F screen, so this feature will not be further discussed in the present text.

This is something that roboticists interested in haptic design are generally reacting against, as it erases the complexity of human touch (reducing it to the touch of an inert glass-like surface).

On the relevance of spatial organization in the enactment of the robot’s agency, see also Alač et al. 2011.

The involvement of touch in interaction has been discussed by Aug Nishizaka in his research on activities at a Japanese midwife house. Nishizaka describes how touch features in a midwife’s work of palpating a pregnant woman’s abdomen to specifically focus on referential practices. In his 2007 paper, Nishizaka shows how the tactile reference to specific locations is shaped by the action sequences in which it is contingently embedded, while his 2009 paper (2011) discusses the accomplishment of reference through touch that is not accompanied by visual access. When the teacher in Line 3 touches the robot’s forearm, she, however, does not refer to a specific location of the robot’s body, nor does she tactilely differentiate the sensed location from any other section of the robot (except the screens, perhaps). That the teacher engages the robot’s arm (rather than the rest of its body) may simply have to do with the spatial configuration of the group at the specific moment in interaction: T1 engages with the robot’s forearm so that the child can access and touch the robot himself.

For this kind of engagement with the digital materiality, see Alač 2011.

This supposition is in accordance with the ideas implemented in the design of the machine. As mentioned earlier, the robot is designed with a goal to eventually recognize and provide its users with an individually targeted educational content.

That the teacher’s action implicates Irma in the orienting of the others toward the robot is also seen in Line 21 where Irma’s individual actions are highlighted so that they now participate in doing the orientation collaboratively. We see again that the teacher talks about Irma in the third person rather than addressing her directly (“Irma really likes he:r,”), which suggest that the utterance is once more designed for the overhearers (Goffman 1981), namely those located by the books. To them, it indicates that the direct engagements with the robot are the actions that need to be attended to. At this point of interaction, however, the teacher (who followed Brian) is located in the area of distance experience, so that, with this turn, she seems to also display her orientation toward the actions around the robot, despite her new location.

In my previous work (Alač et al. 2011), I highlighted the embeddedness of sociable robots in the laboratory setting. While the autonomy of sociable robots is imagined as a future possibility, they presently live in laboratories. Here, as I trace how the actions of the adults participate in achieving the robot’s social character, the prominence is given to teachers (rather than researchers). Nevertheless, it should not be forgotten that we are still in the laboratory. Not only does the robot encompass the laboratory voices in its design while its moves are coordinated with the researcher’s actions, the events featured here, as noted earlier, are a part of the iterative design cycle that researchers perform as they design the robot.

For example, between Irma and T2 in Line 24, or T1 and Brian in Line 22.

Despite the teacher’s positive assessment of her actions, the situation this time is also distinct from the one in Line 11 as Brian and the teacher are not located right next to her (and the robot) anymore. In Line 24, Irma starts to turn away.

See, for example, Tanaka et al. 2007.

It seems plausible to suppose that a person (instead of a thing) would not be touched as frequently.

The teachers may, for example, use the actions of others as models (Lines 9, 11, and 35)—turning those who display familiarity with the language game into associates co-participating in orienting attention toward the robot.

They enable the instruction to be performed “from the robot” where the technology is designed to be used by a single user (we also saw how, when Brian and Irma engage the robot at the same time, the design predisposes for that engagement to be done in parallel but not in concert). As an element of that individualized instruction (Alač in preparation), the roboticists and their educational consultants project the future development where the robot would not only “deliver” but also “collect information.” If children touch its screen, the robot would be in a position to collect and measure their performance on a learning task and, eventually, provide each child with a “targeted” treatment. The acquisition of vocabulary where words to be learned are framed in relationship to their reference to certain visual representations is an example of that kind of instruction.

Barthes talks about the smoothness produced through a perfect junction of the car’s components: “one keenly fingers the edges of the windows, one feels along the wide rubber grooves which link the back window to its metal surround.” (88).

Those gestures may also remind us that sociable robots are often imagined as primarily consumer goods. An example is Sony’s entertainment robot Aibo, modeled on a pet dog (and also adopted as a research platform, together with Sony’s QRIO robot). When in production (between 1999 and 2005), a consumer could open a catalogue of a luxury leather and clothing brand, Hermès, and on the glossy pages of the catalogue—displaying fur jackets, bags, jewelry, fragrances, cutlery and porcelain—find an advertisement for Aibo (e.g., Le Monde d’Hermès 2001, N 39, Volume II). Apparently, Hermès also produced bags specifically tailored for carrying the robot. Those bags, whose contours follow the shape of the robot’s body, allowed owners to exhibit the possession of the commodity (the bag, while being iconic to a medium-sized dog, is unmistakably distinct from a dog carrier, not to mention leashes and other pets’ accessories that a brand could produce), while functioning as an envelope that protects the precious object it transports.

Some of the previous versions of the robot discussed here could be also labeled as a touch-sensitive affective sociable robot even though they didn’t embed haptic technologies in the real sense of the word. In those versions, the robot—remotely controlled—would, for example, vocalize crying and retract if a child touched it aggressively.

For example, Tanaka et al. (2007) dealt with the problem of haptic interaction focused on a QRIO robot that—assisted by a human operator—responded to touch by giggling. The authors focused on how a single user touches the robot with a goal to show that over time children start to engage the robot as a peer. In other words, they aimed at indicating that the robot can be treated as a person.

It may also be mentioned that the teachers also explicitly attribute value to this material aspect of the robot. For example, after the robot’s first visits to the preschool, the roboticists’ requested feedback from the teachers. They responded by asking for the movements of the robot to be limited, at least initially. In interviews, they repeated that children should see the robot as a thing, similar to toys and educational materials encountered in the classroom.

For related accounts on how seemingly contradictory features can unproblematically coexist in practice, see Verran (2001), Mol (2002), Alač (2011). For a semantic account, see Faucconier and Turner (2002).