[This is a continuation of my (now numerous) reflections on the Uncanny Valley. You can read the previous seven installments here if you’re interested.]
Late last semester, as a part of my department’s semi-regular Philosophy Film Series, we screened Mechanical Love, a 2007 documentary about the ever-evolving relationships between humans and robots. The film focuses chiefly on two robot projects, both from Japan: (1) IRC’s (Intelligent Robots and Communication Laboratories) Geminoid Project, which manufactures very human-like androids, and (2) AIST’s (Advanced Science and Technology) project that created Paro the “therapeutic robot.” Both Paro and the Geminoids fall under the general category of “social robots,” which are designed to mimic human interactive social behaviors and which, the film suggests, might even come to replace other humans for our social relationship needs. I’ve discussed the Geminoid Project in this series already, so I want to focus more on Paro here. As you can see from the picture above, Paro takes the form of a plush baby seal and is programmed to respond (positively and negatively) to human touch and voice. Paro isn’t meant to actually “replace” a human being so much as it is meant to replace a non-human animal, like a pet. (For many people, of course, relationships with their pets already replace or mimic relationships with real human beings. More on that below.) If the depictions that we see in the film Mechanical Love are true, Paro is very good at doing what it was designed to do. So good, in fact, that Paro seems…. well, uncanny.
Interestingly, the design plan for Paro indicates that the roboticists behind it made a concerted effort to avoid the uncanny valley phenomenon that I’ve been discussing in this series so far. In the film, we’re told that Paro’s general size and weight is meant to approximate the feel of a small child. Also like a small child, Paro can’t speak, so it cant and isn’t meant to generate complex or sophisticated interactions, like a conversation partner. Paro needs its owner to “care” for it (by stroking and petting), respond to it when it “cries,” and also to “feed” it (by plugging it in periodically). That is to say, Paro is designed to be dependent in the way a small child is, and also to make its owner feel indispensable and necessary, like I imagine the parent of a small child feels. Unlike scientists at the Geminoid Project, however, the AIST designers didn’t want to produce a robot-version of a small child, but rather a robot-version of a non-human animal pet that also happens to be the size of a small child. No offense to all the parents out there, but I think AIST’s fundamental intuition is right, namely, that the similarities between our relationships with pets and our relationships with small children, even if not identical, is remarkably similar. The dependency of pets and small children on us fosters a particularly strong relational bond, if only because that dependency forces us to respond, and to be responsible, to them.
So, why not a small cat or dog? As it turns out, the original design of Paro was in the form of a small cat, but that form was rejected by almost all of its test “owners.” The reason for this, the Paro designers explain, is because most people are familiar enough with the look, feel, sound and behaviors of real cats to be immediately turned-off by the robot cat’s imperfect approximation. (That “imperfect approximation” is, as I have contended in this series, the root of the cause of the uncanny valley effect.) When the designers switched to a form that was equally cute but not nearly as familiar (the baby seal), Paro’s uncanny problems disappeared. In this way, I think Paro accomplishes the same effect as Speilberg’s creature E.T. and does so by many of the same strategies. Paro’s designers were correct that most people have no idea what the up-close-and-personal interactions with baby seals are like– and, with a few highly improbable exceptions, nobody knows what the up-close-and-personal interactions with extra-terrestrials are like– so choosing the form of an unfamiliar-animal or an unfamiliar-being to serve as the host for very-familiar interactions and behaviors gives designers quite a bit of latitude. What Paro’s creators needed to capture, like Speilberg, was the essence of what makes these creatures the sort that enable relationships. It turns out that the basic characteristics needed to manufacture that effect are simple and few: big eyes, small stature, minimal capacity for communication, some element of dependency, complete lack of malice or aggressiveness, and a rudimentary simulation of the basic biological functions (eating, sleeping, mortality*) by which we understand beings to be capable of Being. Oh, and not incidentally, they need to be skilled at fostering and maintaining that thing captured in the title of the documentary: “mechanical love.”
[*Paro, despite the fact that it’s a machine, is what we could analogically call “mortal.” Machines do die, as anyone with any kind of machine knows.]
In Mechanical Love, Paro is placed with an elderly woman in a nursing home named Vera, with the hopes of giving her enough social stimulation to maintain (if not improve) her diminishing cognitive capabilities. The story of Vera and Paro is, quite frankly, heart-wrenching. Vera immediately takes to her new robot companion, and she is obviously happier with Paro than without it. The other patients in her nursing home, as well as the nursing home staff, find Paro annoying and bothersome. They ridicule Vera for treating Paro as if it’s real. The staff suspects Vera’s attachment to Paro is crazy. But when we see Vera’s interactions with her fellow patients in the home, who appear far more “robotic” and “mechanical” (given their compromised social and cognitive faculties) than Vera or even Paro, it’s not difficult to surmise why Vera finds her relationship with her robot pet so meaningful. It’s what we humans might consider a “social relationship” stripped-down to its bare minimum essentials, but those essentials are undeniably present. Paro is, as it was designed to be, a “therapeutic robot” for Vera and the therapy that it is dispensing is something that uncannily simulates real human social interaction.
So, the question is: what difference does it make whether or not Paro is an actual non-human animal or a robot? Don’t our relationships with all “real” domesticated pets function in the same basic way for humans as Paro does? One of the really interesting, and for the most part unexamined, questions raised by Mechanical Love is the extent to which our relationships with domestic (non-human animal) pets says a lot about the liminal space between our relationships with other humans and our relationships with machines. Even as technology gets more and more characterized, defined and understood by the rules of “social” networking, most of us still find it difficult to project onto machines the human characteristics that we easily project onto pets. Still, we forget that our projections onto non-human animal pets are nonetheless projections and, as such, are projected no differently onto non-human animals than they are onto non-human animal-like robots.
But, of course, there is a difference that makes a difference. Non-human animal pets can SUFFER and DIE, while robots/machines can only “break” (or stop working). So, it seems like there must be something more significant that we’re capable of investing into our relationships with sentient beings than whatever it is that we invest into our relationship with machines, which we know are not sentient, however much we may pretend otherwise. The interesting phenomenon exhibited in Mechanical Love, in the relationship between Vera and Paro, is that “sentience”– or, rather, the lack thereof– is not really a dealbreaker when it comes to what makes for a meaningful, even therapeutic, social relationship. In fact, there are many sentient beings about which the ability to suffer and die is not in the least questionable (like, for example, Vera’s companions in the nursing home, or the staff of that home) who nevertheless not only fail to inspire and/or sustain the basic sentimental intuitions necessary for a meaningful relationship, but who may also undermine those same intuitions.
Is Vera’s attachment to and love for Paro both meaningful and therapeutic? Undoubtedly so. Is it “crazy”? Arguably not. What does that say about the prospect of “replacing” real sentient beings, like humans and non-human animals, with machines? I don’t know.
The fundamentally aleatory nature of our relationships with other human beings (and “domesticated” non-human animals) is, of course, what makes those relationships unique. But it’s also what makes them mutable, grossly unpredictable, constitutionally fragile and, as a consequence, quite often extremely painful. As anyone who has a pet– or a real human friend, for that matter– knows, the option to “program” one’s partner in those relationships such that the possibility of their under-performance might be circumvented is a very attractive option, even if only self-interestedly so. What’s interesting about Paro is that Paro remains uncanny, though not because it too-closely approximates the appearance of a human companion or a non-human animal companion, but rather because it too-closely approximates the very companionship we have with human and non-human animals.
Paro approximates those relationships in every way except their imperfect unpredictability… which leads one to wonder after our attachments to those imperfections. In many ways, this particular iteration of the uncanny reminds me of “God’s Song” (provocatively subtitled “That’s Why I Love Mankind”), penned by Randy Newman and performed best by the late, great Etta James. In that song, Etta James recounts, from the point of view of God, the kind of blind trust with which mankind believes in and depends upon its relationship with God. Here’s the song, if you don’t know it already:
What Paro is able to accomplish, which no human or non-human animal could, is in the end something truly divine, viz. the ability to constitute itself as a relationship-partner whose “mechanical love” is thoroughly reliable, predictable, determinate and unconditional. Would that it were so with the rest of us imperfect beings!
**NOTE: To those of you who have written to me concerned that you couldn’t access this blog recently, I’ll just say that I took it down for a few weeks. Why is not important. It’s back online now and I’m glad to have you here.**
Great post, Leigh. I think you're nailing the issue. This sort of behavior makes me feel relatively Levinasian: the Other's qualities need not include vulnerability-in-fact in order to demand a tender response.
I wonder what you would think of Charles Stross's novel _Saturn's Children_. (There, though, the robots are sentient so much of the real tension you're pointing to is missing.)
Also: have you published any of your stuff on th Uncanny Valley yet? I'm thinking of trying to work my paper on race, face, and implicit bias into something publishable beyond the blog, and I wonder if you've found good journals for this work.