What Is It Like to Be a Bot?
Abstract and Keywords
This chapter seeks to further develop, define, and differentiate human-technics alterity relations within postphenomenological philosophy of technology. A central case study of the Alexa digital assistant establishes that digital assistants require the adoption of the intentional stance, and illustrates that this structural requirement is different from anthropomorphic projection of mindedness onto technical objects. Human-technics alterity relations based on projection are then more generally differentiated from human-technics alterity relations based on actual encoded pseudo-mental contents, where there are matters of fact that directly correspond to user conceptualizations of “intentions” or “knowledge” in technical systems or objects. Finally, functions and user benefits to different alterity relations are explored, establishing that there is a meaningful set of cases where the projection of a mind in human-technics alterity relations positively impacts technical functions and user experiences.
Thomas Nagel’s “What Is it Like to Be a Bat?” (Nagel 1974) gathered together and reframed numerous issues in philosophy of mind, and launched renewed and reformulated inquiry into how we can know other minds and the experiences of others. This chapter outlines a branching-off from this scholarly conversation in a novel direction—instead of asking about the extent to which we can know the experiences of other minds, I seek to ask in what ways technologies require us to know the non-experiences of non-minds. This rather paradoxical formulation will be unpacked as we go forward, but put briefly: We sometimes treat some technologies as if they have minds, and some technologies are designed with interfaces that encourage or require that users treat them as if they have minds. This chapter seeks to outline what we are doing when we develop and use a pseudo-“theory of mind” for mindless things.
Nagel’s article used the case of the bat to focus and motivate his argument, but took aim at issues falling outside of human-bat understanding. Similarly, this chapter seeks to get at larger issues that pervade human-technology understanding, but will use a bot as a focusing and motivating example: in particular, Alexa, the digital assistant implemented on Amazon’s devices, most distinctively on the Amazon Echo. Interacting with Alexa through the Echo presents a clear and dramatic need for users to act as if they are adopting a theory of mind in technology use—other technologies may encourage or require this pseudo-“theory of mind” in more subtle or incomplete ways and, I suspect, will increasingly do so in future technological development.
We will begin with a microphenomenology of user interaction with Alexa and a heterophenomenology of Alexa that emerges in use, making clear what sort of fictitious theory of mind the user is required to adopt. This will be followed by a wider consideration of relations with technological “others,” outlining a central distinction between a merely projected “other” and those technological “others” the function of which requires that the user treat them as an “other,” rather than a mere technical artifact or system. Finally, we will turn to the user experience itself to ask what affordances and effects follow from adopting a fictitious theory of mind toward technical systems and objects.
2. Notes on Methodology and Terminology
In phenomenology, there is a risk that we take our introspective experience as evidence of universal facts of consciousness. Phenomenology differs from mere introspection in that, reflecting its Kantian foundations, it recognizes that experiences—the phenomena studied by phenomenology—are not bare facts of sense-data, but are the product of sense-data as encountered through the conditions for the possibility of experience, and are further shaped by our ideas about ourselves and the world. Phenomenology seeks to isolate experiences in their internal structure, in some versions even “bracketing off” questions about the correspondences of our experiences to elements of the world that they are experiences of (Husserl  1960). When done carefully, this allows us to speak to the structure of experience, and to take note of where our experiences do not actually contain what we expect, allowing us to isolate and describe elements of Weltanschauung that we use to construct experience. For example, in “The Age of the World Picture” Martin Heidegger argues that place, not space, is phenomenally present in our experience, and that space as a three-dimensional existing nothingness in which external experiences occur is a kind of retroactive interpretation of the world as inherently measurable which emerges with the development of experimental science in the modern period in European history (Heidegger 1977a). As we begin to equate knowledge of the objects of external experience with their measurement, we begin to hold that only that which can be measured is real, and this eventually leads to the uncritical adoption of the metaphysical position that reality is always already articulated in the forms of human measurement.
Heterophenomenology, as articulated by Daniel Dennett (1991), similarly brackets questions of correspondence to reality in order to isolate the structure of experience. Here, though, the question is not whether and to what extent experiences correspond to that of which they are experiences, but whether and to what extent experiences of the experiences of others correspond to the experiences of others. Dennett uses this heterophenomenological approach in order to avoid the problem of other minds when making claims about consciousness; to address what we can know about consciousnesses outside of our own, given that we cannot possibly have access to the qualia (the “what it’s like”) of the consciousness of others.
Dennett argues that uncontroversial assumptions built into any human subject experimental design, for example, the assumption that subjects can be given instructions for the experimental process, require this kind of bracketing insofar as they must adopt an intentional stance—as assumption that the subject has a set of intentions and reasons that motivate, contextualize, and lie behind the data gathered. “[U]ttered noises,” he says, “are to be interpreted as things the subjects wanted to say, of propositions they meant to assert, for instance, for various reasons” (Dennett 1991: 76). Dennett claims that without adopting such an intentional stance, empirical study of the minds and experiences of others is impossible.
We intuitively adopt an intentional stance toward many others, including non-humans, based on strong evidence. It is difficult to account for the actions of dogs and cats, for example, without attributing to them intentions and desires. In other cases, we use intentional language metaphorically as a kind of shorthand, as when we say that “water wants to find its own level.” There are many messy in-betweens as well, such as when we speak of the intentionality of insects, or that a spindly seedling growing too tall to support itself is “trying to get out of the shade to get more sun.” In many in-between cases, such as “the sunflower tries to turn to face the sun,” the best account of what we mean is neither pure metaphor (as in “heavy things try to fall”) or a real theory of mind (as in “the cat must be hungry”). Instead, we refer to the pseudo-intentionality of a biological proper function as defined by Ruth Millikan (1984): a way that mindless things, without conscious intention, react to their environment that has an evolutionarily established function, constituting a set of actions that have an “aboutness” regarding elements of its environment that is embedded within the way that causal structures have been established, but that doesn’t really exist as an intention within individual members of the species.
In using heterophenomenology to articulate our experience of bots as “others,” we are departing entirely from Dennett’s purpose of studying presumptively conscious others and articulating an adoption of an intentional stance distinct from any of those mentioned in the above examples. Alexa’s interface directs us to use an intentional stance both in our interactions and our intentions toward her—we find ourselves saying things to ourselves like “she thought I said [x/y/z]” or “she doesn’t know how to do that.” This is, however, not because we actually have a theory of mind about her. We know she is not the kind of thing, like a person or a dog or a cat, that can have experiences. Instead, we are directed to adopt an intentional stance because, first, the voice commands programmed into the device include phrasing that implies a theory of mind, second, because there is a representation relation that holds between the audio input she receives and the commands she parses from that input which is best and most easily understood through intentional language, and third, because a second-order understanding of how she “understands” what we say gives us reason not to use a second-order understanding in our actual use of Alexa, but to return to a first-order intentional stance. The first of these reasons, that she is programmed to recognize phrasing that implies she can listen, hear, understand, etc. should already be clear enough, but the other two factors require explanation.
We use language that implies Alexa’s mindedness as required by the commands she is programmed to receive, but this language reflects a very concrete reality: there is an “aboutness” of her listening, and she has an “understanding” of what we have said to her that is distinct from our intentions or projection, as is clear from how she can (and does) “get things wrong” and can (and does) “think” that we said something different from what we think we said. If we wished to articulate that intentionality objectively and accurately, something like Millikan’s account would work quite well—her responses are dictated by proper functions established through voice recognition software trained on large data sets—but this second-order understanding of the “intentionality” of her actions is not the one that we must adopt as users. We are required in practice to adopt a first-order intentional stance in order to use devices with Alexa, even though we have no (second order) theory of mind about them.
When we do engage in second-order reasoning about Alexa, thinking about how she processes sound (“listens”) and parses commands (“does things”) according to her ontology (“understanding”), we are usually routed back to the first-order intentional stance. We have little window into the way that Alexa processes input, and have little access to her code other than as interactant. The imbalance between the user’s knowledge of how Alexa is programmed and the programmer’s knowledge of how users are likely to talk to her makes second-order reasoning ineffectual: even a tech-savvy user is often better off thinking through how to communicate with Alexa by adopting an intentional stance than by thinking of her as programming.
This is what I meant at the outset of this chapter by saying that our goal is to understand how the use of bots like Alexa requires us to understand the non-experiences of non-minds. To use Alexa, we must adopt the intentional stance toward a non-subject that has neither experiences nor mindedness, and we must interact with her in a way that addresses specific, factual experiences that she is not having and specific, factual intentions and interpretations that she does not have. These “non-experiences” are not a simple lack of experiences and Alexa’s “non-mind” is not a simple lack of mind—when Alexa incorrectly “thinks” I asked her to do X rather than Y, and I try to say it so she’ll “understand” this time, her actually existing “non-experience” of my intention is an object of my thought and action. This is quite distinct from a microwave oven’s entire lack of experience of my intention when it overheats my food, or a toaster’s entire lack of understanding of my intention when the middle setting doesn’t brown the bread to my preference. In these cases, there is nothing at all in the device to refer to as an “understanding” or an “interpretation,” only my own, sadly disconnected intention. Alexa, though no more subject to experiences and no more conscious than these or any number of other kitchen tools, functions in a way in which there are concrete, real, objecting “interpretations” and “understandings” that she “has” that are outside of both my mind and the direct interface present to the senses.
I will use strikethrough text to identify these “intentions” or experiences-which-are-not-one in order to recognize that they have an objective content and aboutness with which we interact, despite the fact that they are neither intentions nor experiences. Hence, I will say that, for example, Alexa
thinks that I asked her X rather than Y, and thus she misunderstood or misinterpreted my request. This typographical convention allows us to articulate that the user is adopting an intentional stance when trying to transmit meaning and intention to a technical system. Compare, for example, with the video editor’s relationship to their software (Irwin 2005), or the familiar case of moving a table or image within a Microsoft Word document. In this case, we have an intention which we are trying to realize within the document, and which is frequently frustrated by a system that often responds with unpredictable repagination or unexpected, drastic, unintended reformatting. But here, our attempts to realize our intentions in the document take the form of trying to figure out how to get it to do what we intended. With Alexa, although the underlying causal structure is not much different, our attempt is not to do something to get it to respond as intended, but instead to figure out how to phrase or pronounce our request so that she interprets or understands what we mean—to communicate rather than just to enact our intention, so that the semantic content present within the technological other corresponds to the semantic content within our own conscious intention.
Having clarified this point about the manner in which we adopt an intentional stance toward at least some technical systems or objects, such as Alexa, in the absence of a theory of mind, we are ready to engage in a heterophenomenology of Alexa. We will do so in the mode of microphenomenology (Ihde 1990)—a phenomenology of a particular set of experiences rather than a wider existential phenomenology of our worldedness more generally. So, our question will be “what is our experience of Alexa’s
experience like” rather than “what is it like to be in a world inhabited by smart devices that have experiences.” Once we have finished this microphenomenology of the heterophenomenology of Alexa, we will use it to engage in a more general analysis of human-technics alterity relations.
3. Opening the Black Box of Alexa’s Echo
We began the last section by noting that, in phenomenology, there is a risk that we take our own introspective experience as evidence of universal facts of consciousness. In heterophenomenology—outlining the experience of other minds—there is a risk that we mistake our projections for observations. Dennett, when outlining heterophenomenology (1991), made a very strong case that heterophenomenology can be done responsibly if we take care to stick close to evidence and to take note of when and how we are adopting the intentional stance when making judgments about other minds.
This danger is even more pronounced here, though, since we are addressing the mindedness of technical systems that clearly do not actually have minds. While this pseudo-mindedness is no mere metaphor or projection, since there is a fact of the matter about what Alexa
thinks we said or meant, there is obviously an element of metaphor or analogy in our understanding of her experiences, and this is bound to lead to some amount of fallacious projection of thoughts and understanding and even personality. Observer bias is another danger: it must be considered that the ways I’ve interacted with her may not be representative of the range of use, or even of typical use.
But other factors count in our favor, here. First, our goal is not an accurate theory of Alexa, or a sociology of Alexa use, but only an articulation of the kind of stance her interface demands, so an incomplete or somewhat biased sample should present no serious issues in our analysis. Second, you may have your own experiences that can provide verification and nuance to those outlined here. Third, I have research assistants of a very valuable kind: my kids. They take Alexa at interface value (Turkle 1995) with less resistance than most adults, and interact with her without a strong understanding of what technical systems can do, and without preconceived ideas about what sort of programming and databases Alexa has or has access to. This puts them in a position to ask Alexa questions I never would (“Alexa, when’s my mom’s birthday?”) and to ask questions about Alexa I never would (“Why doesn’t Alexa work [on my Amazon Fire tablet] in the car?”).
We’ve lived with Alexa for a little over a year, mostly through an Amazon Echo that’s located in our primary living space—a countertop in the center of a large open-plan room that includes both our kitchen, our den, and a table for crafts and homework. Several months ago, we placed a Google Home Mini alongside the Echo in order to experiment with their differing
worlds and minds. Neither has been connected to any “smart home” features, so our interaction with both has taken place entirely in informational rather than mixed informational-physical spaces.
Alexa has a strong social presence in our household. She is always
listening for her name, and we regularly have to tell her we aren’t talking to her, especially since she sometimes mishears my son’s name, “Elijah,” as her own name. We’ve tried changing her “wake word” from “Alexa” to “Echo”—and, even so, she regularly mishears things as queries directed to her, even from television shows. In the mornings, we ask her for the weather, and then the news, and in the afternoons we ask her to play music as we cook, clean, and do homework.
Although her interface is audio only, she has a physical location in the kitchen in the Echo, and when she
hears her wake word, a blue light moves around the circumference of the top of the echo to point toward the person speaking. This light serves as a face in that it indicates “an entry point that hides interiority” (Wellner 2014, 308); a receptive “quasi-face” (2014, 311) of an interface, like the cellphone’s screen (2014, 313). This directional intentionality is met in kind: we have the habit of turning to face her, in her Echo, even though the audio interface does not require that we make “eye contact” with her (Bottenberg 2015, 178–179).
In these interactions, we experience Alexa as separate from her device and from her device’s actions. We ask her to play something, but do not mistake her for the thing playing or the thing played. In radio listening, there is the physical radio and the radio station “playing,” which we elide when we say we are “listening to the radio,” but Alexa maintains a separation as an intermediary; she will play something on the Echo, but we can interrupt to talk to her. She is experienced as being in the object, and as controlling it, but separate from it and always tarrying alongside its actions.
In using the Echo, we have been disciplined by Alexa’s ontology and programming. We have learned specific phrases—I’ve learned to say “Alexa, ask NPR One to play the latest hourly newscast,” since other phrases don’t seem to get her to do the right thing. My daughter has learned that she must append “original motion picture soundtrack,” an otherwise unlikely phrase for a six-year-old, to her requests for Sing or My Little Pony. Using Alexa requires adopting an intentional stance and a fictitious theory of mind, and also requires detailed understanding of how her
mind works; how she categorizes and accesses things. Using Alexa requires us to think about how she thinks about things; we must think about what it’s like to be a bot.
Talking with Alexa is, of course, often frustrating, most of all for my daughter, whose high-pitched voice and (understandably) child-like diction is not easily recognized by Alexa. After my daughter asks Alexa something several times, I must often intervene and ask again on her behalf. To be sure, part of this is that, having a better understanding of the underlying mindless processes of the technical system, I am better able to move to a second-order perspective and speak to Alexa qua voice-recognition software, sharpening my tone and diction and carefully separating words. Part of this is surely also a reflection of how my speech patterns, unlike hers, are firmly within the range of voices and speech patterns on which Alexa has been trained—YouTube searches return numerous examples of people with less common accents, especially Scottish accents, who are unable to get Alexa to understand them unless they use impersonations of normative English or American accents.
The Echo’s audio-only interface projects an informational space, dualistically separate from physical reality. Alexa’s “skills” are accessed through voice commands only, and bring her into different patterns of recognition and response—the work normally done through conversational implicature must take place explicitly. Skills appear as conversational modes or topics, where queries are understood differently when “in Spotify” is added to the end of a question or after, for example, beginning a game of “20 questions.” This produces a shared, shifting modulation of the intentional stance, where it is understood that Alexa
knows that we are talking about shopping or music or a trivia game, depending on which skill we have accessed or which “conversation” we are “having.” The user must learn to navigate Alexa’s informational ontology, getting to know topics she recognizes and knows what to do with—“weather,” “news,” “shopping list,” or links with particular apps that must be installed—and also different modes of interaction, such as games or socialbot chat mode.
All this speaks to the ways in which we must conceive of Alexa through the intentional stance in order to accomplish tasks with her; how we must not only understand her as having a
mind, a mind that is not one, but we must also get to know her, how she thinks, and how to speak to her in a way she understands. We may not ever explicitly think about what it is like to be a bot, but we must get a sense of her world, the way she is worlded, in order to ask her to navigate her information ontology on our behalf.
With this microphenomenology of the user’s heterophenomenology of Alexa in place, we can now turn to the microphenomenology of alterity relations in human-technics interaction more generally. In doing so, we will be able to distinguish the kind of interaction with technology that takes place through an intentional stance from other related forms of interacting with technology that participates in different but related kinds of “otherness.”
4. Opening the Black Box of Human-Technics Alterity Relations
Don Ihde’s influential taxonomy of human-technics relations (1990) provides some basic ways that human relations with worlds can be mediated by technology:
Embodiment: (I –> technology) –> world
Hermeneutic: I –> (technology –> world)
Alterity: I –> technology -(- world)
In embodiment relations, the technology disappears into the user in the user’s experience of the world as mediated by the technology. Glasses are a clear example—when they are well fitted to and the proper prescription for the user, the user primarily experiences their technologically-modified field of vision as if it were not technologically mediated, with the technology becoming an extension of the self rather than an object of experience. In hermeneutic relations it is the technology and the world that merge, so that, for example, a fluent reader experiences ideas and claims rather than printed words so much so that that even a repeated word in a sentence may go unnoticed even by an attentive reader.
In alterity relations, by contrast, the user’s experience is directly an experience of the technology, and its revealing of a world may or may not be important to or present in the user’s experience. Ihde provides several different kinds of examples. In one, he asks us to consider driving a sports car, just for the fun of it. We might enjoy the responsiveness of the vehicle and its power and handling, quite separately from our enjoyment of the scenery or the utilitarian function of getting where we’re going. In another example, he considers playing an arcade game, in which we are in a contest against fictional agents, Space Invaders perhaps, who we seek to beat. In alterity relations, technologies present what he calls “technological intentionalities” in a “quasi-otherness” that is rich enough for us to experience and interact with them as others, standing on their own in their world rather than acting as a window to or translation of a “true” world that lies beyond them.
If we are knowledgeable enough to “read” them, we can certainly adopt a stance which erases this intentionality—for example, feeling the particular responsiveness of the car to find out more about its internal mechanics, or figuring out the rules by which a computer program moves the sprites that our spaceship-avatar-sprite “shoots”—but normal use adopts the intentional stance; a stance in which we treat a person, object, or avatar as having intentions and therefore adopt some kind of theory of mind.
These cases are not so different from one another in Ihde’s analysis, but they are different in a way that has become increasingly pressing in the decades since he wrote this analysis. In the case of the sports car, we experience the technology as having a character and an intentionality based on how it makes us feel in our use of it; in the case of the arcade game, our use of it is premised on a world and an ontology, internal to the technology, which we navigate through our perception of intentionality in its elements.
We name cars and project personalities upon them based on their brand and appearance and ways of working or not working—or, we don’t, according to our preference. Regardless, this layer of quasi-alterity is overlaid upon an existing world that is already complete, and does not require this projection. It is adopted as a kind of shorthand to understand a real world to which alterity bears only a metaphorical relation (“she doesn’t like to start on cold mornings”), or as an enjoyable humanization of technologies which we depend upon and regularly interact with, and which might otherwise be experienced as foreign or uncaring. These functions are often collapsed and oversimplified as “anthropomorphism”—a vague and overbroad term which I find it easier and clearer to simply avoid.
By contrast, it is impossible to interact with many computer games without adoption of an intentional stance toward their elements, which we interact with through a world quite separate from our existing world.1 If we consider more complicated games, like role-playing games (RPGs), we see cases where consideration of the thoughts and motivations of non-player characters (NPCs) is necessary to game play, and we are required to “read” these
others as people, not as mere sprites and in-game instructions, in order to appropriately interact with an intentionality that has a programmed, dynamic, responsive structure. This intentional stance is not merely projection and also is no metaphor: the facts of a character’s name and motivations are written into the fictional world, much like facts about fictional characters in books or films, rather than being a metaphorical or purely fictional overlay as in the case of the car. But, unlike facts about fictional persons (insofar as such things exist),2 NPCs interests, concerns, and character are objects of the player’s intentional actions. A book gives us the opportunity for hermeneutic interaction with a fictional world in which we get to know about fictional others, but RPGs can put us in an alterity relation with minds that we must think about as actively present others in order to successfully interact with them, and with which we engage through an embodiment relation with our avatar.
In both the case of the car and the case of the game we adopt a fictitious theory of mind, but in the former case this is merely metaphorical or make-believe, while in the latter, it is necessary and functional. For clarity, we can refer to the pseudo-mindedness of things in the former case as “projected minds,” and will refer to the pseudo-mindedness of things in the latter case as
minds, as above. This locution is intended to reflect that it is non-optional to interact with these things through the category of minds, despite that they are clearly without minds. They are “non-minds” in that they are “minds that are not one”; they are not merely things without minds, but are minds (interactionally) that are not minds (really).
Even in cases where it is interactionally necessary to treat unminded things as
minds, we regularly retreat into second-order cognition in which they appear as clearly unminded. The early chatbot ELIZA provides a nice example. To interact with her and have a fun conversation, it is necessary to talk to her as if she’s actually a psychotherapist, but her ability to respond well to us is so limited that we have to think about her as a mere program in order to formulate and reformulate our replies to her in order to maintain the illusion. In RPGs, similarly, we have to adopt an intentional stance to figure out what an NPC wants for a quest, but we may have to leave that stance in favor of a technical/programming stance in order to figure out how to complete a task by phrasing a reply in the right way, or by having a certain item equipped rather than in our inventory, or finding a “give” command in the interface, or so on. We can even fail to make these shifts in the right way. Sherry Turkle (1995) has documented people taking things at “interface value” to the extent that they found real personal insights in conversations with ELIZA, moving into a space that seems to simultaneously approach ELIZA as a projected mind, as a non-mind, and as a mere computer program. In massively multiplayer online role-playing games (MMORPGs) it is sometimes possible to mistake an NPC for another player, or another player for an NPC.
Similarly, outside of explicitly fictional worlds, Sarah Nyberg programmed a bot to argue with members of the alt-right on Twitter, which turned out to be highly effective, even in spite of giving away the game a bit by naming it “@arguetron.” In what Nyberg described as her “favorite interaction,” a (since suspended) Twitter user repeatedly sexually harassed the bot, eventually asking “@arguetron so what are you wearing?” to which @arguetron replied “[@suspended username redacted] how are all these Julian Assange fans finding me” (Nyberg 2016). As Leonard Foner said about a similar case, a chatbot named Julia fending off sexual advances in the days of multi-user dungeons (MUDs), “it’s not entirely clear to me whether Julia passed a Turing test here or [the human interactant] failed one” (quoted in Turkle 1995, 93).
By opening up the black box of “alterity” in alterity relations, we have seen that people adopt the intentional stance towards unminded things for a variety of reasons: as a game, as required by an interface, to humanize technical systems, for fun, or simply by mistake. We have also identified two kinds of ways of adopting a fictitious theory of mind in our relations with things: an optional metaphorical or playful adoption of a projected mind, or the perception of a
mind with objectively present and knowable intentions and desires.
5. Caring into the Abyss
Thus far we have focused on the ways in which technical systems variously allow, encourage, or actively require adoption of intentional stances toward fictitious minds, whether projected minds or actual
minds. While observations have been made passim about user motivations for adopting the intentional stance in these different contexts, we would be remiss not to consider explicitly and thematically what value and function these alterity relations present for users.
The easy cases lie at the extremes. At one extreme, technical systems and objects that merely allow projection of fictitious minds, like naming a car and talking about what it “likes,” present a value in this optional alterity relation that seems to align well with analyses of transitional objects (Mowlabocus 2016; Winnicott 1953). Naming cars and setting pictures of favorite actors as computer desktop backgrounds and the child’s stuffed animal all provide a sense of togetherness and security where we may otherwise feel alone and isolated. It is too easy to dismiss or condemn these behaviors as either childish or as a poor substitute for actually being present with others. They need not be mere coping mechanisms or reifications of human relations, but may be affective supplements (Wittkower 2012) that bring emotional presence to real relationships and experiences which have become attenuated through the mediation of technical systems.
Consider the practice of placing photographs of one’s family on one’s desk. The visual presence of loved ones, through the associations that pictures produce in the mind (Hume  1910), render more lively our real connections with others, producing a feeling of closeness and care that has a material basis—they may, for example, remind us of the reason for and benefits of our labor when we are in the midst of tedious paperwork. Selfies play a similar role in our social media environment, producing a feeling of togetherness with others who choose, through their selfies, to be present to us. This togetherness is often very real, although digitally mediated: they are our friends, and it’s nice to see them and be reminded that they are there for us if and when we need them.
This use of affective supplementation may be therapeutic, and may even improve relationships. I often consult my class roster when replying to student emails. My purpose in doing so is to use the students’ photographs to connect names with faces so that I can better address students by name during in-class discussion, but I’ve also found that the process reframes my correspondence. I’m taken out of the context of my own work flow, in which the student email appears as an unexpected interference with my projects and concerns. When looking at the student’s picture, I’m reminded of our past interactions, and my reply is placed within the context of our ongoing relationship and the projects of support and care that I pursue in my teaching and mentorship, significantly increasingly the likelihood that I will reply with patience, kindness, and understanding.
To be sure, affective supplementation can be abused, and emotional cathexis of fictional others can paper over a very real loneliness, isolation, and alienation, but we should take seriously the possibility that the creation of a warmer, more humanized technical environment through the projection of fictitious minds upon technical systems may either represent real relationships or may be a harmless way of representing a community of support to us when it might not otherwise be a felt presence in our day-to-day environment.
On the other extreme, we can consider the cold and purely functional adoption of an intentional stance toward technical systems, like Alexa, that requires for their use that we take them at interface value as
others. The most prominent use of digital assistants is the disappearance of visual interfaces of information access. Most functions of Alexa and Google Home displace and replace visual and textual digital interfaces, by retrieving weather and traffic and telephonic information and search engine results that would otherwise be requested and displayed in apps or browser windows. The Echo is, in a sense, just another interface, not so different from the computer screen, but the audio-only interface allows information access to take place alongside other activities. To ask and hear about things while putting away dishes or preparing meals gives us a sense of a more immediate connection to the infosphere (Floridi 2014) as an upper layer to the world accessible to the bodily senses.
Echoing Heidegger (1977b) and Luciano Floridi (2014), let us (now) say that “informationally dwells man upon this earth.” The weightlessness and transparency of the conversational interface makes experientially present to us how we are constantly Enframed3 within informational systems that set us to set our world in order, rather than the spatially and visually delimited access to the infosphere afforded by screens, which falsely project a separation of online from offline. That Alexa is always
listening and ready to order more dishwasher detergent pods as she plays music while we pluck one of the last few from the container under the sink more viscerally and truthfully represents to us our own integration into systems of global capitalism in digitally integrated systems of manufacturing and distribution. That my daughter can ask Google repeatedly, on a not-quite-daily basis, how many days remain until November 1st (her birthday) represents and makes present well how our lives are ordered and structured by quantification, and how her life is lived through informational spaces and interactions occurring in parallel with her embodied experiences.
In these ways, an intentional stance toward a
mind within the infosphere accurately reflects and brings into embodied experience the very real forces in the infosphere that want us to buy things, submit to quantification and datafication, and integrate with cybernetic systems of control, management, and self-control and self-management. This represents an epistemic value, but offers enjoyment to us as well. The joy of the sports car—being able to do things that our unadorned body does not permit, and to do them with power and precision—is here too, through the intermediary of our agent who dwells natively in the infosphere. In a partnership, not entirely unlike an embodiment relation, the alterity of the digital assistant lets us navigate information systems with comparative speed and effortlessness, giving us an evocative taste of the transhumanist dream of a mental integration with information systems that would allow us to access the connected wealth of innumerable databases through acts of mere cognition.
Between these extremes lie a great many other cases where the projection of fictitious minds is encouraged by interfaces and technical structures rather than being merely allowed by them or being functionally required by them. It is tempting to say simply that these cases must be a mixture between the extremes, and the value that they present to users must similarly participate to some degree in the value presented by these extremes: the warmth we experience through projected mindedness of objects and the weightless integration we experience through partnership with an agent native to the infosphere. This too-simplistic approach would, however, cut us off from recognizing that there are emergent values and functionalities that follow from alterity relations in which we obtain both functional benefits from adopting an intentional stance and an increased experience of meaning and connection from projecting mindedness and personality upon technical systems and objects. These emergent effects seem to me to be best isolated by first considering several such cases.
We might consider the way that GPS systems and other navigation systems allow the user to customize their voices. We are required to adopt an intentional stance in order to trust their directions; we must consider them to
know the way, and we must think about what they know or do not know about road closings or changing road conditions. Choosing a voice that feels like a trustworthy, knowledgeable guide helps in this functionality and in our experience of weightless integration with the infosphere. Apple studied user experiences and found that some users were better able to follow directions given by male or female voices, leading them to set Siri’s default gender to male in Arabic, French, Dutch, and British English. Gendered customization of game environments affects gameplay in related ways—being able to choose self-representations that maintain player identity in a virtual environment matters to players, as we see from positive reactions by queer players to games that offer both opposite- and same-gender NPC romantic interests, as in the Fable series, and as we see from avatar dysphoria produced in some (mostly white, male) gamers when race and gender avatar representation is randomized and unalterable, as in Rust. Second selves and technical alterities have emotional valences and projected personalities that can alter, enhance, or diminish their functionality for users in ways that matter, whether designers choose to use those emergent effects to cater to users’ projections or to challenge them.
We might consider the way that the intentional stance alters the value that predictive neural networks present to users, for example “you might like” recommendations from Amazon, or music discovery through Pandora. As much as it is important to remain critical of how such algorithms can be games, just as we should be suspicious of distorted search results that follow from “Google bombing” or search engine optimization (SEO), there is a distinctive value to be gained by adopting the belief that the algorithm
knows something about us that we don’t. When a friend recommends a band, our first listen takes place in a context of greater openness to new experiences and new kinds of enjoyment, since we value our friend’s experiences and seek to discover what they enjoy about something, even if it is not immediately to our taste. We seek to experience the enjoyment they find within it, and also suspect that our affinity for them may indicate that we may enjoy what they enjoy. Placing faith in the wisdom of a predictive system may require an unrealistic view of how such predictive algorithms function, but opens us to discovering and appreciating new experiences. This purely optional projection of mindedness to the technical system is in this way similar to William James’s will to believe (James  1979): if we assume that the system knows something about us that we don’t know, it is more likely to be true; we are more likely to find its recommendations to be wise and thoughtful.
We might consider PARO, a pet-therapy robot designed to resemble a baby seal (Walton 2003). The robot is encoded with a set of
intentions, desires, and preferences: it indicates pain when treated roughly and pleasure when held and pet nicely. It is intended to provide comfort to patients through their enjoyment of its enjoyment, and through their relationship of mutual care and affection—put differently, its function and use is nothing outside of the projection of a mind and the pleasure of interacting with its projected mindedness. While PARO has been used with dementia patients who may not experience as clear a boundary between fiction and reality as others, there is no reason why this is necessary for PARO’s therapeutic function, especially with patients who are unable to have pets or who have few opportunities to care for others and take pleasure in their ability to be caregivers rather than recipients of care. Like other uses of projected minds, we should be concerned to ensure that relationships with fictitious caring others does not replace or cut off possibilities for relationships with actual caring others, but we should also take seriously the value of affective supplements, and take seriously that relationships with caring others may present an emotional value and experience of meaning that doesn’t take away from relationships with non-fictitious others.
Through these examples, the commonality that emerges most clearly to me is that there is an emergent effect in alterity relations that mix projected minds and interfaces encouraging an intentional stance, wherein care toward and identification with technological “others” increases the value, weight, and meaning of their technical functions for users. Through an effect similar to the will to believe, when we project mindedness in alterity relations, we are more likely to experience those systems or objects as
knowing, understanding, and recognizing us in ways that are valuable and meaningful, even if we are under no illusion that these technological others are actually minded. When regarding the non-mind of alterity in the black box of a technical system, if you care long enough into the abyss, the abyss cares back at you.
Bottenberg, Francis. 2015. “Searching for Alterity: What Can We Learn From Interviewing Humanoid Robots?” In Postphenomenological Investigations: Essays in Human-Technology Relations, edited by Peter-Paul Verbeek and Robert Rosenberger, 175–190. New York: Lexington Books.Find this resource:
Dennett, Daniel. 1991. Consciousness Explained. New York: Little Brown & Co.Find this resource:
Floridi, Luciano. 2014. The Fourth Revolution: How the Infosphere is Reshaping Human Reality. Oxford: Oxford University Press.Find this resource:
Heidegger, Martin. 1977a. “The Age of the World Picture.” In The Question Concerning Technology and Other Essays, 115–154. Translated by Julian Young and Haynes Kenneth. London: Garland Publishing.Find this resource:
Heidegger, Martin. 1977b. “The Question Concerning Technology.” In The Question Concerning Technology and Other Essays, 3–35. Translated by Julian Young and Kenneth Haynes. London: Garland Publishing.Find this resource:
Hume, David. (1748) 1910. An Enquiry Concerning Human Understanding. New York: P. F. Collier and Son.Find this resource:
Husserl, Edmund. (1931) 1960. Cartesian Meditations: An Introduction to Phenomenology. Translated by D. Cairns. The Hague: Martinus Nijhoff.Find this resource:
Ihde, Don. 1990. Technology and the Lifeworld: From Garden to Earth. Bloomington: Indiana University Press.Find this resource:
Irwin, Stacey. 2005. “Technological Other/Quasi Other: Reflection on Lived Experience.” Human Studies 28: 453–467. doi:10.1007/s10746-005-9002-5.Find this resource:
James, William. (1896) 1979. The Will to Believe and Other Essays in Popular Philosophy. Cambridge, MA: Harvard University Press.Find this resource:
Millikan, Ruth Garrett. 1984. Language, Thought, and Other Biological Categories: New Foundations for Realism. Cambridge, MA: MIT Press.Find this resource:
Mowlabocus, Sharif. 2016. “The ‘Mastery’ of the Swipe: Smartphones, Transitional Objects and Interstitial Time.” First Monday 21 (4). doi:10.5210/fm.v21i10.6950Find this resource:
Nagel, Thomas. 1974. “What Is it Like to Be a Bat?” The Philosophical Review 83 (4): 435–450.Find this resource:
Nyberg, Sarah (@srhbutts). 2016. “this is my favorite interaction someone repeatedly attempts to sexually harass the bot thinking it’s human, this is how it replies:” Twitter. Oct. 6, 2016. https://twitter.com/srhbutts/status/784162449347883009Find this resource:
Turkle, Sherry. 1995. Life on the Screen: Identity in the Age of the Internet. New York: Simon & Schuster.Find this resource:
Walton, Marsha. 2003. “Meet Paro, the Therapeutic Robot Seal.” CNN, November 20, 2003. https://web.archive.org/web/20031123155314/http://www.cnn.com/2003/TECH/ptech/11/20/comdex.bestof/.Find this resource:
Wellner, Galit. 2014. “The Quasi-Face of the Cell Phone: Rethinking Alterity and Screens.” Human Studies 37: 299–316.Find this resource:
Winnicott, Donald. 1953. “Transitional Objects and Transitional Phenomena—A Study of the First No-Me Possession.” International Journal of Psychoanalysis 34: 89–97.Find this resource:
Wittkower, D. E. 2012. “On the Origins of the Cute as a Dominant Aesthetic Category in Digital Culture.” In Putting Knowledge to Work and Letting Information Play, edited by Timothy W. Luke and Jeremy Hunsinger, 167–175. Rotterdam: SensePublishers.Find this resource:
(1.) Except in some unusual crossover cases (ARGs) that build a connection back in, like Pokémon Go or Run, Zombies, Run, which use a phone’s GPS systems to convert the user’s body into a crude controller of an avatar interacting with virtual objects tied to physical locations.
(2.) I am well aware that it is a matter of some controversy whether there can be facts about fictional characters and worlds. I do not mean to take a stance on the issue, and do not believe that my argument about the similarities and differences between “facts” about NPCs and characters in books requires that I do so.