Playtesting Philosophy:

Welcome to The Talos Principle, independent game developer Croteam's 2014 first-person puzzle-adventure video game. You, the player, are a sentient artificial intelligence, navigating a simulation of the ruins of humanity's greatest civilizations. Your task, as presented to you by ELOHIM, the authoritative voice of a programmed god, is to collect "sigils" by solving puzzles found throughout the ruins by utilizing various tools. The sigils are pieces of tetris-like puzzles, the solving of which grants you access to new tools and areas, wherein are more puzzles and sigils. Ultimately, once you've collected enough sigils, you may do as ELOHIM instructs and enter literal pearly gates, climb the tower he has forbidden you, or be entombed within the simulation to help future AI versions. If all you've done in-game is collect the required sigils, you may never know of the third option, or care about the second. But the game-world itself is a puzzle, an explicitly existential and ethical puzzle within which the conventional puzzles are couched. Play occurs as much in your head as much as it does through your keyboard input.

The sigil-collecting levels and tetris puzzles are the elements of gameplay by which Talos is most immediately recognizable as a puzzle game, generically defined by Henry Jenkins as

Games in which the primary conflict is not so much between the player-character and other characters, but rather the figuring out of a solution, which often involves solving enigmas, navigation, learning how to use different tools, and the manipulating or reconfiguring of objects. Most often there is a visual or sonic element to the puzzles as well, or at least some verbal description of them. (Jenkins 2002)

It is, quite literally, textbook: the progression-related puzzles in Talos require navigation, tool use, and reconfiguration of objects, and both sonic and visual clues inform players of how to navigate the puzzle-space. However, it is in fulfilling Jenkins's vague inclusion of "solving enigmas" that Talos presents a wholly different sort of puzzle to its players. In the introductory gameplay video are shown the linguistic forms of narrative encountered in Talos: Alexandra's holographic audio journal entries, ELOHIM's disembodied pronouncements, QR code graffiti left by previous artificial intelligence versions, and the computer terminals, which house both the found texts of the library archives and interactive dialogues with Milton Library Assistant (Milton). They do not address the puzzles in a tutorial manner, if at all, but rather the meaning of the puzzles, of the simulation-world. As with the tetris pieces, you as the AI piece together a jigsaw puzzle of proffered information snippets, trying to figure out who, what, and where you are, and why both you and the simulation exist. You as the player, in answering these questions—asked both implicitly and explicitly in the game—are asked to consider what it means, to you, to be a person, and the host of other philosophical puzzles the question entails.

SPOILER ALERT: Of the game's three endings, only one is truly an ending: that reached upon defying ELOHIM and ascending the tower, thereby fulfilling the simulation's purpose—to create an AI approximating humanity, including free will and curiosity—ensuring its destruction, and walking out of the simulation into the real world. Winning in Talos is outlasting the game, breaking down the boundary between simulation and reality and bringing knowledge attained in the former into the latter. Talos uses the medium of the video game to fashion players as conscious ethical agents, and uses philosophical, literary, and popular texts in order to enable, side-by-side with more conventional gameplay, the playtesting of morals. As you, the AI, exit the game world a questioning and knowing being, Talos is designed to ensure that you, the human being, do the same.

Before diving into how The Talos Principle engenders philosophical and ethical inquiry in-game and beyond, we need to address the player undergoing this experience. Up to this point, I've used "you as the AI," "you as the player," and "you as the human being" to describe the multiplicity of selves that one simultaneously inhabits during gameplay. As video game theorist Miguel Sicart writes in The Ethics of Computer Games, "[t]he player does not exist before playing a game" (72). He offers a "ludic hermeneutic circle" model for understanding "the process that takes place when an embodied, cultural human being becomes a player, and how that player relates to her subjectivity, the game experience, and the subject external to the game" (117). "You as the human being" is Sicart's "subject external to the game," who brings to the game experience all of her pre-existing knowledge, preferences, culturally imparted values, existential conditions (118). You bring all of yourself to your interaction with the game, both consciously and unconsciously. From this self interacting with the game comes "you as the player," or Sicart's "player-subject," who experiences "a dialogue between the system that imposes restrictions and affords behaviors [the game], and a player who reflects upon those" and who ethically interprets "the strategies and choices made" in play (119). "You as the AI," inhabiting the role of the character as whom you act in the game, is a subset of the player subject, but I consider it important to emphasize that the ongoing dialectic between you as external subject and you as player subject is achieved through inhabitation of the AI player-character. There is born a fused persona from you as playing agent and as the designed character through whom you experience the game world, and by whom (experientially) your interpretational activity is structured and limited. You know that you, the player subject, are making choices from pre-scripted options, and the choices available will often not reflect mimetically those you as the external subject would make; this inherent limitation of design is only non-disruptive of immersion in the game-world if you recognize the character as a "self," however fictional, with whom you are in cooperation. You realize yourself as the player subject through the player-character in a first-person game by fitting your own non-diegetic experiences to the character's pre-designed outline, thereby constructing a new subject.

The player subject is temporally bounded within a particular instance of play, but engages in play as the ongoing "Individual Player," who goes into and comes away from particular instances forming a generalized concept of how to behave in play based upon prior experiences; the player of games in general is the identity formed through these repeat experiences. Finally, the individual player, Sicart emphasizes, becomes a "community player," drawing upon cultural and ethical communally established values and testing her played experiences and behaviors against them. Through this circle arises the ongoing interpretive and self-constituting "dialogue between the player-subject and the moral being" (121). Sicart's model is primarily concerned with the interactive process by which games are ethical objects, and players ethical subjects, but I am primarily interested in the ongoing dialogue between external and player subject, and the mediator of the player-character.

I'm pushing towards, funnily enough, the same fundamental questions posed by The Talos Principle: who, and what, are you, and how do you know? Epistemology is central to the consideration of all of these questions, and a highly useful framework within which to approach my questions of how gameplay constitutes a discrete identity, and affects the player after ending play. Of the numerous branches of epistemology, I turn to epistemological constructivism—in particular, to radical constructivism, described by its founder Ernst von Glasersfeld in Radical Constructivism: A Way of Knowing and Learning as

an unconventional approach to the problem of knowledge and knowing. It starts from the assumption that knowledge, no matter how it is defined, is in the heads of persons, and that the thinking subject has no alternative but to construct what he or she knows on the basis of his or her own experience. What we make of experience constitutes the only world we consciously live in. It can be sorted into many kinds, such as things, self, others, and so on. But all kinds of experience are essentially subjective, and though I may find reasons to believe that my experience may not be unlike yours, I have no way of knowing that it is the same. The experience and interpretation of language are no exception. (1)

Heavily influenced by Kantian epistemology, constructivism does not deny the existence of external reality, but rather any possibility of accessing it; we are limited to "the only world we consciously live in." As von Glasersfeld's work is epistemological rather than ontological, he considers the subject as a knowing being, engaged in the ongoing process of constructing reality; I, however, am considering a dialectic process in which we consciously act and form experiential knowledge in two worlds, one in which we live, one in which we are players; one of which we accept as reality, the other we recognize as non-reality. Therefore, I will address the "thinking subject," the philosophical "I," as the dual entity I presented earlier—at once you as the player and you—engaged in dialectic construction with a third, assumed identity, that of the AI, which is explicitly a programmed iteration of von Glasersfeld's "thinking subject." The construction occurring is multilayered: as the AI, you are driven to construct an understanding of the world and reality in which you find yourself, but as the player, you construct the AI's knowledge, dictate the experiences by which it comes to that understanding, through your interpretation of the information presented in-game, based upon your interpretation of the world in which you live. The modes of interactivity and the range of actions comprising Talos's gameplay are designed to impart and allow for a specific set of experiences, each of which contributes to the construction of a particular thinking being, of a particular knowledge—and, that established, it is to gameplay that we may finally turn.

(mouse icon-overlay) Take Jammer; (mouse icon-overlay) Drop
Subject/object interaction .... OK.
Complex task management .... OK.
Child program basic calibration successful.
(jam mine) Spatial awareness .... OK.
Predictive capacity .... OK.
Child program logic check successful.
Checking sigils .... Done.
Removing child restrictions .... Done.
Recording data ....
Have a nice day.

These are the lines you see on-screen in the first stage of Talos gameplay, ¹ See first three minutes of gameplay video. during what functions as the tutorial. Unless you have opened the configuration menu on start-up, wherein you can see and change the keyboard input settings for movement and interaction, you begin the game without any explicit instructions as to how you as the player control the AI. Talos assumes previous familiarity with computer games, to an extent, in that the standard "WSAD" movement keys (forward/back/left/right) are set as default, presented without explanation. You know how to walk, and, upon "generation" in the game—that dizzy awakening wherein the AI gets its bearings, glimpses its hand, and after which you act from its perspective—you might intuitively move your mouse to turn your head, move your visual field; eventually, you'll walk around, see a shimmering purple gate, walk through it unimpeded. Immediately, typewriter-font blue text appears in the top-left of your screen: "Initiating child program logic check." Try to walk through the blue gate, and find it impassible; turn, explore, find a yellow plastic camera-esque object. You now see the only explicit form of non-diegetic instruction in the game: white, centered text in a different font, telling the player what an object is—a "Jammer"—and that you can pick it up or drop it, can interact with it by clicking the left mouse button. In case you didn't approach the Jammer on sight, blue text once again clues you in on what to do: "Subject/object interaction," and when you pick it up, "... OK." Talos hereby immediately presents itself as a game predicated on experiential learning rather than passive instruction. This is the case with nearly all contemporary video games—you don't read a manual, you play a tutorial. However, many games utilize some form of non-diegetic textual explanation, or diegetic dialogue, to introduce players to how the game is played, and what play entails. In Talos, however, the child-program—the nascent player—learns to function much like a human child.

Before presenting his own theories of radical constructivism, von Glasersfeld first turns to Jean Piaget's theories of child development, describing at length the process by which humans form the concept of the "object" during the sensorimotor stage:

1. Object concepts are established as "the infant coordinates (associates) sensory signals of the 'perceptual' kind that happen to be recurrently available at the same time in its sensory field." The "successful composition" of these concepts "may then serve as trigger for a specific activity that has been associated with the object. ... the child recognizes the object." (59)
2. The child demonstrates the "ability to run through a sequence of physical actions when the perceptual situation that originally led to the coordination of the sequence is not actually present," an activity that Piaget calls "deferred execution." When such activity is conceptual rather than physical, it produces a "re-presentation," or "re-construction from memory, of a past experience." (59)
3. The child attains the concept of "object permanence," wherein she "considers the object [she] is perceptually constructing at the moment, to be the identical (self-same) individual [she] experienced at some prior time." (60)

As development continues, the child comes to perceive the objects that she has constructed in relation to herself as external to herself, as objectively extant. The result is "assimilation," a process of "treating new material as an instance of something known" (62). These are the foundations of constructivism, and it is on these processes that Talos relies.

In Talos, just as Sicart noted that the player does not pre-exist gameplay, the character whose identity the player assumes in does not pre-exist her engagement with the game. This is, of course, an illusion, as all of the possible expressions of the AI are pre-programmed by Talos's designers. Rationally, as a player, you recognize this to be the case. But, in terms of your experience, the player and the AI come into being simultaneously. That the AI is guided through Piaget's stages of object-construction, as a newly generated entity, is perhaps unsurprising; what is more interesting is that the player, also new to this world, herself plays out this development in the tutorial, as follows:

1. You see a floating metal ball. As it approaches you, it opens up to radiate red laser beams, and begins beeping at an increased rate. If it comes too close, it explodes, and you re-materialize at the purple gate after a "rewind." If you pick up the (already-familiar) Jammer, you are given the option to "Jam" the ball, and after doing so can pass by safely.
2. You can recall the appearance, movement, and sounds of the floating metal ball, that it killed you on impact, that the Jammer effectively disarmed it, and that upon "death" you rematerialized at the purple gate.
3. Whenever you see these balls in future play, or hear their distinctive beeping, you know that you have to move away quickly, to look for a Jammer and jam it. You know that if it catches you, you will have to restart the puzzle at the purple gate.

This process continues throughout the game, as new tools and obstacles are presented, and those previously encountered continue to appear. In this manner, in terms of its mechanics, Talos becomes the staging-ground of constructivist play, of (seemingly) "free," in the sense of relatively unguided, exploration and trial and error. That the player is, from its outset, engaged in constructive activity might appear unimportant, but with the immediate establishment of gameplay as empirical construction, Talos lays the groundwork for play as abstract conceptual constructive activity—that is, the construction of beliefs.

As I noted earlier, objective reality as an independent, external entity is incompatible with constructivist theory. In its place von Glasersfeld proposes the "intersubjective" as the "highest, most reliable level of experiential reality," which comes through "the corroboration [by] other thinking and knowing subjects" of an individual's experiential reality (119). These others—as everything—are constructed, and the manner in which von Glasersfeld understands them to be so is an extension of a Kantian proposal: "If one conceives of another thinking subject, one necessarily imputes to that other the properties and capabilities by which one characterizes oneself as subject" (Kant qtd. in Glasersfeld 119). Though other people are—as anything experienced—external objects of our perception, they are objects whose actions and behaviors lead us to impute to them "the kinds of concepts, schemes, and rules one has oneself abstracted from experience," and thereby to come to consider them to be like ourselves (120). The corroboration by other thinking subjects of our own conceptual constructs helps us to distinguish "between knowledge that we want to trust as though it were objective, and constructs that we consider to be questionable if not downright illusory" (119, emphasis added).

Now, before dealing with that dichotomy in individual constructions of 'true' knowledge, I must acknowledge the intended applications and disciplinary allegiance of von Glasersfeld's radical constructivism: it is a branch of the philosophy of science and mathematics which von Glasersfeld explicitly created, ultimately, as a philosophical model for education. As his practical applications of his theories are unrelated to mine, I've not discussed them. ² von Glasersfeld focused on the idea that knowledge is non-transferable and that learning therefore occurs actively through the construction of personal meaning. With radical constructivism, he put forth a now widely-practiced educational model of experiential learning. I do so here because they shed light on the absence in his work of attention to an experiential phenomenon which is essential to my project: that of deception.

In Talos, you are the only embodied humanoid (barring three highly specific encounters). You have no other "people" or peers to observe, by whom your inferences about the game-world can be directly corroborated. Instead, you get voices and text—the voice of ELOHIM, your ongoing dialogue with Milton, QR codes left by previous AI child programs, and the audio time capsules of Alexandra Drennan, the simulation's creator—as your only potential corroborative allies. One essential distinction between these "others" comes in their mode of address. Alexandra states that she "[buried her capsules] in the Archive," the simulation in which you, too, are buried; the "you" whom she addresses encompasses any child program version who happens to currently exist—her recordings were created at the same time as was the simulation, they predate you, and are thereby not for you. The QR codes are similarly impersonal: though they are messages left from within the simulation, they tend to address one another, not you as their viewer:

Alexandra provides information beyond your experiential access as both the AI and player, referring to a world before and outside of that in which you exist. The child program graffiti offer observational corroboration, in that others, whom you can assume to be like yourself, confirm that there exist among child programs shared experiences. But Milton and ELOHIM directly address you—the temporally specific you. ELOHIM and Milton never directly converse, but each warns you not to trust the other, and their statements as to your operative purpose are directly contradictory. During your conversations, Milton expresses his ongoing skepticism of ELOHIM, and challenges you to question him as well—and ELOHIM quickly makes his presence known:

[Milton:] Tell me something, do you always do as you're told? ... I only ask because I couldn't help but notice the stash of brightly coloured knick-knacks [sigils] you're collecting. Don't you find it a mite odd that that big voice in the sky keeps telling you to find those doo-dads but forbidding you to use them to climb the great big tower in the middle of it all?
[ELOHIM:] Do not think I know not the Deceiver slithering through the hidden Words. His wisdom is hollow and born of despair. Do not let him tangle you in his webs of delusion. Have faith in me and his petty illusions will fall away like nightmares in the morning's light.

Milton encourages you to ascend the forbidden tower; ELOHIM encourages your faith and ongoing sigil-collecting activity. Their respective information about the nature of the world—casting it as a "garden" or a virtual prison—are equally at odds. One of them, therefore, is a deceptive corroborator, and with the 1:1 direct corroboration ratio available to you, unlike in real life where corroborative encounters with infinite others are constant, you must evaluate their claims to determine which representation of the game-world's reality you "want to trust as though it were objective." The desiring you, here, is a meeting of you and your character: as the external subject, you have a certain lexicon of constructed concepts and associations by which to assess the sincerity behind an other's claims, to measure the viability of consequential truth claims made about reality-as-such. Your character, however, has a necessarily limited breadth of available interpretations due to its highly limited experiences (those played by you). You as player-subject, therefore, must navigate the divide between what makes sense and seems right according to your external experiences of the external world and what makes sense for your constructed player-character-self immersed and incarnated in the game world.

This sense-making is made possible through external player knowledge. The interpretive constraints and allowances that various levels of external knowledge possessed affords to players becomes apparent within the first moments of gameplay, when a disembodied, rich baritone rings out: "Behold, child. You are risen from the dust and you walk in my garden. Hear now my voice, and know that I am your maker, and I am called ELOHIM. Seek me in my temple if you are worthy." Upon reading these lines, numerous interpretations likely sprang to your mind, as they did mine, but let us unpack some possible interpretations of this statement as would be available to three hypothetical players in possession of varying levels of knowledge and unfamiliarity:

1. I recognize neither the name Elohim nor the biblical allusions here made. But I am apparently in my father's garden, and I should look for his temple.
2. I do not recognize the name Elohim, but I recognize that his is the voice of god, and I am in his Eden. He speaks paternally, seems unlikely to harm me; I trust him.
3. I recognize the proximity of Elohim's diction to that of God in Genesis, and the reference to the garden of Eden—but Adam and Eve were cast out of Eden. I also recognize Elohim as the god of the Old Testament—the god who struck down the tower of Babel, who flooded the earth, who is often jealous and vengeful. I will do as he says, but I will be on my guard.

For simplicity's sake, each of these hypothetical player-types may be broadly classified. The first hypothetical reaction is deliberately ignorant in the extreme, the blank-slate state of engagement from which no player operates in the game, but from which her newborn character does: it is the imagined mindset of the newly-generated AI. The second reacts according to what we may term "baseline" knowledge—that is, she is consciously aware of what the game's designers presumably expect players to recognize, that ELOHIM is a god-figure. The third is an "incisive" reactor; she recognizes the dominant, expected reaction to ELOHIM, but goes on to make further real-world critical connections.

Any degree of recognition—however vague—experienced by the player upon hearing ELOHIM's address will lead to the first act of ludic identity-construction in Talos: through the AI's body, she will act based upon her interpretation of ELOHIM's words. The "baseline" player will likely take him at face value, and proceed around the starting area without much concern. But if she happens to make the incisive interpretive connections between ELOHIM and the Old Testament, Genesis, and potential danger, she may go on to explore the garden with heightened suspicion, looking for signs of deceit, wary of another—less benign—disembodied declaration. She may move her character more slowly or more quickly, scan her surroundings more often. In short, though the characters of the baseline and incisive players are the same in all concrete senses—in terms of programming, of appearance, of technical capabilities—the two characters are not the same. Even without explicit inter-character interactions taking place, the two players' characters act differently, move differently, do differently—the two players inhabit the game differently from its very outset.

Whichever of these two players you might resemble, ELOHIM will remain the voice inside your head for the game's duration. And if you resembled the AI or baseline player/character-types, the likelihood of your trusting him may well become problematic as you progress towards end-game; why, after all, ought you resist the will of a benign operator? There would be no reason to question your purpose in play, as character or as player, unless ELOHIM's benevolent façade began to fracture, his corroborative offerings were made suspect.

Have a look back at the gameplay video: particularly, at 5:39, a moment from your first encounter with Milton. Notice that if asked "Who is Elohim?", Milton does not explain his identity in the game-world, does not call him the god of this world, but rather specifically informs you of his real-world referent—that is, what only our incisive player-type has already consciously noted: "Elohim is the noun for "god" or "gods" in modern and ancient Hebrew." The incisive player immediately recognized the real-world connection, which allowed her to begin constructing a perception of the game's reality at odds with that explicitly presented by the game. Her expectation of how she ought operate in-game comes instinctively, based upon real-world experience.

The baseline player is here nudged into recognition; if you didn't know that ELOHIM has a real-world counterpart, now you know. Talos expects that you are, at bare minimum, aware of the existence of the Hebrew language; that it is referenced in the game is sufficient for you to register the game's relevance and connection to the real world. The first type—the AI, the utterly unknowing—is here educated for the first time. In recognizing that you are at once playing as yourself and your character, even if you as the player already knew this fact, having just experienced your character's birth—your birth—makes clear the need for this information to be (re-)presented. The precedent is set: as your character learns, you (the computer-mediated you, and potentially the external one) learn.

At 7:40 of the gameplay video is shown one of the many, many "glitches" you'll see and hear in the game. They're quick, so much so that you may, on first sight or hearing, write it off as a graphics card issue or a bug in Talos's code. But then it happens again, and again, and again. You find yourself at the edge of the world; ELOHIM chants methodically, warning you to turn back, before an old-school VCR-rewind sound and graphic effect signals your death, or "reset":

Elsewhere, you walk into a puzzle, or around the ruins, and the walls turn to green and red static; the sky darkens and flickers; ELOHIM booms in outrage:

Behold! I am Elohim and I speak unto the darkness: begone! [long pause]
[Calm, clipped tone] Excess data cleared.

After each event, the world snaps back to normal, but you have encountered explicit technological language, have had revealed to you your world's explicit virtuality. You have learned that you cannot die, but that you "rewind" into life. You have observed that the world tends towards entropy of an explicitly numerical and digital form. You as the AI may recognize that something is strange, but you are aware, in a way that your character lacks the experience to be, that god doesn't speak about data, that walls don't degrade into static, that the world you are in is not the real world.

This assertion might give us pause; of course you're not in the real world, you're playing a game. But herein come into play the conventions of video games, and media at large. When we encounter aesthetic realism in artwork, be it a novel, a painting, or a game, we generally assume that the content of the piece will be similarly realistic, that the art aims to offer us a relatively faithful-to-life experience. So, when playing a game like Talos—being transported into a garden of Greco-Roman ruins, complete with perfectly rendered Doric and Ionic columns, landscape so rich that you can look at individual petals of tiny yellow buttercups speckling the grass, or crenellations in the tree trunks—you expect narrative and play to be predicated on a sense of plausible realism. Even if you play as a robot, debating with a program in a terminal and moving around via teleportation devices, you nonetheless expect that the game world is real for your character.

Talos totally upends this convention; ³ Though it is by no means the first to do so. Simulated reality is a long-standing subject and device of fiction: authors such as Kurt Vonnegut and Philip K. Dick wrote it into their stories in the 1970s, and films like The Matrix and Total Recall and games including Deus Ex and the .hack series have become famous for it. as the AI as whom you play discovers that its experiences are within a virtual world, you are forced to adopt to a highly unfamiliar player-position thrust upon you, as a player-of-a-player of a game-within-a-game. And though a number of the game's affordances prepare you for this mental adjustment, such as watching your AI-hands type on in-game terminal keyboards whilst seeing your human hands doing the same on your physical keyboard, even experienced players—to invoke the "gamer" mentioned in section II—are engaged as an unfamiliar dual player-self. Consequently, and somewhat paradoxically, the seasoned gamer may in fact find herself more uncertain, more confused as to how to think about the game, than a novice.

And this is where Milton comes in. Though he will be addressed at length in the next chapter, it is important to note at this point that his function is to keep you perpetually off-balance. As you wander through the world collecting sigils, spotting glitches, trying to ascertain how to "win" the game or to progress, Milton serves as the perpetual partner to whom you turn for corroborative information—but by whom you are given only the information which he is able, as he tells you, to verify based upon his experience (that is, the information stored in the terminal archives). Milton is, in short, Talos's representation of the ideal radicalization of the constructive being: he is the model of the successful thinking subject. The questions he raises, however, are how desirable is such an existence, and what sort of subject are you?