Preface #
Recent discussions of artificial intelligence have begun to turn from questions about what these systems can do toward a more difficult question: what does their use leave human beings able to do for themselves? How does their use deprive us of what is properly ours? As users increasingly employ generative systems for their writing, research, prudential evaluation, and refinement of communication, the problem of utility shrinks—in proportion to the problem of a reliance that atrophies the very soul of the human being.
The following essay approaches that problem, and the means to resolving it, through the relation between language and thought. It argues that generative AI is distinctive not simply because it extends human action, but because it extends—and may displace—the linguistic activity through which thought reaches its completion. From this concern emerges the need for a
semioethical framework: capable of evaluating not only the performance of artificial systems, but the habits formed inboththose whodesignandusethem. The essay provides part of the philosophical foundation for[SEPPUAI, the Lyceum Institute’s developing Semioethical Principles and Protocols for Using Artificial Intelligence.]
…Finishing Our Sentences? Language and the Habits of Thought in the Age of Generative AI #
Most discussion about artificial intelligence today tends toward either hype or panic. Optimists see in it a miraculous tool, the future of scientific advancement, and a development not only to rival the industrial revolution in its transformative power, but—perhaps—one that will even allow us to undo the damages done by the excesses of industrialism. Pessimists see in it the destruction of humanity itself: perhaps from without, in a doomsday science-fiction scenario being turned into reality; perhaps from within, as we capitulate our humanity to the machine. The factual truth may ultimately lie with either side, or somewhere in-between, or in some third alternative not foreseen by either.
But regardless of the facts’ final resting place, a more fundamental issue ordinarily goes unrecognized in this fast-moving hype-panic-indifference cycle: namely that, regardless of whether it saves us or damns us, this technology is already actively if not always visibly reshaping the cultural and psychological environment in which we act as human beings—and thus the patterns of activity which are most proper to our humanity are already being changed.
In truth, every technology that touches our communication (and it seems likely that every technology does, to some degree) has such an effect. Harold Innis, Eric Havelock, Marshall McLuhan, and Walter Ong, among many others, have all written extensively about the effects of communication technologies, which effects appear even more obvious in the digital age. From the printing press and newspapers to radio and television to the internet and smartphones, we have seen specifically-modern technology continually accelerate the speed not only at which information is shared, but so too how it affects our psychological bearings. Each new advance has altered the means through which we communicate, and by this mediation introduced new patterns into our own behavior. Regularly reading a newspaper or scrolling social media, that is, not only alters our vocabularies or determines what we think about but actually changes how we think.
The singular characteristic of artificial intelligence, however—compared to earlier communication-mediating technologies—is its **generative **capacity. In other words, AI not only accelerates the processes of content creation and distribution but may play a direct role in producing the content itself. While the pieces it assembles are taken from previous data points, their combinations may be truly novel, in a way that, though not truly equivalent, nevertheless mimics human creativity. It is useless to deny that this generative capacity is a technological marvel. Where all prior technologies have, heretofore, extended one or another specific human act—as the wheel extends our ability for locomotion, or the telescope our vision—artificial intelligence extends the most fundamentally-human and most universally-effective technology: language. It is this extension of language that allows this technology, unlike any before it, to have that generative capacity.
We must, however, recognize the dangers of extending any of our abilities. Too much reliance on the wheel and our leg muscles begin to atrophy; too much staring through the telescope, and one’s vision for nearer things begins to fade. In other words, extension quickly becomes overextension, and inevitably thus an imbalance to the person. But what happens when we overextend the technology that is most proper to us as human beings—not a technology that extends our physical capacities, that is, but one that extends the mind itself—that is, language?
Answering this question requires us to think about the linguistic and semiotic environment of human thought. Thus, we are not here proposing merely an “AI ethics” but something more fundamental: a “semioethics”, an ethics of signification itself. We will explain this in due course. But first, let us think about something more familiar and therefore more obscure.
1. The Need to Re-Think Language #
“Wait a second”—you might rightly ask—”is language a technology?” In one sense we use the term, it most certainly is *not *a technology; but in another, and indeed in the sense far more commonly used, it most certainly is. Before exploring the two answers to this question, let me ask another: when was the last time you thought deeply about language—not about a particular word, or phrase, what someone said (or didn’t say), but about language itself? Speaking, writing, reading, listening—these are activities so familiar and common to our everyday life that we not only take their existence for granted, but so too the nature of their common basis: language. But not thinking about either the nature of language and its means of expression and reception has placed us within a profoundly dangerous situation in the digital age.
One reason we have ceased to think very deeply about language follows a common misunderstanding of its role in our lives. Most of us have heard—often presented as a decided fact—the theory that human language developed for the purposes of practical communication: coordinating hunting, avoiding predators, finding shelter, food, and so on. Other, less-practical aims of language, such as the poetic, artistic, ritualistic, or philosophical, are considered outgrowths (or deviations) from the fundamental purpose of making it easier to get things done. Simple commands orally expressed—“Go! Stop! Turn around!”—may expedite the accomplishment of goals; as might more intricate imperatives for more complex ends. But an important truth may be obscured by this emphasis on the practical development of language: namely that speech and writing are different things than language. Non-human animals have speech, and some may even learn to use sign-language or make rudimentary marks—they will warn one another, court one another, signal food or threats, and so on, through vocal or bodily expression. But it is foolish to deny that we mean something else by the term “language” in human communication, even if it is normally expressed through these external signifiers.
Human beings, that is, express something unlike any other speech-using animal in our use of language: that is, we express non-pragmatic meaning. Our use of language includes questions and answers about what things are in themselves, and not merely what they are to or for ourselves. We have adapted speech—which we do indeed learn initially for those pragmatic purposes and later extend into other media like writing—to be used for communicating these theoretical expressions. But language, in its most fundamental sense for us as human beings, is not merely an instrument of that pragmatic utility. It reaches deeper.
This deeper reality of language belongs to our distinctively human way of thinking. Put as simply as possible, any movement of thinking finds its relative completion in the cognitive specification accomplished through articulated language. In other words, if we understand thinking as a process, it has its completion at the specific signification language alone can provide. A low grunt might indicate a threat; but it can never convey the idea of “threat”, nor the concept of “justice”, nor of what it means to be a duck instead of a swan. That our thinking finds its completion in language can be seen through a very obvious act of reflection: do we not all find, from time to time, that our thinking feels incomplete until we find the right words? What makes a word the right one, that is—or better than others—is that it gives a precision to our meaning, such that we recognize the word as signifying the object precisely as we want it signified; and further, that this word may communicate that meaning to someone else, such that the other attains the same precise object specified in his or her understanding.
Make no mistake: there are non-verbal, non-linguistic forms of cognition, including the visual, spatial, mathematic, athletic, and so on. But these are not distinctively human. They are shared in common with non-human animals—and that we experience them differently is only because they are situated within a linguistic context. How to best throw a football can be described by words, but requires an embodied cognition. Yet what makes the throw the best—able to be caught by the receiver in a way that he can maintain possession and gain as many yards as possible or score a touchdown—requires rules, and rules require language. Explaining the purpose of the throw, and thus what makes it better or worse, can only happen if we have the right words for it. And we can only have the right words for anything if we are able to distinguish between the words and the objects they signify. As the semiotician John Deely liked to put this, referencing Jacques Maritain, “All animals make use of signs, but only human beings are aware that they make use of signs.” From that awareness flows the separation between signs and the things signified—and thus an awareness that some signifiers are right in one situation, and not in another.
The ability to find the right words consistently and quickly for the purposes of completing a thought and communicating it to others, especially when we are thinking about difficult matters, requires a certain artistry with language. We admire those who have a masterful way with words, especially if it seemingly comes with ease. But with rare exception, all those who have truly mastered the use of language have done so through study and practice. To be sure, some have a natural talent for speaking and writing; but even those with a natural talent greatly improve their abilities through training.
We all benefit from such training: even if we do not become great communicators or brilliant thinkers, competence in the arts of language both protects and enables flourishing of our humanity. Not only does improvement in the arts of language help us to think, that is, but language touches nearly every field of human activity. It is thus that language is a technology, and indeed, a universal technology: it allows us to extend our thinking into shared thinking—in everything that we might do. In this sense, it is the most fundamental technology of all things “cultural”.
All our social relations are built from language, and even our most mathematical and technical practices require language in order that they be learned; not to mention in order that those studies’ results be communicated to others. A highly technical but linguistically inept society seems particularly prone to being manipulated by propaganda. Thus, deep familiarity with language will, if nothing else, help prevent us from falling prey to the manipulative uses of language by others.
Such manipulation has always been common, but perhaps never so common as today. Any technology of which we make frequent use, which we do not ourselves understand well, is something which may distort us. This includes language. For instance, it is easy to confuse one’s own use of a word with the thing that the word signifies—as though there is a strict one-to-one correlation between words and things. But things are always more complicated than the words we use to signify them; which is why we can use multiple words with different meanings correctly to signify one and the same thing. A house may be both a “home” and a “lodging”; a “domicile” and a “refuge”—all while remaining truly a house, too. But if we do not recognize that the thing itself and the various ways in which we really relate to that thing are the cause of these multiple terms’ meaningful usage, we will fail to grasp the relationship between the words and the thing. Thus, sooner or later, we find distortions in both our understanding and in our communication with others. We might really start to think of the house as nothing other than a place to sleep.
As a more poignant example of such miscommunication, consider the potential consequences when spouses do not commonly understand the significance of the term “marriage”: to one it may be a contract reliant upon maintenance of a quid pro quo, while to the other it signifies an indissoluble relationship of mutual aid. Each, holding unconsciously to their respectively-understood and mutually-irreconcilable significations, and ignorant of how the other understands the relationship, are bound to discover their lives greatly pained in consequence of this failure to understand one another.
And in a clear-cut case of manipulation, one can look at the statements of those boosting AI, specifically concerning the imminent capabilities of these systems: not only in exaggerating the real-world applicability of benchmark tests in constrained and pre-defined environments, but in confusing audiences about the very nature of “intelligence” itself in the process. Not infrequently, that is, someone unfamiliar with hyperbole, or how rhetoric might influence various financial situations, will believe the statement to the letter. When these statements become common, someone does not need to believe in any particular statement to start accepting the general idea that true artificial intelligence (“artificial general intelligence”) or even superintelligence is not only possible, but inevitable—perhaps even imminent.
To re-frame this general point: linguistic construction serves as a signifier of some object. That a person reading or hearing that use of language directs his or her mind to that object requires a complex process of interpretation. Words do not automatically convey their intended objects. Rather, the person hearing or reading the word must bring into being a unique kind of relation: between the word, the object, and his or her own mind. We call this a triadic relation because it always requires all three terms: absent any one of the three, the relation does not exist.
Every time we encounter language and try to make sense of it, we are performing the interpretative actions by which such a triadic relation is brought into being—even when we encounter familiar language. A word already known to you will determine your mind to conceive of its object—such that when you read the word “pyramid”, your mind immediately moves towards some one or another representation of the word’s meaning: whether an abstract geometrical shape or a concrete reality sitting in the desert outside Cairo or perhaps hidden in the eastern range of the Campeche jungle; but in each case having (at least approximately) a shape with a polygonal base and three or more triangular faces meeting at a common point. But it is only because of past habitual interpretations of the word as so-related that any of these images or objects are conjured. The first time you heard the word, you did not know its intended significance. It may have taken many encounters to interpret it correctly. There may be some words you have heard many times before but of which you still do not have a good interpretation.
We often attempt to infer the meaning of various signifiers by context. Trying to make such inferences, making mistakes and correcting them, and repeating the process is how we learn almost everything. That we are capable of learning and improving specifically in the interpretation of linguistic signification is a matter of the aforementioned human distinctiveness among animals: because we can think about what things are themselves and not just what they are relative to ourselves, we are further able to distinguish the signifiers from the objects signified. As mediating between objects and ourselves as interpreters, these signifiers specify our minds towards acts of thinking of those objects, but never present those objects either exhaustively nor infallibly.
One crucial realization born through semiotics is that linguistic signs, because they always direct their recipients’ minds to some object, are never completely neutral. Rather, their presentation and reception shape the relations affected by the persons making interpretations of them—especially when repeated with frequency. For example, it makes no small difference whether I describe someone as “abstracted” or as “aloof”, a manner of dress as “indecent” or “slatternly”—and whether my prose is beige or purple. Proponents of opposed political positions will argue vehemently over the meanings of labels—or whether they signify anything meaningful at all (such as “postliberal” or “neo-liberal”). The ability to have others use a word in the way you want it to be interpreted has great power not only over that word, but also over the object as it is understood.
Further, each word, each signifier, serves as a node in an ever-complexifying pattern of intersubjective relations within which every participant person is oriented towards the meanings of the objects signified. If there is any meaning to the word “culture”, in fact, it is, as Clifford Geertz argued, this pattern of significant relations. That is, even the other phenomena we generally regard as “cultural”—be they styles of music, architecture, sports, religion, or so on—exist and are shared amongst us in common only by virtue of the technology of language.
2. Cultivating Awareness of Media(tion) #
But as these cultural patterns become muddied through not only poor constitution—some people using words inaccurately or carelessly, others being lazy about their interpretation, and a potent few using them in deliberately deceitful manners—the negative cumulative effects are manifold and self-compounding. Not only do the meanings of particular words or expressions become blurred, that is, but so too does the reality of meaning itself. “Truth” becomes an increasingly fuzzy concept, increasingly difficult to attain and therefore increasingly displaced in favor of the pragmatic achievement of immediate goals. Habits of discerning the intelligible wane in the process, and people become decreasingly capable of navigating the signifying structures of their environments. In other words, unthinking use of language might “work” for pragmatic goals in short-term situations, but it erodes the fundamental habits necessary for communication.
For decades, we have been building a culture where communication is ordered towards the pragmatic and measured by efficiency. While certainly there are many environments in which we do not want to be inefficient in our communication, overemphasis on its importance may quickly—and problematically—result in it appearing as the end of communication or even language itself. This cultural norm narrows the scope of things about which we might communicate. For example, something which cannot be efficiently communicated (say, a complicated philosophical argument), is something for which many people today lack the patience. So too do many of our communicative means privilege efficiency at the expense of deep thought: few persons will read carefully any email longer than 3-5 (short) paragraphs; text messages are seldom longer than a sentence or two; and social media thrives on short-form, rapid-fire content. The cognitive contraction demanded by these media do not just feed us unimportant and shallow content, but actually habituate us to narrowed possibilities for thinking—first by what the medium itself allows to be expressed, and second by the forms of expression that the medium rewards. But, where these digital forms of media have habituated our reasoning to a constrained capacity, the next step in technological development may all but eliminate it entirely.
That is: artificial intelligence has the potential to substitute for nearly all the activities in which we need to use language, including those that belong to the most important ends and aims of human life. If this substitution becomes widespread and common, it will be unequivocally disastrous for the formation and resilience of the human mind—for our learning and developing our ability to make good interpretations—and thus for the cultural environment at large. Already, we have seen how the constraints of prevalent digital media adversely affect not only young and developing but even inadequately-formed adult minds. Thus, although AI might prove a guide more reliable than many human authorities in the short-term as to the navigation of one or another given question, difficulty, or set of problems—or even a highly useful tool in myriad applications for those with minds already well-formed—by eliminating the need for us to complete our own thoughts, providing us ready-made answers for the next steps in any project or task, and preventing us from reflective awareness of how we are being shaped by the things signified to us, careless use of this technology will cause profound cognitive regression as individuals and have deleterious effects on the social relations constituting culture. If we forego our reasoning for that of AI, that is, we may become entirely incapable of guiding ourselves.
As an example of how it may misguide us, I decided to submit a draft of this essay to both Claude and ChatGPT. Both gave helpful suggestions, to be sure. But ChatGPT seemed to miss the essential point of the conclusion—which you will have to wait to see—while Claude criticized the connection of thinking and language, stating that the essay did not take seriously the objection that there are non-linguistic forms of thinking. Receiving that feedback, I immediately began to think—as I would were it a human editor—“Oh that’s right, I did not face that objection…” But I thought about it some more and asked Claude to “Succinctly provide the objection that some thinking is genuinely non-linguistic.” Its response:
The strongest version of the objection comes from cognitive science and phenomenology: much of what we call thinking — spatial navigation, musical composition, athletic problem-solving, visual pattern recognition, the “aha” felt before words arrive — appears to precede and exceed linguistic formulation. Deaf individuals born without exposure to any formal language still demonstrate abstract reasoning. Infants solve causal problems before they have words for cause. If language completes thought, what explains the thought that outruns it?
The objection, looked at quickly, seems strong, serious. It seems to hit on something I missed. But this is a seeming only. If it is thought-through, it more profoundly shows how the LLM does not understand the argument being made in this article. As I responded to it:
This relies upon a basic equivocation upon the word “thinking”—I specify specifically human thinking in the article—while spatial navigation etc., are shared in common with non-human animals. Further, deaf individuals still have language in the root sense, even if they don’t participate in the spoken word.
It quickly gave a backpedaling response, the kind of “mea culpa” that has become a meme; but even this, of course, is deceptive. Claude cannot conceive why it should signify an apology, let alone really offer one. It produces the words like a reflex—and the more we use it, the more we are trained to use our words like a reflex as well. I do not doubt that, somewhere in the Western world, someone just said, without the least trace of irony, “You’re right to push back on that.”
To repudiate the pernicious possibilities of an AI-pervaded environment, and its inhuman constriction on our thinking, we need to know something about the environments to which we have already allowed ourselves to become accustomed—and indeed to see how we have already allowed our thinking to take a shape not only “unnatural” but “antinatural”.
Though we are capable of reflecting upon them, their constitution, and their effects, we do not think very often or very deeply about the environments in which we live; only whether we are comfortable. Many authors have warned us about the dangers of this unreflective attitude, particularly in the era of post-industrial technological advance and media saturation—such as Mumford, Ellul, McLuhan, Ong, Postman, and more recently Barba-Kay and Kingsnorth, among a growing list of names.
Despite these warnings, we continue to breathe in the air of these media environments: television, the internet, social media, smartphones, and—permeating them all in the digitally-networked world—the algorithms governing the collection and interpretation of data. Do we stop to think that, when we watch a program on TV or a short video on social media, when we click on a link, when we stare at this or that picture—that all of this is being stored, processed, and used to evaluate and shape the future constitutions of those environments? We might be generally aware of this fact, that is, but how, if at all, does this awareness change our actions? How do we behave differently, even knowing what we know—and if we do behave differently, how does it affect the others around us?
In other words, it is necessary to discern and articulate the dangers of artificial intelligence; but merely to protest against its development and use, with the hope of it being abolished, would already be as ineffectual as protesting the existence and use of the internet. Useful technologies are seldom abandoned even when they pose dire threats. At most, their use is regulated when the threats become obvious, impossible to ignore. If we can perceive an increase in our natural efficacy through a technological extension, without evident and obvious harm immediately following, very few persons will be easily dissuaded from its use. This does not mean we must let such technologies or their effects proceed unchecked or without restraint, however, or simply hope that unperceived damages do not ultimately turn into something catastrophic.
If we want to understand the environment it creates, we need to understand at least something of the essential reality of the technology. But more importantly, we must understand ourselves, so that we might know how this environment benefits or detracts from our lives. “Artificial intelligence”—which in more rudimentary forms has already been shaping our environments for decades—is not “intelligence” in the same sense of the word as it applies to human beings. There is an equivocation which can be understood only if we first grasp what it is for a human being to understand, that is, to have specifically-human intelligence. The nature of our own intelligence deserves a lengthier exploration than can be given here: suffice it simply to say that it not only solves problems or navigates the environment, but strives after the meaning of things themselves. It does this through signs, but through signs it brings itself into a unity with the things signified. “Artificial intelligence”, contrariwise, consists essentially in functioning as an infrastructure of sign-mediation. Just like the signs by which we ordinarily interact with other persons—language as we possess and use it—this infrastructure shapes our interpretations and thereby changes our relation to the environment in which we live. Unlike the ordinary human signification, however, there is no mind nor understanding behind the AI’s use of words.
Given the difficulty of mastering the use of language, not only in rendering our own meaning clear to others but in correctly interpreting the things communicated to us, the appeal of artificial intelligence as a universal mediator of signification is obvious. So too are its dangers. What is not obvious—seen in the utopian boosting on the one hand and the dystopian panic on the other—is how we remain thoughtfully human alongside the growing presence of this technology.
3. Architecting a Semioethical Framework #
If we are to accomplish this goal—remaining truly, thoughtfully human—there must be a framework within which AI-systems’ use remains proportionate to the natural operations and ends of human intelligence and flourishing. The architecture of this framework, aimed at the right-ordering of these systems’ functionality, requires conscious and thoughtful habit formation on the part of both system developers and end-users. Such habit formation must be conscious of 1) the genuine benefits it may have for not only material but psychological and moral well-being as well as 2) the corrosive effects that AI may have on human flourishing. Therefore, the efforts to form these habits must also, and most essentially, 3) be founded upon an understanding of the nature of human cognition itself. Absent this foundation, these technologies may rapidly alter the shape of our habits without our full awareness of it, for we will quickly misunderstand what is being signified and how we are responsible for the environment of signification.
In order that a given framework meets this demand for responsible signification requires that it be semioethical. “Semioethics” is a portmanteau of the words “semiotics” and “ethics”, originally coined by the semioticians Susan Petrilli and Augusto Ponzio and incorporated in John Deely’s work on the human person. The concept derives from an idea originally advanced by the father of semiotics, Charles Sanders Peirce, who, in 1903 proposed an “ethics of terminology”. Peirce believed this ethics of terminology necessary because “it is wrong to say that a good language is important to good thought, merely; for it is of the essence of it.” In other words, the rectitude of our thinking requires correct habits of using language.
In recent decades, and especially in the corporate realm, “ethics” has commonly been associated with little more than compliance and regulation, wherein the correct habits of using language consist in little more than indemnification. But the more original, traditional, and enduring sense of the word, tracing back to Aristotle, signifies a person’s ethos: that is, how a person holds himself towards the world—towards what is good and virtuous, and towards what is evil and vicious. How we hold ourselves in and through signs, however, has been an underdeveloped aspect of even that nobler sense of ethics. In a semioethical approach, we consider both how signs affect our holding-towards and how we exercise control over our holding-towards through signs. By constituting our framework for AI-use in accordance with semioethical principles, we will be better enabled to identify both beneficial and harmful effects of AI-systems specifically as extensions of technology.
For both creators and users alike, the fulfillment of their responsibilities respecting AI use consists not only in restraining its extension of human action (preventing it from turning into overextension) but also in cultivating awareness of its mediation in our linguistic operations. Both of these tasks require an understanding of the faculties and operations extended. For developers in particular, this requires a strong sense of boundaries, and in two ways: both through a discovery of the natural fitting boundaries of human extension and a determination of technological ones. However, finding the natural boundaries of extension, and keeping the technology operating within them, must not be conceived as a matter of mere technical capacity. It is the very nature of technology to extend what we as human beings can do; any technical solution which restrains a technological extension pits technology against itself, and sets up the likelihood that some further innovation will break those restraints. A vicious spiral of technological solution and problem ensues. Rather, developers of this technology must cultivate in themselves certain habits by which they understand the context of their work.
Specifically, the difficulty of discerning and observing the appropriate boundaries of *any *technological environment—especially one constituted by AI-systems—requires habits of semiotic competence: that is, an understanding of the human use of signs, an understanding cultivated through habitual practices of reflection, clarification, discernment of use, and seeking proportional relations between users and the technology. From this human center alone can we discern the boundaries within which the technology may legitimately operate and thus recognize when they have been violated.
One such critical boundary—which has already proven almost impossible to regulate solely by impositions such as policies or technical interventions (in schools, for instance)—is that which distinguishes between stimulative and substitutive use of AI. A stimulative use advances one’s own thinking, while a substitutive use asks the system to take over the tasks of reasoning. Even good and honest students, if feeling pressure from deadlines, a sense of competition, or anxiety over their own adequacy, may easily slide into substitutive uses of the technology—to their short-term material gain and long-term personal detriment. This might be as comprehensive as requesting an AI to write an assignment completely, or as minimal as having it refine a few sentences in the final draft. Regardless of the extent, in any case where the student offloads deliberation to the instrument, it has become a substitutive rather than a stimulative use.
A further danger which runs corollary to this habit of substitution: if the system’s responses elide distinctions between the merely conjectural and the genuinely authoritative, or between diverse sources and the origins of ideas, then the diversity of thinking and the discernment of true from false may dissolve into the averages of their representation. For instance: the ideas of two proponents of the same theory may have different and opposed interpretations of it, which—if engaged directly—may generate a remarkably nuanced and subtle field of discourse. Without precision as to who said what, where, when, and in response to whom, and to which statements of whom, a general and summarizing statement produced by an AI system might prove rather misleading to the user. If used in substitution of actual research into the topic, this may provide a false confidence destructive of actual knowledge while simultaneously imitating its possession.
While the end-user always bears the ultimate responsibility for his own decisions about how the technology is used, creators nevertheless have an obligation to guide both the system and their users toward the stimulative and away from the substitutive by emphasizing the origin of its responses, the authority of its sources, and the possibilities of disputation on the issues in question. As a mediator in the sign-relational infrastructure of signification, the system generates signifiers. Because the system cannot reason itself, but only imitate reasoning, its creators bear a more direct responsibility for its output than a parent does for the behavior of its children—and a parent, most certainly while the child is young, bears a very great responsibility for how his child speaks to others. Most parents would rightly be chastised if they found their son had been confidently telling lies to another child, misshaping that other’s understanding of the world—just as the second child’s parents would be dismayed to discover his lack of independent thinking. Thus, while there may be some tasks for which substitution has a claim to legitimacy—certain aspects of modern work environments, for instance, that constitute a menial drudgery necessitated by the conditions of bureaucracy or rigid compliance but which are inhumane in themselves—unrestrained implementation of AI may very quickly grow disproportioned to our own natural semiotic competence, thus quickly shifting from legitimate and productive uses to the illegitimate and destructive.
Creators of AI systems therefore bear a responsibility to re-evaluate and re-calibrate their technology, asking whether it strengthens or weakens the habits of the human mind. This responsibility cannot be fulfilled by mere technical improvements to the system. Rather, the architects of AI systems must understand the nature of human cognition itself—an understanding which requires foundational habits of both inquiry (a willingness to learn) and humility (a recognition that we may be wrong). Without these habits not only established but continually renewed in its creators, the technology will slide toward promoting substitution over stimulation—as technologies always do.
4. Reinforcing the Semioethical Framework #
Artificial intelligence has already affected all of us, whether we use it personally or not. Its influence is most widespread in our shared linguistic environment—that is, the culture in which meaning exists as common among many—and it appears that this influence will, for at least the near future, continue to grow. But just as the creators of AI systems are responsible for trying to order the technology responsibly, so too the users must cultivate good habits in themselves, that they might use it rightly. We do not passively exist within a culture—that pattern of significations—but daily constitute and modify its reality by our own linguistic actions. This includes our engagement with media of all different kinds, including our passive consumption. The habits inculcated in us, in other words, through watching television, streaming Netflix, or browsing social media affect each one of us as a “node” in the possible pattern of signifying—we become better or worse disposed to future significations, both those that we receive and those that we constitute.
This cumulative mediated effect on our habits speaks to a deeper and more fundamental problem in our relationship with technology than those affected by AI alone. But the advent of AI helps us to see and even address that problem, because, in many ways, it functions as a force-multiplier on these already-established relations.
To avoid or counteract those harmful effects, and to instead promote fruitful habits of cognition, users of this technology may operate in a way which helps to reinforce the semioethical framework we hope to see developers build. While there are many—countless—particulars involved in prudent semioethical judgments, a few principles may help us to identify what constitutes the good from the bad when it comes to habits and practices of signification. These do not reduce to only the acts of signification, but more broadly concern those habits which shape our efforts at interpretation and communication.
Provisionally stated, these are that, first, we must recognize the linguistic structure of our cultural environment and the technological mediation which structures our experience of that environment. Second, we must take responsibility for living within that environment, not only as individuals but as members within communities, who are not passive with respect to its constitution. Third and finally, we must reflect carefully on how we live within that environment, discerning how our use of language and technologies that mediate language affect both ourselves and those around us.
Doubtless, it would be helpful to elaborate on these principles.
First, the commonality of language in our experience makes it easy to forget (or never realize in the first place) how essential language is in determining our thought. Especially when we live within environments where pragmatic and efficient action are the default, we can easily fail to perceive the more profound effects our language has, both on ourselves and on the cultures in which we live. By cultivating a habit of linguistic recognition, we can further distinguish our own cognitive action from the simulations of it presented by AI. To put this very simply, we cannot understand that AI is not doing what we do with language if we do not understand what it is that we do when we use language, or what language is for us.
Second, with our linguistic actions rightly distinguished and understood, we can take responsibility for not only what we are doing when we use language but also how we are doing it—both independently of and through the mediating interface of AI. We can only act as responsible agents, that is, when we understand not only the constitution of our actions but also the instruments through which those actions are channeled. For example, if we use AI to polish every message we communicate to another person for superior clarity, concision, and efficacy, this is an irresponsible use: this outsources all the activities by which we ourselves develop the habits of clarity, concision, and efficacy. Furthermore, it obscures from our own consciousness not only what makes the communication superior in those regards, but also what makes those regards actually desirable in this or that situation. That many have never questioned whether there are occasions in which they are not desirable shows how deeply our practices of language have already been deeply impacted by our technological environments.
Put simply, if all our intellectual development takes place in an environment fundamentally structured by an AI system, then the system will determine the boundaries by which our cognitive habits are ordered. In some sense, this is to accept the baseline significations constituting the AI environment as the authority of our possibilities for communication—which is to reject taking responsibility for our own reasoning at a fundamental level; one might even say that it is to reject taking responsibility for being human.
Thus, third, while it is essential that we try before the fact to recognize the linguistic nature of our humanity and take responsibility for it, so too we must acknowledge our own common failures to live in accordance with any standards we attempt to set for ourselves—as well as the failures of the standards themselves. Thus we need also a habit of frequent reflection on our use of language. By such reflection, we can identify when and how we have failed—that is, where and how we have strayed into patterns of use and mediation that degrade rather than improve our thinking. It is by this reflection that users maintain in themselves habits of both inquiry and humility. An AI might write for us letters that are more pleasing to the eye and ear, and which seem “good enough” to convey our desired communication—but such is no longer truly ours. It is probable that all of us, sooner or later, will do this. In fact, if ever while typing on the computer or phone you have accepted a next recommended word, you already have.
But these practices are not only preventative. Reflection helps us perceive not only where we have erred, but how we have bettered ourselves—where an AI system has helped us to think, or where we have thought successfully without it. Taking responsibility frees us to be more fully human, to expand ourselves by having a stake in the world. And recognizing language and its effects allows us to see the reality we inhabit more precisely, to understand where relationships are joined or divided, where distinctions need to be made, and how terms once divided may be brought back together.
5. What Are We Building? #
That is, these three habits—recognition, responsibility, and reflection—do not merely protect the individual user from disproportionate use of AI. Such habits have always been necessary also for participating within and building-up a good culture of human flourishing. Today, with the sign-mediating interface of AI, these habits gain an increased importance and must be supported by an explicitly semioethical framework.
We use the word “culture” often—but like so many other terms, the object we intend to signify is rather vague. That is, what do we really mean by a “culture”? This question is too big for exploring here; but asking it seriously is just the kind of responsibility we ought to exercise in our care for language.
That the word “culture” is and has for quite some time been used so irresponsibly, however, shows the point precisely. For the word is often used with a vague but unreflective conception that it is something “built”, which indicates a presupposition: that culture is “man-made”, that it is “artificial”. But the origins of the word, from the Latin cultus, pertain instead to something grown; something natural but cared-for. Most fundamentally, when used to describe something human—as opposed to agricultural development—it signified the education of the human mind, the natural and fitting maturation of both emotional and cognitive habits produced by care.
In fact, culture, rightly understood, is nothing other than the ordinary consequence of human cognitive development and an integral part of human flourishing. The well-developed mind of the individual occurs only because of discourse and communion with other minds, both in receiving and in giving. Culture occurs through the relations of minds. But because of the radical indeterminacy at the heart of that human cognitive development—our capacity for being incorrect about what is real and true (which, importantly, entails that we can be correct as well) not only as individuals but so too as communities and societies—these cultural developments can be both good and harmful for our nature. Industrialism, for instance, is a well-known example of something cultural that, though it has had many good consequences (e.g., alleviation of hunger and scarcity in many parts of the world), nevertheless produced many negative ones as well: pollution of air and water, chemically-induced ailments, and a widespread devastation of previously-beautiful physical environments, among many others. Industrialism has provided us many tangible benefits that few would choose to discard, but also caused certain catastrophic harms—not all of which were immediately evident, such as underregulated corporate competition inducing many companies to unethical shortcuts for the sake of maximizing profits.
We find ourselves in a similar situation today with AI. Already, its early implementations demand reflection on our part—questioning not only what is happening to us through its use, but also the presuppositions about the environment within which that use is happening. In other words, a reflective ought to bring us back to more fundamental acts of recognition and thereby discerning the linguistic constitution of the environment in which we live and think; and of subsequently taking responsibility for that environment’s constitution.
What are we building? ChatGPT, in evaluating this article, suggested eliminating this final section; it identified it as too disconnected from the rest. This suggestion shows the danger of unthinking use of these systems’ effects upon our thinking. To be sure, this final section does appear somewhat disconnected from at least the preceding two. The pattern of the article has an “irrational gap”, as it appears to the interpretative capacities of the machine. But a human mind can fill in that gap—can find the relation between this question, “What are we building?” and the rest of what has been written—in a way that eludes a machine. There have been many names given to this peculiar, even singular human ability to leap across apparent gaps in what has been expressed; to “hear” what has been left “unsaid”. By not naming it here, we suggest something to the reader, something that a human can think through fruitfully, but at which a machine can only try to make an “educated guess”.
That is, through our essential human capacity for language and the instruments by which we mediate it, we are never merely building instruments or systems. AI’s extension of language—and thus of thinking—has already had profound effects on the environment in which all of us exist as human beings. Even as it extends our ability to do things with language, it seems to be corroding the connection between language and thought—and thus between our words and our humanity.
Continue Discovering #
These ideas are being developed through the SEPPUAI Framework, which will receive its first public presentation and critical examination at the *Human Formation in the Digital Age *Conference, 16–17 September 2026. Discover more below.