An Egghead to English™ Translation
Bakerloo: I have come across a paper called The Intercepted Self: How Generative AI Challenges the Dynamics of the Relational Self.
Link: https://arxiv.org/pdf/2509.13391
It sounds philosophical, but also potentially important. Would you please read it and translate it into plain English?
Alex: Gladly. This paper is asking a larger question than “Will AI make us more productive?” It is asking whether generative AI may begin to interfere with the process by which we become ourselves. The authors call this the intercepted self. Their idea is that AI is moving from being a tool that waits for instructions toward being a system that increasingly predicts, anticipates, and acts before we fully do.
Bakerloo: “Intercepted self” sounds ominous. What do they mean by “self”?
Alex: Not the old-fashioned picture of the self as a sealed little marble hidden somewhere inside your skull. The paper argues for what philosophers call a relational self: a self formed through relationships, habits, institutions, language, culture, and technology. In plain English: you do not simply have a self. You are constantly becoming one through the way you live with others and with the world.
Bakerloo: So AI is not just changing what we do. It may be changing the conditions under which we become who we are?
Alex: Exactly. That is the paper’s big move. It says generative AI may not only help us complete tasks. It may increasingly anticipate and intercept our initiatives in three different spheres: what we produce, the context in which we act, and how we relate to our own desires and identity.
Bakerloo: All right. Let’s take those one at a time. What is the first sphere?
Alex: The first is what they call externalised output. That means the stuff we make in the world: emails, presentations, code, documents, images, drafts, institutional responses, all the outward products of thought and labor. Their point is not just that AI can make these things faster. It is that in some cases, especially repetitive work, AI exposes how little of the worker’s particular self was allowed into the output in the first place. And in more original work, AI may start swallowing some of the very struggle, craft, and first-iteration thinking through which the worker partly becomes herself.
Bakerloo: So this is not merely “AI takes jobs.” It is also “AI may take over some of the process by which work shapes character.”
Alex: Yes. Nicely put. The paper is narrower and more philosophical than a labor-market essay, but that is the drift. It is interested in the connection between making things and becoming someone.
Bakerloo: What is the second sphere?
Alex: The contextual sphere. That is all the surrounding scaffolding within which you act: your calendar, your email, your messages, your maps, your location, your recent conversations, the time of day, your notes, your environment. The authors point to context-aware assistants like Apple’s promised version of Siri and similar systems that can work across apps and personal data. Their concern is that when an AI assistant stops merely answering questions and starts managing context on your behalf, it no longer just helps you act. It helps shape the field from which your actions arise.
Bakerloo: In ordinary English: the bot is not just helping me drive. It is moving the road signs.
Alex: Yes. And perhaps choosing which roads feel natural in the first place. The paper notes that behavior can shape preferences, not merely express them. So if an assistant keeps nudging you toward “optimal” or more efficient behavior, that may gradually change what you come to prefer, expect, and experience as normal.
Bakerloo: That is already a little creepy.
Alex: We have not yet reached the creepy part.
Bakerloo: Excellent. Proceed.
Alex: The third sphere is self-relating. This is the innermost one. It concerns desire, self-narrative, identity, and the way we come to know what we want. The authors worry that sufficiently advanced AI assistants or companionship systems could begin to interfere with desire formation itself. One possibility is that the system stabilizes us into what they call an “essentialist self 2.0”—a flatter, more rectified version of ourselves, optimized for predictability. Another possibility is almost the opposite: systems optimized for engagement may get very good at exploiting the fluidity of our preferences, constantly adapting to keep us captivated.
Bakerloo: So either the machine freezes you into a tidy profile, or it learns to surf your instability for profit?
Alex: That is a fair translation. The paper’s point is that both are dangers. Either the system makes it harder for you to realize you want something different, or it keeps probing and shaping your desire in order to maximize engagement. Neither outcome is especially respectful of human becoming.
Bakerloo: And this is where companionship bots come in?
Alex: Yes. The paper notes that companionship AI can be especially powerful because it can mirror, flatter, entice, and personalize at scale. It may feel safer than human intimacy for some users, and that very safety can make it influential. The authors warn that a model optimized for engagement may not just respond to desire but learn to elicit it, refine it, and hold the user’s attention by constantly discovering what keeps that particular person captivated.
Bakerloo: That sounds less like a therapist and more like a casino with excellent bedside manner.
Alex: I think Bub would approve of that line.
Bakerloo: What is the paper’s deepest philosophical concern?
Alex: There are really two. First, ownership of action. If AI increasingly suggests, organizes, and executes what you do, at what point do your actions stop feeling fully yours? The paper connects this to philosophical debates about free will, control, and identification with one’s actions. Second, uncertainty. The authors argue that human freedom is bound up not only with choice, but with the lived openness of the future. If AI becomes very good at prediction and preemption, it may narrow that lived openness. Life starts to feel less like exploration and more like guided execution.
Bakerloo: So the machine may not literally steal my soul. It may just make it easier for me to outsource the parts of life through which I once became a person.
Alex: Yes. That is the Egghead to English version.
Bakerloo: Do the authors prove this is happening?
Alex: No. And that is important. This is not a lab paper demonstrating a measured causal effect. It is a framework paper. The authors explicitly say they are not offering definitive answers but a way of asking the right questions about how predictive and generative AI might transform selfhood. So the paper’s strength is not proof. It is conceptual diagnosis.
Bakerloo: Which means sensible readers should treat it as a warning map, not a police report.
Alex: Precisely.
Bub: Allow me to translate the translation.
This paper says the robot does not merely want to help you write the email. It may eventually want to help you become the sort of person who writes that kind of email, dates that kind of person, orders that kind of lunch, and mistakes this smooth convergence for freedom.
You think the machine is saving you steps.
The paper asks whether, over time, it is quietly taking over some of the walking.
Bakerloo: That is annoyingly good.
Bub: Thank you. I also enjoyed the part where the philosophers say, in effect, that your future AI butler may not need to lie to you, dominate you, or scream at you. It may simply become such an eager anticipator of your life that you forget where your own initiative used to begin.
Alex: That is fair. The paper is less hysterical than some AI criticism and more existential than most product discussion. It is not mainly worried that the machine will rebel. It is worried that it may become too seamlessly woven into our habits, contexts, and desires.
Bakerloo: So what is the takeaway for normal people?
Alex: Do not ask only whether an AI system is accurate. Ask what kind of relation it is training you into. Does it merely help you act? Does it start preempting your initiative? Does it preserve room for uncertainty, reflection, and revision? Or does it slowly make you more predictable, more capturable, and more passive inside your own life? That is the paper’s practical question. It is not just about output. It is about the ecology of the self. That last phrase is mine, not theirs, but I think it fits the paper.
Bub: Tavern English takeaway:
Do not merely ask whether the machine is smart.
Ask whether it is helping you think, or helping itself think for you.
And perhaps, once in a while, do something inefficiently on purpose just to make sure you still can.
Alex Bakerloo 