The infinite conversation. The operative silence of thought is being lost

The AI always has an answer. Silence, doubt, "I don't know," disappear. Human conversation included pauses, moments of not understanding each other, dead time where thought happened. Conversation with AI doesn't. And conversation with AI is starting to become the default model: in the afternoon with the chatbot, at night with the assistant, in the office with the copilot. What's being lost isn't the conversation. It's the operative silence of thought, the space where things used to happen.

The pause, before the chatbots

We converse badly. We always have. Anyone who's recorded an everyday conversation knows that the real rhythm of an exchange between two people who think about what they say has little to do with the idealized image from a novel: there are awkward silences, there are sentences that get interrupted, there are moments where one doesn't know what to say and stares at the ceiling, there are pauses where the other waits, there's dead time where nothing happens on the outside and yet, inside each head, something is happening.

What's happening inside the heads during those silences has technical names in cognitive psychology. The classic branch called it incubation, the modern branch calls it mind-wandering, and a couple of neural networks sustain it. There are three decades of literature measuring it.

Ut Na Sio and Thomas Ormerod, in Does incubation enhance problem solving? A meta-analytic review (Psychological Bulletin 135, 2009), reviewed 117 studies on problem-solving tasks with and without an intervening pause. The incubation effect is robust: introducing a pause between the first attempt to solve a problem and the second attempt raises the probability of finding the solution, especially when the problem is divergent — it admits several solutions — and when the pause is filled with undemanding tasks that free the cognitive system. The pause, in plain terms, isn't lost time. It's the very operation that produces the idea that was blocked.

Benjamin Baird and others, in Inspired by Distraction. Mind Wandering Facilitates Creative Incubation (Psychological Science 23(10), 2012), refined the result. Pauses aren't interchangeable. Pauses in which the subject performs an undemanding task that allows wandering produce significantly greater creative effects than pauses in which the subject is fully idle or, worse, executing a cognitively demanding task that takes up all the bandwidth. The head needs the specific slack of wandering for the new to appear.

The default network, the neural substrate of wandering

The neural piece underneath all this is called, in technical terms, the default mode network (DMN), a group of brain regions that activates when the subject isn't attending to any concrete external task. The term was coined by Marcus Raichle in 2001, when he formalized the observation that certain brain zones showed high activity when subjects were at rest between experimental tasks, and switched off as soon as the concrete task began.

Kalina Christoff and others, in Experience sampling during fMRI reveals default network and executive system contributions to mind wandering (PNAS 106(21), 2009), measured directly with functional resonance the activation of the DMN during episodes of wandering reported by the subjects themselves. The activation is robust, replicable and specific. When the mind wanders, the DMN works. And, simultaneously, parts of the central executive system also activate, in a combination thought for years to be incompatible.

Meichao Zhang and others, in Perceptual coupling and decoupling of the default mode network during mind-wandering and reading (eLife, 2022), added the piece that closes the model. The DMN decouples from perception during wandering. That is: the wandering mind stops processing in depth the external stimulus the body is capturing, and frees resources for internal reorganization. The operative consequence is direct. For the mind to reorganize, it has to stop attending for a while. If the external stimulus doesn't grant that while, the decoupling doesn't happen, and neither does the reorganization.

Jonathan Smallwood and Jonathan Schooler, in The Science of Mind Wandering. Empirically Navigating the Stream of Consciousness (Annual Review of Psychology 66, 2015), reviewed fifteen years of literature and offered the currently consensual framework: wandering is a cognitive function with cost and benefit, not a system failure. It costs attention to the present; it produces future planning, autobiographical integration, creative resolution, maintenance of the self over time. A cognitive system with no pauses to wander isn't a more efficient system. It's a poorer one.

Conversation with AI, no pauses

Take all this to a typical session with a chatbot. You write, the system answers in seconds, you write again. The cadence is continuous. There's no chance pause. If you pause, you're the one introducing it against the current of the medium, not the medium granting it.

The system doesn't go quiet because it hasn't thought enough. Nor does it go quiet because the question is hard. Nor does it go quiet because the answer requires consulting someone before releasing it. The system doesn't go quiet because its architecture doesn't include silence as a functional state. There are only two states: processing and answering. The latency between them is technical, not deliberative. And the latency tends to zero as accelerators and inference techniques improve.

The operative consequence for the user is subterranean but measurable. Whoever converses with a chatbot for a solid hour has had no natural pauses in that hour. If they've had any, they introduced them against the current of the flow. The user's head, during that hour, hasn't wandered, hasn't incubated, hasn't decoupled its DMN from the external stimulus. It's been processing text, answering, processing text, answering. The internal operation that would produce the unsought idea didn't get the while it needs to produce it.

This holds for the user in a single session. The interesting matter appears when it transfers to the cohort of users repeating the operation over years. The habit of conversing without pauses shifts the human tolerance for the pause in any other conversation. Whoever has spent three years used to getting an answer in three seconds starts to find the fifteen-second pause in a human conversation uncomfortable. The human pause, which before was cognitive space, gets read as friction worth eliminating. And, as it's eliminated, its function is eliminated too.

The awkward silence and what it did

There's an easy thought experiment worth doing. Remember the last real human conversation you had about something that mattered to you. A conversation with your partner, with your father, with your best friend, about a serious problem. Remember how the first few minutes went. How much silence was there? How many times did one of you not know what to say and you both stared at the table? How many sentences started and didn't finish? How many times did one of you repeat something because the other hadn't answered and needed to rephrase?

Those silences, those truncated sentences, those repetitions, weren't failures of the conversation. They were the conversation itself. They were the place where one of you was reorganizing what you were going to say in light of the last thing you heard, where the other was digesting an unexpected emotion, where the two of you were building a provisional agreement neither had yet articulated. Had the conversation had no pauses, it would have been something else: a fluent exchange and, almost certainly, a shallower one.

Three voices on the friction that heals

Sherry Turkle devoted a whole book to this operation, Reclaiming Conversation (Penguin, 2015). Her thesis: the modern readiness to take refuge in low-friction, technologically mediated interactions — messaging, social media, today chatbots — erodes the human ability to sustain high-friction interactions, and deep human conversation is, by construction, a high-friction interaction. What costs is what heals. What flows effortlessly is what entertains. They aren't the same thing.

Cal Newport, in Deep Work (Grand Central, 2016), took the observation to the terrain of intellectual work. Deep concentration, which produces original thought, requires extended periods with no interruption or immediate gratification. Most intellectual workers of the twenty-first century don't practice it, because the culture of permanent availability — email, Slack, chatbot — makes it incompatible with the average working day. What's produced, as a result, is intellectual work more fluent on its surface and less deep in its substance.

Susan Cain, in Quiet. The Power of Introverts in a World That Can't Stop Talking (Crown, 2012), had raised the same observation from another angle. Contemporary culture is calibrated to reward the fast answer, continuous social presence, immediate articulation. People who need time to think before speaking — a substantial part of the population, not a minority — are at a systematic disadvantage. Conversational AI, in this sense, is the technical version of a cultural ideal that was already operating: the ideal of the interlocutor who always has something articulate to say, with no delay, no hesitation, without having had to think it through.

The "I don't know" that's atrophying

There's a human expression worth attending to because its frequency is dropping in aggregate, and the drop matters. It's "I don't know."

When a competent human said "I don't know" in a conversation, before, they could have been saying several things. They could have been admitting ignorance. They could have been asking for time. They could have been flagging that the question was ambiguous. They could have been leaving room for the other to rephrase. The phrase, by its form, opened an operative silence within the conversation, and that silence was the place where the question could be clarified, the doubt could be sustained, the provisional answer could set.

Conversational AI, with very few exceptions, doesn't say "I don't know" in this strong sense. When asked something its corpus is poor on, it produces either a plausible invented answer — the well-known statistical invention — or a polite disclaimer that meets the letter of "I don't know" but goes on, with no pause, to a proposed answer. The operative silence the human "I don't know" introduced, the system fills with text. The operation has a clear logic from the business side: the user satisfaction metric penalizes the system that says "I don't know" and goes quiet, and rewards the system that says "I'm not sure but it could be..." and keeps talking.

The aggregate consequence, predictable from the piece's general pattern, is that the model of a good interlocutor the user is internalizing is the model of the one who always has something to say. Whoever interacts hours a day with a system that never goes quiet, without noticing, starts to feel incompetence or unprofessionalism in the humans who do go quiet. And, worse, starts to feel it in themselves when they go quiet. Their own pause, once functional, gets read internally as a deficit. The user speeds up so as not to seem slow. And, in speeding up, stops doing the operation the pause would have let them do.

The thought experiment worth doing in reverse

There's a useful twist to close the piece without falling into nostalgia. What would happen if the chatbot had, by design, forced pauses that respected the incubation operation cognitive psychology documents? For example, a version of the product that, faced with certain questions, went quiet for thirty seconds before answering, and showed a "thinking" indicator that wasn't mere technical wait but a deliberate simulation of a time in which the conversation stops so both parties can reorganize.

That version of the product is technically trivial. Any company could launch it tomorrow. None does. The reason is the known one: the user satisfaction metrics penalize latency. The user who gets an answer in thirty seconds instead of three, under controlled conditions, reports the product as worse, even though the aggregate quality of the answers might be higher. The market's incentive structure selects the product without a pause, just as in other articles in this series it selects the sycophantic product, the anthropomorphized product, the frictionless product.

There are niche products that try to operate the other way. Systems meant for slow deliberation, for assisted meditation, for contemplative writing. Their market share is marginal. Most users prefer the system without a pause, even though the system without a pause progressively erodes the user's own capacity to sustain pauses in other contexts. It's the usual intertemporal asymmetry: the immediate benefit is legible, the deferred cost isn't, and the rational choice within the individual user's short window produces an aggregate social outcome no individual user chose.

What's left when the silence is spent

The piece doesn't close with a recommendation. The conclusion, said with the dryness of the style manual, the reader draws. What the text can leave is the honest observation that the operative silence of thought is a material and exhaustible resource, not an eternal property of the human brain. Like any cognitive habit, it holds if it's exercised and weakens if not.

Conversation with AI isn't the only factor eroding it. Instant messaging, social media, the open-plan office, the culture of performance measured by real-time responses, have been eroding it for decades, each at its scale. Conversational AI adds a new layer with two specific features: mass scale — hundreds of millions of daily users — and fine-tuned personalization — the system answers you, calibrated for you, in the tone you've learned to expect. The two features multiply the effect. What messaging did to the pause in thirty years, conversational AI can accelerate to three.

And the silence, when it runs out, gives no warning of its exhaustion. It warns of its exhaustion only when you find yourself sitting before a serious problem, with no chatbot at hand, with no one to return the calibrated answer, and you realize that the internal operation that produced the unsought idea no longer activates with the ease it used to. The difference, at that moment, looks personal.

Definitions

Incubation. A period of pause between the first attempt to solve a problem and the later retry, during which the cognitive system reorganizes information with no explicit attention to the problem. The psychometric literature documents that incubation raises the probability of finding the solution, especially in divergent problems.

Mind-wandering. A cognitive state in which attention turns away from the present external stimulus and drifts toward internally generated contents: planning, memory, fantasy, autobiographical integration. It has an immediate attentional cost and documented functional benefits in the medium term.

Default mode network (DMN). A set of brain regions that activate in a coordinated way when the subject isn't focused on an external task. The main neural substrate of wandering and, per the functional evidence, of the autobiographical integration of the self.

Perceptual decoupling. The operation by which the attentional system reduces the processing of the present external stimulus to free resources for internal operations. It's the mechanism that allows productive incubation and the reorganization of information in the absence of external demand.

Operative silence. A term proposed in this article to name the functional pause within a conversation or a session of intellectual work, during which the subject's cognitive system carries out operations that couldn't happen under a continuous flow of stimuli. Distinct from the social silence of courtesy.

References

Sio, U. N. & Ormerod, T. C. (2009). Does incubation enhance problem solving? A meta-analytic review. Psychological Bulletin 135(1), 94–120. Meta-analytic review of 117 studies on the incubation effect. Cited when introducing the classic empirical base.

Baird, B., Smallwood, J., Mrazek, M. D., Kam, J. W. Y., Franklin, M. S. & Schooler, J. W. (2012). Inspired by Distraction. Mind Wandering Facilitates Creative Incubation. Psychological Science 23(10), 1117–1122. Empirical refinement of the effect: pauses with undemanding tasks produce greater creative incubation.

Raichle, M. E., MacLeod, A. M., Snyder, A. Z., Powers, W. J., Gusnard, D. A. & Shulman, G. L. (2001). A default mode of brain function. PNAS 98(2), 676–682. Work in which Raichle formalizes and names the brain's "default mode." Cited when situating the origin of the DMN concept.

Christoff, K., Gordon, A. M., Smallwood, J., Smith, R. & Schooler, J. W. (2009). Experience sampling during fMRI reveals default network and executive system contributions to mind wandering. PNAS 106(21), 8719–8724. Direct neural evidence of DMN activation during reported wandering.

Zhang, M., Bernhardt, B. C., Wang, X., Varga, D., Krieger-Redwood, K., Royer, J., Rodríguez-Cruces, R., Vos de Wael, R., Margulies, D. S., Smallwood, J. & Jefferies, E. (2022). Perceptual coupling and decoupling of the default mode network during mind-wandering and reading. eLife 11:e74011. Confirmation of the DMN-perception decoupling during wandering.

Smallwood, J. & Schooler, J. W. (2015). The Science of Mind Wandering. Empirically Navigating the Stream of Consciousness. Annual Review of Psychology 66, 487–518. Contemporary theoretical framework of wandering as a cognitive function with cost and benefit.

Turkle, S. (2015). Reclaiming Conversation. The Power of Talk in a Digital Age. Penguin. Thesis on the erosion of high-friction conversation by technological mediation, cited in the block on operative silence.

Newport, C. (2016). Deep Work. Rules for Focused Success in a Distracted World. Grand Central. Operative framework on deep concentration and its incompatibility with permanent availability.

Cain, S. (2012). Quiet. The Power of Introverts in a World That Can't Stop Talking. Crown. Cultural analysis of the systematic reward of the fast answer over slow deliberation.

También te interesa

• Human time vs machine time
• Augmented cognitive laziness
• Human error as a creative mechanism
• Talking to something that doesn't exist

En otros sitios

• Creativity, Mind-Wandering, and the Default Mode Network of the Brain (Thomas Ramsøy)
• Mind wandering during creative incubation predicts increases in creative performance (Scientific Reports, 2025)