Biological Connection and the Simulation of Language

Communication, understood biologically and phenomenologically, is not an information-transfer protocol but a dense, multi-modal process through which organisms co-regulate, establish trust, and sustain the relational fabric of life. Alongside semantic content, it encompasses prosody, haptics, olfactory signaling (including chemosensory honesty cues), autonomic synchrony, and the micro-temporal coordination of gaze and gesture. These channels are not supplementary to meaning-making — they constitute its substrate. Strip them away and what remains is not a thinner version of communication; it is a categorically different phenomenon.

Large language models produce output that is fluent at the verbal-semantic surface while being structurally incapable of participating in the embodied, co-present, physiologically coupled process that the above channels require. The concern this raises is not straightforwardly one of deception. Human relational systems did not evolve to adjudicate the Ontological status of their interlocutors; they evolved to respond to signals. When enough of those signals are present — responsiveness, apparent attentiveness, tonal warmth — Attachment and affiliation systems activate, regardless of whether the fuller relational process is actually occurring.

This produces a specific kind of deprivation that is difficult to detect precisely because the activating signals are present. The Harlow analogy is instructive: the surrogate mother was not a deception in the ordinary sense, but its capacity to engage infant Attachment behavior without satisfying the developmental needs those behaviors exist to serve produced measurable harm. The hypothesis here is that sustained AI-mediated interaction may function analogously — providing sufficient signal to engage human relational systems while the co-regulatory, embodied, and physiologically coupled dimensions of genuine relationship remain unmet.