Why Ambiguity Is the Least of the Problems in Human Communication
Year: 2026
Type: article
Abstract: This paper argues that the imprecision of language is not the main obstacle to human communication. Before speech there is no formed meaning, only a "cloud" of unexpressed directions. An utterance does not translate this cloud into words. It imprints it: it fixes the cloud partially, disturbs it, and the speaker then takes the words as what he meant all along. Listeners mostly do not listen. They prepare their own expression and understand only through a small stock of shared formulas, the keys, which they mostly borrow from others. Ambiguity is therefore often the speaker's best option, and the decisive obstacles lie elsewhere: listeners' inattention, the narrowness of what speaker and listener share, and the impossibility of verifying that two people understood the same thing. The argument is formalized with the mathematics of unsharp measurement. The cloud is a density operator, an utterance is an unsharp Lüders instrument of a given sharpness, and keys are the commuting sector that speaker and listener share. The paper proves an information–disturbance tradeoff for self-expression, an optimal nonzero ambiguity, the convergence of a monologue to a sharp thought, a characterization of decodable content by shared algebras, and limits on telepathy and on verifying shared understanding. Further results cover the degradation of a memory by retelling it, the dominance of the first speaker, the sincere lie, rereading, the decay of shared formulas across generations, and an estimator that measures sharpness from independent readings. A second part develops a theory of the diagnostic interview, in which a human or artificial interviewer reads a person's cloud on a topic the person chooses. It proves a disturbance budget, shows that hypotheses stated before probing contaminate the answers, and shows that axis-by-axis interviews cannot see the connections between motives. It gives a stopping rule for refused forced choices, bounds the interviewer's error by the rate of off-menu answers, and requires interpretations to be sealed by cryptographic hash before the answers come in. Every interview must probe, indirectly, the four basic programs of sex, expansion, cooperation and death. A first sealed interview, between the author and a language model, is reported together with its failures. Generalizations treat mixed operational and expressive speech, with a test–retest criterion that makes the central premise refutable. They also treat group conversation, which converges to the shared sector of its participants, as well as framing power, strategic ambiguity and deniability, the avoidance of threatening directions, and synthesis, understood as the acquisition of incompatible conceptual frames. A table sorts all results by the parameters they require. Most quantitative predictions depend only on quantities that can be read from texts. The paper includes counterarguments and responses at every step, 16 empirical illustrations, a program of experiments, and appendices with Python code that verifies all theorems numerically, together with the sealed texts of the interview. Keywords: ambiguity; meaning; language and thought; predictive processing; unsharp measurement; Lüders instrument; information–disturbance tradeoff; shared knowledge; common ground; choice blindness; first-speaker effect; diagnostic interview; preregistration; strategic ambiguity; group communication; philosophy of language; artificial intelligence License: Creative Commons Attribution 4.0 International
Author: Kriger, Boris
Language: English
Subfield: Atomic and Molecular Physics, and Optics
Field: Physics and Astronomy
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7234449550
