Why We Care About Understanding: Competence through Predictive Compression

Summary

This 10,400-word conceptual paper asks why understanding and compression are so tightly linked and why human understanding takes a heavily compressed form. Against the compressionist tradition (Chaitin, Hutter, Schmidhuber, Delétang et al.) that identifies comprehension with compression, and a philosophical literature that barely mentions it, the authors advance three theses. First, the Proxy Hypothesis: the concept of understanding is an efficient proxy for robust competence, letting us identify whom to trust and whom to learn from. Second, to understand a domain is to possess a structure-sensitive mental model enabling prediction; predictable data need not be stored separately, so compression is the 'representational shadow' of comprehension, not its identity. Third, social pressures — demonstrability and transmissibility, joined to predictivity and storability — drive human understanding toward principled simplicity, formalized as a minimum-description-length objective L(D|M)+λL(M) and mapped into nascent, schematic, elegant, and opaque understanding. Objections (zip compression, stenographers, Sheffer-stroke logic; tacit phronesis) are answered by asking which structure a compression reflects. Limitations: the argument is purely conceptual, with no empirical support for the genealogical Proxy Hypothesis; 'mental model' spans explicit and tacit forms, risking unfalsifiability; λ is never operationalized; and LLM classifications are asserted rather than tested, with the relevant empirical work cited only in footnotes.