One Direction Under Pressure: A Pre-Specified Test of Mindsponge Theory's Single Effort-Share Prediction in Repeated Risky Choice
Description
Mindsponge theory (MT) holds that a mind weighs new information through two routes, one that leans on what it already trusts and one that evaluates the content itself, and that a single effort share sets the balance between them. When evaluation becomes costly, the share falls. If one dial really sets the balance, then every change in behavior caused by rising cost must move along one fixed direction. This is prediction P1b of the MT Testing Program, and it had not been tested before. We tested it on open data from an experiment in which 73 people made 324 choices each between a safe and a risky lottery under three levels of time pressure. The decision rule, an admission simulation on the real design, and one declared revision were fixed and fingerprinted before any choice was read. As pressure rose, sensitivity to expected value fell from 0.69 to 0.25 and the lean toward each person’s own habitual option rose by 0.33 logit, so that people chose their habitual option about 80% of the time under high pressure against 75% under low pressure. The two successive shifts were parallel, as P1b requires: their cross product was 0.002 (bootstrap SE 0.030, z = 0.08), and a likelihood-ratio test of the one-direction constraint found no loss of fit (LR = 0.01, p = 0.92). The prediction was supported. The result held in model-free summaries, with time on task controlled, in an independent re-fit, and with any single participant removed. It rules out accounts in which pressure moves value sensitivity and habit independently, and it is inconsistent with a pure loss of precision. It does not separate MT from one-parameter resource-rational accounts, which predict the same structure. For MT, this is the first confirmation of a structural equality derived from its formal core, rather than of a direction of change alone.