Research gap at a glance
- Question: Can a human paradigm measure predictions about how an internal bodily state will change under its own dynamics, as opposed to predictions that a cue signals an experimenter-imposed stimulus, and do such forecasts differ in anxiety or depression?
- Why it matters: Computational and transdiagnostic accounts locate psychopathology in mis-set predictions about the body, which presupposes a model of how bodily states unfold. That arc of the sensory-control loop has no human instrument, so the clinical theory currently rests on its least measured part.
- Theme: Mechanism
- Status: open
What the wiki already knows
- Petzschner et al. name forecasting, predicting how internal states change through action and through their own dynamics, as the arc of the regulatory loop that few frameworks have formalised, especially in the internal domain (Petzschner et al. (2021), Forecasting (internal-state prediction)).
- The cleanest existence proof is in mice: thirst neurons terminate drinking minutes before blood osmolarity changes, so satiation reports a predicted outcome (Forecasting (internal-state prediction)).
- The breathing learning task measures trial-by-trial interoceptive prediction, but participants learn that a visual cue predicts an imposed resistance; the body has no dynamics of its own in the design (Breathing Learning Task (BLT), Harrison et al. (2021)).
- Anticipation paradigms likewise ask what an arbitrary grey cue predicts about an experimenter-applied stimulus; two lineages that do not cite each other fail to reach forecasting in the same way (Wilzok et al. (2023), Discrepancy awareness).
- computational psychosomatics and Interoceptive psychopathology rest their clinical account on this unformalised arc (Homeostasis (and feelings as homeostatic sensations)).
What would answer it
- Let a bodily state evolve without experimenter control (for example thirst or hunger across a fixed deprivation interval, or heart-rate recovery after exercise), collect repeated participant forecasts of its trajectory alongside the measured physiological time series, fit a dynamic model that scores forecast error and bias, and compare anxious or depressed participants with controls.
Connections
- Raised on: Forecasting (internal-state prediction), Petzschner et al. (2021), computational psychosomatics, Homeostasis (and feelings as homeostatic sensations), Discrepancy awareness
- Related concepts: Forecasting (internal-state prediction), The sensory-control loop, Allostasis, Homeostatic reinforcement learning (HRL), Active inference
- Related debates: inferential?
The Forecasting (internal-state prediction) page describes itself as mostly a gap. Both major frameworks for body regulation, Homeostatic reinforcement learning (HRL) and Active inference, need a model of how internal states will evolve, and neither writes one down for the internal domain. Allostasis and predictive homeostasis name regulatory strategies; forecasting names the predictive machinery those strategies require, and the wiki holds substantial material on the first and almost none on the second.
The two closest human paradigms are instructive failures. In the Breathing Learning Task (BLT) and the Accuracy of Interoceptive Anticipation and Interoceptive Discrepancy paradigms, the experimenter decides what happens to the body and the participant predicts it from a cue. That is the design that gives the experimenter control of the outcome, which is why it is used, and it is also why it cannot contain forecasting: the body does nothing under its own dynamics. The wiki records the terminological hazard that follows, with “anticipation of interoceptive states” used for three different things across the clinical and computational literatures.
What a forecasting study needs is the reverse arrangement: let the body run and ask what the person expected it to do. That is harder to standardise, but states such as thirst, hunger and cardiovascular recovery have known time courses that can be recorded, which makes forecast error computable.