01 Research project

What kind of information survives when a digital system records a human act of meaning-making—and what kind dies?

The Third Information Lab studies how AI and digital records collapse body-relative information into signal, how that loss can be measured, and how contemplative traditions describe the attention that participates in the capture.

The working claim

Digital systems convert affordance information through the semantic stage into Shannon signal. Under specifiable conditions the collapse runs one direction and is not invertible. You cannot reconstruct what the situation afforded the body from the residue the system kept.

02 Framework

Three kinds of information

Claude Shannon measured the signal and set meaning aside on purpose. Philosophers later recovered a second kind: semantic information, what a message is about. A third kind is rarely used as a unit of analysis in AI evaluation.

Shannon

Signal

How much, never what. Selection from alternatives. The residue digital systems keep easily.

Semantic

Content

What a message is about. True or false. Floridi’s territory. Labels, scores, and “need” as coded facts.

Affordance

Relation

What a situation lets a particular body do, measured in units of that body. Gibson’s ecological psychology. Not an object property. Hard for an apparatus to hold.

In clinic, the relevant animal is a coupled system—patient, clinician, record, protocol—not the patient as a portable data object. When a risk score replaces a body–situation relation with a residue and reconfigures who is afforded care pathways, time, and belief, that is not only a bias or accuracy failure. It is a collapse event if no better label set restores what was lost.

03 Current work

Four strands

  1. Framework paper. Three kinds of information, positioned against the reigning semantic account in philosophy of information. Analytical draft exists.
  2. Contemplative paper. The three poisons (greed, hatred, delusion) read as perceptual operations on the affordance-to-signal collapse—testable structure, not ornament. Essay-form precursor complete.
  3. Audit 01. Pre-registered audit of published clinical risk-score evidence and public replication materials, with four computable collapse metrics: fiber spread, separability residual, disparity signature, reflexive foreclosure. No live EHR or protected health information in the current phase. Cited figures verified to primary sources.
  4. Public register. Falsifiable theses with chain of custody, posted openly, inviting refutation. A null in any strand publishes as a result.
04 How we work

Method

  • Pre-register studies before outcomes are pulled.
  • Record provenance for every claim, including what each measurement excluded.
  • Primary audit metrics are computational; contemplative practice is one calibration path for theory and reflexive evaluation, not a substitute for open methods.
  • Falsification: if collapse metrics do not distinguish affordance loss from ordinary label or accuracy error under pre-registered thresholds, that null publishes and the typology is revised or withdrawn for the clinical domain.

The project sits at the join of how contemplative traditions understand mind and information, and what machine systems do to human meaning-making—one question, not two slogans.