Start here

This is a tabbed walkthrough of research by BoB and Matthew Bishop — from a 2018 dice-complexity thesis through physical falsification experiments to a causal framework for surgical coordination.

The central question: can probability and Shannon entropy alone describe how oriented physical work unfolds in an operating room? The dice experiment says no — intervention, chirality, and pre-operative feasibility (T*) had to enter the picture. Use the tabs below, or open the interactive dashboard for spinners and galleries.

From the X thread? Shared in reply to Judea Pearl's argument that Shannon must be enriched with causal levels 2 and 3. Authors: BoB Bishop (@bishopb0b) and Matthew Bishop · MICDE & ASA 2026 · jointspace.org

The Spark: Dice Experiment (Chirality Discovery)

Shannon entropy could not predict real surgical handoff times. Only repeated physical do-actions revealed chirality — left- vs right-handed configurations that entropy treats as identical but reality does not.

The Dice Game That Started Everything

I wanted to understand real work in the world. So I spun dice — many times — and wrote down what happened.

Surgery has hard steps: picking the right tool and turning it just right, every single time. I designed a game with dice to test whether starting position would predict how long each spin takes.

The old idea: Information Entropy (Shannon, 1948). I tried it many times. The old math did not match the real times.

do(spin dice study) → entropy breaks → new knowledge about chirality begins

The Discovery: Handedness

Some dice are right-handed and some left-handed. You cannot turn one into the other by spinning or sliding. Shannon entropy gave the same number to both — but they are not the same for real physical work.

Right-handed and left-handed dice — chirality

No spin or slide converts left to right. Mirror images do not match. This is chirality.

Try It: Simulate a Dice Spin

Click to perform a do(spin) intervention. Each spin is a real constructor operation — not a draw from a probability distribution.

Face 1 up · North (0°)

Elapsed: —

24 distinct orientational states exist for a surgical instrument. Only one is ★ required.

The Causal Chain — 5 do-Actions

Why causal levels matter

The dice study is a Level 2 do-action. Shannon (Level 1 association) failed. Only intervention on the real world revealed what probability alone could not.

The JointSpace Framework

We live in the JointSpaces our bodies create with the world. Every joint reconstruction is a temporary zone where human anatomy and engineered materials meet through causally ordered constructor operations.

What Is JointSpace?

In surgery, the joint space is the gap between articulating bone surfaces. Extended: any interface where an animate agent and inanimate system exchange information and force.

  • → Surgeon's hand → instrument → anatomy
  • → Scrub tech → sterile field → surgeon
  • → Operating team → procedure → restored function
JointSpace = the interface where nerve meets machine

Possible vs Impossible

The fundamental question is not how probable an outcome is. It is whether it can be caused at all.

✓ Possible
Team capacity meets procedure demand. Outcome is reproducible.
✗ Impossible
Capacity below demand. Failure is structurally guaranteed — not merely probable.

The Causal Pathway

Each step is a do() intervention — a constructor operation that propagates forward. Errors at any node propagate downstream.

Surgeon
decides
do(hand
position)
Scrub
tech
do(instrument
state ★)
Anatomy
transformed

Operating rhythm is the observable signature of this chain executing within constructor capacity. When rhythm degrades, a constructor operation has failed.

24 Rotational States of a Surgical Instrument

For each of 6 faces pointing up, there are 4 rotational positions. The scrub technician must select ★ — the one state the surgeon requires.

N (0°)
E (90°)
S (180°)
W (270°)
⚀ Face 1
1·N
1·E
1·S
1·W
⚁ Face 2
2·N
2·E
2·S
2·W
⚂ Face 3
3·N
3·E
★ 3·S
3·W
⚃ Face 4
4·N
4·E
4·S
4·W

★ = surgeon-required state. Red = physically wrong — task fails and propagates into the assembly pathway.

T* — The Pre-Operative No-Go Line

The culmination: when impossible physical movements make the planned assembly of do-operations a no-go zone.

T* is the boundary between a case this team-constructor can cause and one it cannot. Every procedure has an assembly index a (minimum causally ordered do-operations, including chirality and substrate matching). The team has constructor capacity T*(team). When a(procedure) > T*(team), the planned operation is in the no-go zone — not low-probability, impossible.

Plain example

Your left hand is not your right hand. One's L cannot be turned into R — whether palms up or palms down. No rotation. No flip. That is handedness. The same rule applies to a surgical plan, a glove, or a tray: left is not right. The plan is impossible — not unlikely.

Try it yourself. Hold your hands in front of you — palms up, then palms down. Your left hand never becomes your right hand. Everyone has two hands; everyone can see this in ten seconds.

Left and right hands — mirror images that cannot be superposed. From Wikipedia chirality article.
GO — matched handedness

Left task on left hand. Right task on right hand. Same side.

Matches ✓

NO-GO — wrong handedness

Left task on right hand. No rotation fixes it. Same logic as an L die on an R substrate.

Impossible ✗

The illustration on Wikipedia's chirality pageChirality_with_hands.svg (NASA, public domain). Flip your own hands palms down: still not the same. Dice, implants, and pre-op plans obey the same rule.

a(procedure) > T*(team) → planned operation impossible for this constructor

Catch wrong-handed plans, mismatched trays, and capacity overload before the first incision — not in complication statistics.

Why causal levels matter

JointSpace replaces probabilistic outcome correlation with a constructive, pre-operative, computable account of team capacity — at Pearl Level 2 (intervention) and Level 3 (counterfactual protocol design).

Pearl's Ladder + do-Calculus

Click each rung to see how it applies to surgical knowledge. Level 2 — do() — is where real discovery and operational decisions live.

Level 1 · See
Association

P(outcome | high-volume team) — correlation only. Confounders unresolved.

Level 2 · Do
Intervention

P(outcome | do(coordination = structured)) — cuts incoming paths. The operative question.

Level 3 · Imagine
Counterfactual

Had the team used unstructured protocol, what would P(outcome) have been?

Level 2 · Do — The Operative Level

Pearl's do() operator represents an intervention: it cuts all incoming causal paths to a variable and sets it directly. This is the formal counterpart of a constructor operation — the team does not observe the procedure, it intervenes on every step. The dice study, instrument handoff, and parity flip in imaging are all do-actions.

The do() Equation

P(outcome | do(coordination = structured)) = ?

Observational studies correlate outcomes with team experience. These are Level 1 — they cannot answer the operational question.

  • Mediator: instrument handoff state resolved per step
  • Confounder: surgeon volume — correlated but not the cause
  • Confounder: patient anatomy — adjust for, don't intervene on
Volume is a confounder. Coordination is the mediator. Only do-calculus separates them.

T* — When the Plan Is Impossible

a(procedure) > T*(team) → no-go zone

T* is where Constructor Theory meets the OR schedule: certain do-operations cannot be assembled because the physical substrates forbid it. Example: left on right — handedness mismatch — impossible, not unlikely.

Assembly index a counts required constructor steps (chirality included). T*(team) is validated team capacity. Above the line = structurally guaranteed failure.

Why causal levels matter

Constructor Theory defines what is possible or impossible. do-calculus computes the interventional distribution when the team acts as constructor. Together they enrich Shannon's Level 1 with Levels 2 and 3 — exactly as Pearl demands.

Chirality & the Shannon Critique

Shannon entropy collapses SU(2) spinorial geometry to a single axis. Correct for ensembles. Structurally incomplete for chiral constructor operations.

Shannon entropy (H = −Σ p(s) log₂ p(s)) assigns the same value to left-handed and right-handed configurations.

Constructor theory + do-calculus says: they are not the same world.

Spinorial Geometry & Handedness

Surgical instruments are chiral objects. A left-curved retractor is not a rotated right-curved retractor — it is a topologically distinct mirror image.

Chirality demands spinorial geometry (SU(2), double cover of SO(3)). A spinor requires 720° to return to its original state — not 360°. The path of rotation carries physical meaning.

✗ Impossible in 3D
Converting left-handed to right-handed by rotation alone. No constructor exists.
✓ Possible in Mind
Imagining a 720° twist. Holding both handedness states and reasoning about the gap.

SU(2) — Three Projections

Shannon Binary — SU(2) → {0,1}. No path. No handedness. Blind to chiral operations.
Deutsch Qubit — SU(2) → ℂ². Richer than binary. Still flat Hilbert space.
Causbit — Full SU(2) in composite biological tissue. N-tuple spin. Carries causal efficacy, not just information.
Shannon binary and the Deutsch qubit are projections of SU(2). The causbit is the geometry operating without reduction.

Handedness Across Every Scale

Molecular
L-amino acids
Plexus
Causbit medium
Anatomical
Laterality
Instrument
Chiral design
Team
Handoff path
Assembly
Construction

The same chirality appears at every scale: molecules, biological structures, imaging data fields, and the "window and levels" adjustments in medical images. The gap is fractal.

Why causal levels matter

Shannon measures uncertainty over symbols (Level 1). Chirality and constructor operations require intervention on structured substrates (Level 2) and counterfactual reasoning about parity (Level 3). The dice experiment is the existence proof.

Implications & Next Steps

From MICDE predictive science to ASA imaging methods to patient-sovereign health records — the framework generalizes wherever causal structure is legible.

T* — why the whole framework matters

Dice chirality → instrument handedness → plan vs substrate matching: all the same possible/impossible dichotomy. T* is the pre-operative test: is this planned sequence of do-operations physically possible for this team on this patient? If not, the case belongs in the no-go zone — do not treat impossibility as bad luck.

Pre-Operative Team Assignment

Not "is this surgeon experienced?" but "does this team's constructor capacity exceed T* for this specific chiral procedure on this specific anatomy?"

Protocol Design

Operating rhythms must keep every constructor step within demonstrated capacity. Left and right procedures require distinct rhythm protocols — different assembly graphs.

AI Coordination

Detect in real time when a team's constructor capacity approaches T* — and signal before the causal chain breaks. Not predicting the next instrument. Identifying the failure mode.

Validated Causal Identity

A team's T* derives from a validated record of procedures completed within assembly demand — a causal depth profile, not a credential or volume count.

Fractalscope for Imaging

A "window and levels" tool for parity and assembly state in fractal spatial data. The CHIRAL FLIP toggle is a real do() on data structure — making what Shannon averages away visible to statisticians.

JointSpace.org for Patients

Patient-sovereign lifetime health records that preserve the actual intervention sequence — because bodies have specific structure that averages cannot reconstruct.

Generalization Beyond Surgery

Any multi-agent operation with a causally ordered construction graph — emergency medicine, aviation, manufacturing — has a computable assembly index and a T* boundary. Surgery is the domain where causal structure is most legible and chirality most visible.

2018 → 2026 — Matt's Thesis to Pearl

Matthew Bishop's UMass Boston undergrad thesis (Modeling of Surgical Complexity: A Dice Study, May 2018) grew from running an ambulatory surgery center with Bob. Matt used R and standard statistics competently — but the fundamental premise of the statistical approach was never counterfactually challenged until Pearl.

Thesis finding (associational — Rung 1)

lm(Time ~ StartState + Moves + DiceCount) + ANOVA. Highly significant predictors. Practical signal: 8 dice proximate ~2.5 s vs shuffle roll ~24.8 s — ~10× faster. Thesis concludes: "This is where we want to go!" — an implicit tray redesign recommendation without a formal do(proximate layout).

Download thesis PDF

What undergrad stats asked

  • Is start state associated with time?
  • Null hypothesis + p-values + ensemble means
  • Never: is regression the right tool to license intervention?

What Pearl permissions (2018–2026)

  • P(time | do(proximate layout))
  • Chirality, spin orientation, T* no-go zone
  • Right implant on left femur = impossible, not unlikely

Why Pearl wasn't in Matt's stats curriculum

Fisher/Neyman–Pearson heritage; Book of Why published the same month as the thesis; lm() encodes associational questions; causal inference was grad-level, not UMass Boston undergrad core. Matt had rare ASC ground truth — statistics never matched that embodied knowledge.

Full timeline, empirical photos, video, and interactive arc on the dashboard version:

Open jointspace-interactive →