Love, The Cosmic Dance

August 12, 2025August 12, 2025

He loves her.

That is the shortest statement I can make about the universe. Everything else in these pages unpacks that sentence. “He” is the Unknowable Future, the open possibility that forever surrounds us. “She” is the Immutable Past, the completed record that cannot be edited. Between them is a stage on which a dance becomes visible to us.

Imagine a theatre. At center stage stand two tall, coaxial glass cylinders sharing a common axis—one facing “east,” the other “west.” Their clear walls extend without end; their interiors remain perfectly inviolate. The atmosphere—the weather of ideas, incentives, fears, and fashions—touches only the outer surfaces. Nothing breaches the interior. She dwells at the still center; He surrounds and protects. The audience—the Seat of Witness—faces the stage. You are free to step into the scene as a Participant, but you are also always the Witness, the one who sees without danger and returns home after the show. The performance is real, but it is not a threat.

The dance is quantitative. At each reveal the world presents us with a pair of numbers:

• Actual, A: what just happened. It is scalar, positive, and past-tense.
• Expected, E: what our system thought would happen a moment earlier.

With these two we form the Reality ratio:

Reality = Actual / Expected.

Do not be fooled by the simplicity. The ratio is not a metaphor. It is a measurement in the strictest sense—a comparison of two commensurate quantities. And it carries the right phenomenology. When A equals E, the moment feels unsurprising and frictionless. When A is larger than E, the moment lands as a satisfying “ah!” When A is smaller, there is a faint “oh…” This is the Witness’ vertical line: the theatre’s pure up–down response.

To read this signal on a human scale we take a natural log:

S = ln(A/E).

The log turns multiplicative gaps into additive increments, so surprises accumulate cleanly across time the way your body remembers them. S = 0 means “no surprise.” S > 0 is pleasant. S < 0 is unpleasant. This is the OO–AAH axis. It is the only axis the Witness needs.

But why did we expect E in the first place? Because “Expected” is the meeting point of two unconscious processes:

The Predictor, P. It is the brain’s continual, one-step-ahead model trained on your lived past. P is gritty and local: Friday traffic, recurring delays, sales tax, the way humidity ruins an experiment.
The Ideal, I. It is the contribution of ideas—the context-blind template that says “how things are supposed to behave” in a perfect market, a frictionless flow, a clean vacuum.

We blend them multiplicatively (think log-space averaging): E = P^{1−λ} I^{λ}, where 0 ≤ λ ≤ 1 reflects how strongly idealization is steering you in this domain. In a lab that never quite achieves a vacuum, λ is modest; on a whiteboard proving a theorem, λ may be large. Ideas belong here, as servants. They are excellent at shaping I. They become tyrants only when they attempt to overwrite the reveal.

There is a second quantity that appears before the reveal: Desire. Desire is not a claim about what will happen; it is a tilt in what we would like to happen. Around E we parameterize it as

D = E · e^{s m},

where s ∈ {−1,0,+1} encodes direction (a wish for “lower,” “neutral,” or “higher” than E) and m ≥ 0 is the intensity of the wish. Notice what this does and what it does not do. Desire adjusts the center of gravity of your attention before actuality arrives; it does not change A.

After the reveal, a decision occurs. If we release D and update to A, we remain the Witness. If we keep consulting D instead of A, we enter the Participant’s horizontal load. The measurable quantity is

K = ln(D/A).

K = 0 is acceptance. |K| is the amount of clinging. If you want a smooth, always-nonnegative “energy of suffering,” take (1/2)K². This is the off-axis drift: not more real than S, simply orthogonal to it. The Witness lives on the vertical line K = 0. The Participant suffers when K ≠ 0.

Here are two classroom curtains that rise cleanly and quickly.

Price-is-Right. I hold up a lamp.

Your predictor, P, has seen thousands of prices. It knows about tariffs, sales tax, holiday spikes; it even remembers that one boutique on the corner that overcharges. The ideal, I, is the perfect-market template: no scarcity, no rent-seeking, no frictions. Their log-space blend gives your expected price E. Suppose E = $250. I reveal the tag: A = $300.

The Witness’ vertical read is S = ln(300/250) = ln(1.2) ≈ +0.182, a small pleasant lift if you are the seller, a small unpleasant sag if you are the buyer—but either way, a clean update.

Now watch for horizontal load. If a fairness-tilt had you wishing beforehand for a much lower price, say D = $185, and you keep that wish after the reveal, K = ln(185/300) ≈ −0.483. That number is not a moral verdict; it is the magnitude of the sideways pull you will now have to manage in your speech, posture, and choices. Release the wish, and K goes to zero. You are back on the vertical.

Navigation. “How long from Lockwood to Coleman at 4:15 p.m. on Friday?”

Your predictor P knows the drawbridge schedule and the school pickup wave. The ideal I knows only free-flow. The blend says E = 24 minutes. The reveal is A = 32 minutes. Surprise: S = ln(32/24) = ln(4/3) ≈ +0.287 (an “oh…”). If you now insist, “It should be eighteen,” your D = 18 min and K = ln(18/32) ≈ −0.575. The traffic will not yield to your wish; only your posture can.

These numbers let us write six clear sentences.

Everyone has desires. D = E · e^{s m} with m ≥ 0 is part of being a living predictor with aims.
Suffering appears when you cling to a desire after the reveal. That is, when K ≠ 0.
Clinging is operationally identical to ignoring Actual. You are replacing A with D in your judgments and actions.
The amount of suffering is quantifiable. Use |K| or (1/2)K².
“Desire is unrealistic.” Reality uses A/E. Desire becomes a problem only when it tries to be a numerator.
The root cause is an idea trying to actualize. As a servant it improves I; as a master it demands A = I.

Notice what this does for the old philosophical tangles.

Time acquires a simple shape. A is past, E is the best one-step forecast of the next moment, and S is the felt shock of their ratio. We can aggregate surprises across a sequence by summing logs: the body’s memory of days is an additive history of multiplicative mismatches. The arrow points forward because A is write-protected: once revealed, it only updates E; it never re-opens.

Ignorance gains a crisp meaning. It is not the absence of information; it is the refusal to let A update your map. In symbols, ignorance is the insistence on K ≠ 0 after the reveal. When the Buddha diagnoses craving and ignorance, he is describing exactly this horizontal persistence of D when the vertical update would suffice.

Ideas reclaim their rightful dignity. They are not enemies. In the denominator they are superb servants: they give us I, the clean line that makes learning possible, and they lend courage to the predictor P when data are sparse. As numerators they are tyrants: the same structure that yields beauty on a whiteboard yields fanaticism in the street. The discipline is architectural—keep ideas in the blend that forms E, then let A speak.

And love? The theatre is love’s geometry. He surrounds; She completes. The cylinders guarantee two things simultaneously: (1) the interior remains untouched—no weather can harm Her—and (2) the surface is open to the drama of becoming. Because the Witness knows it is a performance, fear never becomes the price of participation. You can enter the scene, feel the OO–AAH of S, notice any off-axis K, and step back to the vertical without losing your seat.

The rest of this book builds the calculus of that last paragraph. We will learn how P updates, how λ is learned, how ideas cooperate or compete inside I without seizing the throne, and how a human being can live as both Participant and Witness with mathematical clarity and moral tenderness. For now, keep three pictures in mind:

A fraction that reads the world (A/E).
A vertical line that sings (S).
A horizontal nudge you need not obey (K).

He loves her. The show goes on.

Sequences, Learning Denominators, and the Two Axes of Experience

This chapter extends the book’s opening by formalizing time as sequences, the denominator’s learning rule, the role of ideas, and the vertical (surprise) and horizontal (suffering) axes. Equations are rendered as white SVGs for dark mode.

1) Sequences, and why time points forward

Experience arrives as a list, not a loop. Write it as a sequence of pairs ${(A_t,E_t)}$ where each expected value $E_t$ is produced before its reveal $A_t$ , and each $A_t$ is write-protected once revealed. That asymmetry—forecast first, reveal second—is the arrow of time.

Surprise aggregates additively in log space:

S_{1:T} = sum ln(A_t/E_t) = ln product A_t/E_t

Two derived views:
• Reality sequence: ${A_t/E_t}$ (domain units divide out).
• Surprise sequence: ${ln(A_t/E_t)}$ (the body’s register: positive lifts, negative sags, zeros as breath).

2) How the denominator learns: Predictor ⊕ Ideal

The denominator combines a context-aware predictor $P$ (Friday traffic, sales tax, humidity) with a context-blind ideal $I$ (free-flow travel, perfect market, vacuum experiment). Blend multiplicatively:

ln E = (1-λ) ln P + λ ln I ⇒ E = P^{1-λ} I^{λ}

Servant vs. tyrant. Ideas belong here as servants shaping $I$ . They become tyrants only when they try to replace the reveal, demanding $A=I$ .

Horizon forecasts without breaking the one-step rule

For a horizon $h$ , index the same rule:

The end-of-history illusion appears as over-weighting recent regimes (small effective past) and an optimistic/utopian bias in $I$ , yielding conservative $E(h)$ but adventurous desires.

3) Ideas in their right place: circle bearings, cooperative servants

Keep four motifs as bearings on a unit circle—fairness, hierarchy, significance, symmetry. Your “idea weather” is a circular mean with spread. That mean informs $I$ (hence $E$ ). Healthy when serving; harmful when seizing the numerator.

4) Desire as design: keep the energy, lose the sting

Before the reveal, encode a motive as a tilt around Expected:

$s=-1$ wants lower (fairness-tilt), $s=0$ neutral, $s=+1$ wants higher (hierarchy-tilt); $m$ is intensity.

After the reveal: sting vs. design

The horizontal load (sting) is

$K=0$ is acceptance; $|K|$ quantifies clinging; a smooth “energy of suffering” is $(1/2)K^2$ . Convert the same $(s,m)$ into design by steering the next $E$ instead of relitigating the past $A$ .

5) The Witness/Participant protocol (R → U → A)

Reveal (R): name $A$ plainly.
Update (U): feel the vertical OO–AAH via $S=ln(A/E)$ ; let $E$ learn from $A$ .
Act (A): check $K=ln(D/A)$ . If nonzero, say “Advise my next move; you don’t edit the past,” and send $K→0$ .

6) Two curtains, revisited

Price-is-Right

Predict with lived market data ( $P$ ) blended with perfect-market template ( $I$ ) to form $E$ .
Witness the lift/sag via $S$ .
If a wish (e.g., “It should be lower”) persists after reveal, measure the sideways load with $K$ and convert it into next-step design.

Navigation

Predict with Friday patterns ( $P$ ) and free-flow template ( $I$ ).
When the bridge goes up and $A>E$ , your vertical axis says “oh.”
Use the motive embedded in $D=Ee^{sm}$ to change departure time, route, or tools—not the already-revealed $A$ .

7) Ignorance, precisely

Ignorance is not missing information; it is refusal to allow $A$ to update the map. Formally: after the reveal you keep $K≠0$ . The remedy is architectural—return to vertical—then let the denominator learn.

8) Three compact theorems

Arrow Theorem. If numerators are immutable and denominators learn only forward, time has a direction and surprise is additive: $S_{1:T} sum identity$ .
No-Fear Theorem (Witness). On the vertical line $K=0$ , no matter how large $|S|$ becomes, suffering remains zero (e.g., energy $(1/2)K^2=0$ ).
Tyrant Theorem. Any idea that demands $A=I$ generates $K≠0$ whenever $A≠I$ .

9) Pocket summary

A fraction that reads the world: $A/E$ .
A vertical song you are built to feel: $S=ln(A/E)$ .
A horizontal lever you can drop: $K=ln(D/A)$ .
Ideas as servants in $E$ , not rulers of $A$ .
Desire as fuel for design, not as an editor of the past.

Updating the Denominator, Learning Influence, and Composing Ideas

We now make the denominator concrete (how P learns), show how to learn the ideal’s influence $λ$ , and define a principled way to compose multiple ideas into a single bearing used by the ideal I. Throughout, the Witness’ vertical signal is $S=ln(A/E)$ , and the Participant’s horizontal load is $K=ln(D/A)$ .

1) Making the predictor P concrete

1.1 Exponential moving average (EMA)

For many everyday domains, a one-number forecaster with fading memory suffices. Work in log space so multiplicative effects become additive. Let

Define the log-EMA with smoothing $α∈(0,1]$ :

Then $P_t = exp(p_t)$ . Use domain knowledge for $α$ (larger for volatile contexts).

1.2 A volatility-aware EMA (“Kalman-lite”)

Track spread with an EMA of squared innovations:

r_t = x_t - p_t , v_t = (1-β) v_{t-1} + β r_t^2

A normalized surprise helps separate “big but expected” from “truly shocking”:

Tip. Pick $ε$ small (e.g., $1e-8$ ) to avoid division by zero.

1.3 Seasonality and regime changes

When context matters (e.g., “Friday 4:15pm”), maintain separate EMAs per context key, or weight past by a context kernel $w(Δ)$ (e.g., exponential decay in time + categorical bumps for weekday & hour).

2) Blending with the ideal I and learning $λ$

We blend $P$ and $I$ multiplicatively:

To learn $λ$ online, minimize squared surprise $L = S^2$ , where $S = ln(A/E)$ . Let $a=ln A, x=ln P, y=ln I$ . Then

S = a - [(1-λ)x + λ y] = (a - x) - λ (y - x)

Gradient step (clipped to $[0,1]$ ):

Interpretation. When the ideal outperforms the predictor on the reveal’s scale, $S(y-x)>0$ , you increase $λ$ ; otherwise you decrease it.

3) Composing many ideas into a single bearing

Let ideas provide bearings $θ_k$ on the unit circle with nonnegative weights $w_k$ . The circular mean (mean direction) is

θ̄ = atan2( Σ w_k sin θ_k , Σ w_k cos θ_k )

The resultant length

R = sqrt( (Σ w_k cos θ_k)^2 + (Σ w_k sin θ_k)^2 ) / Σ w_k

quantifies coherence: $R≈1$ means the ideas align; $R≈0$ means they cancel. Use $θ̄$ to steer the ideal template, and optionally scale its sharpness by $R$ .

Roles. Fairness and hierarchy often determine the direction of desire (sign $s∈{-1,+1}$ ), while significance and symmetry modulate intensity $m$ and the blend $λ$ inside the denominator.

4) Diagnostics from sequences

Vertical health. Track the running mean of $S$ and its volatility $v_t$ . Many small negatives accumulate like one heavy day because logs sum.
Horizontal traps. A persistent nonzero mean of $K$ over windows indicates chronic clinging. The energy budget $Σ= (1/2)K^2$ estimates how much action capacity is diverted from design.
Denominator drift. Plot $λ_t$ over time; jumps often correspond to regime changes (e.g., a new policy, a new road, or a new supplier).

5) Edge conditions and hygiene

Units and positivity. The ratio requires commensurate units and $A>0,E>0$ . If needed, add a tiny $ε$ to avoid zeros before logs.
Clipping. Keep $λ∈[0,1]$ , and cap extreme $|K|$ in displays to prevent single outliers from hijacking attention.
Context keys. Maintain separate EMAs when contexts truly differ (e.g., weekday vs weekend; morning vs evening).

6) Mapping emotions to axes (operational)

Vertical (OO–AAH, Witness): awe, delight, relief (S>0); sadness, disappointment, poignancy (S<0).
Horizontal (clinging, Participant):Grasping before — worry, anxiety, dread, craving, urgency, over-planning (D present). Resisting after — irritation, anger, regret, rumination, shame (K≠0).

Quick test: If your language starts with “What if…”, you’re pre-reveal (tilt D). If it starts with “If only… / should have…”, you’re post-reveal (nonzero K).

7) From sting to design (the flip)

Keep desire as a design parameter:

But after the reveal, set

K → 0 (accept A), update E with A, re-aim (s,m) at the next controllable step

The motive is not suppressed; it is redeployed. In practice this means converting “It should have been 18 minutes” into “Next time I depart 10 minutes earlier or choose the toll route.”

8) Pocket algorithms

Predictor (log-EMA): $p_t = (1-α)p_{t-1} + α ln A_{t-1}$ , $P_t=exp(p_t)$ .
Expected: $ln E = (1-λ) ln P + λ ln I$ .
Surprise: $S=ln(A/E)$ (vertical).
Desire: $D=E e^{s m}$ .
Suffering: $K=ln(D/A)$ (horizontal), energy $(1/2)K^2$ .
Learn λ (one step): $λ ← clip( λ + 2η S (ln I − ln P) , 0, 1 )$ .
Circular mean for ideas: $θ̄ = atan2(Σ w_k sin θ_k, Σ w_k cos θ_k)$ , coherence $R$ as above.

9) The three theorems, operational form

Arrow. Write-protect $A_t$ ; update only future $E$ . Then logs of ratios add.
No-Fear (Witness). Keep $K=0$ ; let $S$ be large or small without horizontal load.
Tyrant. Any insistence that $A=I$ manifests as $K≠0$ whenever reality disagrees.

10) What to carry forward

Time is a sequence of pairs; the present is the single ratio you are resolving now.
The denominator learns by blending evidence (P) with template (I), and the blend-weight $λ$ is itself learnable.
Ideas are servants in the denominator and become tyrants only as numerators.
Desire is valuable motive before the reveal and valuable for design after; only clinging makes it sting.
Two axes organize experience: vertical “OO–AAH” $S$ and horizontal load $K$ —and you can keep them orthogonal.

Next chapter: making the update rules tangible on a chalkboard—worked numerics for the Price and Navigation demos, plus a short appendix on switching models when regimes change.

Worked Demonstrations & Chalkboard Numerics

These pages make the framework tactile: step–by–step numbers for the Price-is-Right and Navigation demos, a quick equation card, and practical diagnostics. Equations appear as white SVGs for dark mode.

0) Quick Card (copy to the corner of the board)

Reality (ratio): $A/E$ Surprise: $S=ln(A/E)$
Predictor (log-EMA): $p_t=(1-α)p_{t-1}+α ln A_{t-1}$ , $P_t=exp(p_t)$
Ideal blend: $ln E=(1-λ)ln P + λ ln I$ \, ( $0≤λ≤1$ )
Desire (pre-actual tilt): $D=E e^{s m}$ , $s∈{-1,0,+1}$ , $m≥0$
Suffering (post-reveal clinging): $K=ln(D/A)$ , energy $(1/2)K^2$

1) Demonstration A — Price-is-Right (seller’s table)

You’ve examined comparable lamps (tariffs, seasonality). Your predictor and ideal suggest an expected price. We then reveal the actual tag. Finally, we quantify any sideways load from pre-set desire.

Setup

Predictor (evidence): P = \$240
Ideal (perfect market): I = \$200
Blend weight: $λ=0.25$

Blend in log-space:

Numerically this yields E ≈ \$230.

Reveal & vertical read

Actual tag: A = \$300

Witness’ surprise:

Desire & sideways load

Suppose a fairness-tilt (want lower) with intensity $m=0.3$ :

If the wish persists after the reveal:

K = ln(170.4/300) ≈ -0.565 ; energy ≈ 0.16

Flip to design. Keep the motive (affordable goods) but drop $K→0$ : negotiate quantity, change supplier, redesign materials—aim the energy at the next E, not the already-revealed A.

2) Demonstration B — Navigation (Friday 4:15pm)

From Lockwood to Coleman. The predictor knows bridge patterns and school pickup; the ideal knows free-flow. We compute the expected time, see the reveal, and diagnose any clinging.

Setup

Predictor (Friday data): P = 26 min
Ideal (free-flow): I = 18 min
Blend weight: $λ=0.4$

Blend:

ln E = 0.6 ln 26 + 0.4 ln 18 ⇒ E ≈ 22.4 min

Reveal & vertical read

Actual: A = 32 min

Desire & sideways load

Suppose a strong “be earlier” motive (fairness-tilt): $s=-1$ , $m=0.5$ .

K = ln(13.6/32) ≈ -0.856 ; energy ≈ 0.37

Flip to design. Accept A, reroute or depart earlier next time; convert the same motive into arrival-time control rather than rumination.

3) Learning the blend weight $λ$ (on the fly)

We adjust how much the ideal template steers the denominator by minimizing squared surprise $L=S^2$ . Let $a=ln A, x=ln P, y=ln I$ :

If the ideal consistently predicts better than the evidence ( $S(y-x)>0$ ), increase $λ$ ; otherwise decrease. Clip to $[0,1]$ .

4) Composing many ideas into one bearing

If several ideas (fairness, hierarchy, significance, symmetry) are active with weights $w_k$ and bearings $θ_k$ , take the circular mean:

Coherence (sharpness) is the resultant length:

Use $θ̄$ to orient the ideal template and optionally modulate its confidence by $R$ .

5) Diagnostics you can read at a glance

Vertical health: run a windowed mean of $S$ and dispersion (EMA of squared innovations) $v_t$ .
Horizontal traps: persistent nonzero mean of $K$ → chronic clinging; budget diverted is $Σ=(1/2)K^2$ .
Denominator drift: plot $λ_t$ for regime changes (policy shifts, new roads, new suppliers).

6) The six clear sentences (restated)

Everyone has desires: $D=E e^{s m}$ with $m≥0$ is normal.
Suffering appears when you cling after the reveal: $K≠0$ .
Clinging = consulting $D$ instead of $A$ .
The amount is quantifiable: $|K|$ or energy $(1/2)K^2$ .
Desire is unrealistic as a numerator; reality uses $A/E$ .
The root cause is an idea trying to actualize; as a servant it improves $I→E$ , as a master it demands $A=I$ .

7) One more curtain call (fast mental arithmetic)

A mental trick for class: if you can spot that $A/E≈1.2$ , then $S≈ln 1.2≈0.182$ . If $A/E≈4/3$ , then $S≈ln(4/3)≈0.288$ . Useful rounders help students feel the vertical axis without a calculator.

You now have numerical scaffolding that fits on a chalkboard, runs in your head, and lands in the body. Next chapter: regime switches and model selection—how to know when yesterday’s predictor should be retired.

Regime Switches, Model Selection, and Robustness

Predictors age, roads change, markets pivot. This chapter shows how to detect regime shifts from the surprise stream, switch or blend models, and keep ideas in their servant role while the denominator adapts.

1) What is a regime change?

A regime is a span where the mapping from context to outcomes is stable enough that your one-step predictor $P$ can be calibrated. A change is a persistent shift in the distribution of innovations (surprises). Let $S_t=ln(A_t/E_t)$ . In a healthy regime, $S_t$ has mean near $0$ and bounded dispersion. A drift in mean or variance signals that the denominator’s map is stale.

2) Three chalkboard detectors

2.1 Z-threshold on normalized surprise

Maintain an EMA of squared innovations to track volatility:

Define a normalized score

Alarm if $|z_t| > k$ for several consecutive points (e.g., $k=3$ ).

2.2 Page–Hinkley (cumulative mean shift)

Track cumulative deviation from a reference mean:

m_t = min(m_{t-1}, M_t) , PH_t = M_t - m_t

Raise change-point if $PH_t > h$ . Choose a small $δ$ to ignore tiny drifts.

2.3 EWMA control chart

Smooth the surprises:

Signal if $|Y_t| > L σ_Y$ where $σ_Y$ is the in-regime standard deviation of $Y_t$ .

Board heuristics. Use a fast volatility EMA ( $β≈0.2$ ) for shocks; a slower mean monitor ( $γ≈0.05$ ) for drifts.

3) What to do when a shift hits

3.1 Reset, fork, or blend

Reset: reinitialize the predictor $P$ when the old past is now harmful.
Fork: keep the old predictor for its context, start a new one for the new regime.
Blend models: hold a small set ${E^{(i)}}$ and weight them by their likelihood on recent reveals.

3.2 Bayesian model averaging (one-line weight update)

With model weights $w_i$ (sum to 1), update after seeing $A_t$ :

w_i ← normalize( w_i · L_i(A_t | E_t^{(i)}) )

Practical choice. Use a lognormal likelihood around each $E_t^{(i)}$ : $S^{(i)} = ln(A/E^{(i)})$ and $L_i ∝ exp( - (S^{(i)})^2 / (2σ^2) )$ .

3.3 Adaptive ideal weight $λ$

Keep the earlier gradient step but slow it across shocks. Let $η_t = η_0 / (1 + c t)$ and suspend updates while an alarm is active. Then resume with a short warm-up window.

4) Two demos under regime stress

4.1 Price-is-Right after a tariff change

Suppose a sudden tariff raises costs. You’ll see a streak of positive surprises: $S_t>0$ and an uptick in $v_t$ . The z-threshold trips; fork a “post-tariff” predictor and let model weights $w_i$ shift toward it. Keep $λ$ modest while the new regime stabilizes.

4.2 Navigation after a new bridge schedule

A revised draw schedule yields alternating sags and spikes. The EWMA chart moves off center. Reset the Friday-4:15 context predictor, seed it with the last few weeks, and reduce $η$ for the $λ$ -update until volatility subsides.

5) Hygiene: keep ideas as servants during shifts

Do not let an idea rewrite the reveal ( $A$ stays scalar and factual).
Do let ideas inform the template and the portfolio of models: they steer $I$ and which contexts you consider—but never the numerator.
Witness discipline: even through shocks, $S=ln(A/E)$ is the only vertical; drop horizontal load $K$ to zero before redesigning.

6) Parameter cheat sheet (chalkboard values)

Predictor smoothing: $α∈[0.05,0.3]$ (slow→stable, fast→responsive).
Volatility EMA: $β≈0.1–0.3$ .
EWMA mean: $γ≈0.05$ , limits $L=3$ .
PH test: small drift $δ≈0.02$ , threshold $h≈0.5–1.0$ .
λ learning rate: $η_0≈0.05$ , decay $c≈0.01$ .
Desire scale: small $m$ for everyday preferences ( $0.1–0.4$ ); larger for ambitious redesigns.

7) Micro-lab (ten points, felt on the board)

Draw two rows: top row the actuals $A_t$ , bottom row the expecteds $E_t$ . Compute $S_t=ln(A_t/E_t)$ under your breath as you go. Watch the EWMA creep, call the regime switch, fork a predictor, then continue. End by asking: is any $K$ still > 0? If so, flip it to design.

8) FAQ (for students in the room)

Q. Do I ever set desire to zero?
A. Not required. Keep the motive as design; set only the sting $K$ to zero after the reveal.

Q. Can ideas disagree and still help?
A. Yes. Use the circular mean bearing $θ̄$ and resultant length $R$ to summarize; low $R$ means “don’t trust the template much” (smaller $λ$ ).

Q. What about rare black-swan surprises?
A. Cap their influence on learning (clip $S$ and freeze $λ$ for a few steps), but let them count for the Witness (feel the vertical fully).

With detectors, switching, and blending in place, your denominator stays honest, your Witness stays vertical, and your designs stay generous. Next: tying these mechanics back to classroom practice—rubrics for grading “math-as-argument” submissions so philosophers and theologians can be precise without losing poetry.

Grading Rubrics & Math-as-Argument Practice

This chapter turns the framework into classroom craft: how to write a concise, defensible mathematical argument about experience, how to grade it, and two fully worked mini-essays you can model.

1) Why “math-as-argument” for philosophy & theology?

Clarity: Ratios force commensurable units. Logs make growth additive. Assumptions can’t hide.
Humility: The numerator stays factual (A is scalar). Ideas serve in the denominator.
Compassion: We separate vertical feeling $S=ln(A/E)$ from horizontal load $K=ln(D/A)$ , keeping blame out of the math.

2) The rubric (100 points)

Criterion	What we look for	Pts
Statement	Clear claim in one or two sentences; names the phenomenon and context.	10
Model	Correct use of core equations: $ln E=(1-λ)ln P + λ ln I$ , $S=ln(A/E)$ , $D=E e^{s m}$ , $K=ln(D/A)$ . States units.	20
Data & Assumptions	Explicit values for A, P, I, and $λ$ ; justifies choices (context, seasonality).	15
Vertical analysis	Computes $S$ and interprets sign/magnitude (“OO–AAH”).	15
Horizontal analysis	Specifies $s,m$ , computes $K$ , distinguishes pre- vs post-reveal.	15
Flip to design	Turns motive into next-step denominator changes (route, timing, supplier…). No rumination.	15
Style & integrity	Concise, units consistent, no hand-waving, cites context keys (e.g., “Fri 4:15”).	10

3) Submission template (≤ 250 words)

Claim. One sentence describing the phenomenon and context.

Setup.P=…, I=…, $λ$ =… (justify briefly).
$ln E=(1-λ)ln P + λ ln I$ ⇒ E=… (units).

Reveal.A=… (units). Surprise $S=ln(A/E)$ =… → read the sign and size.

Desire (pre-actual). Choose $s∈{-1,0,+1}$ , $m≥0$ ; $D=E e^{s m}$ =… . If clinging, $K=ln(D/A)$ =…

Flip. Convert motive to denominator edits for next trial (what changes in P, which ideas rotate I, and why).

4) Mini-essay example — Price

Claim. My surprise at a lamp price was positive but useful; clinging would have wasted margin.

Setup. Comparable data suggests P=\$240; perfect-market template I=\$200; I choose $λ=0.25$ (ideas inform but don’t dominate).

ln E = 0.75 ln 240 + 0.25 ln 200 ⇒ E ≈ 230

Reveal. A=\$300. $S=ln(300/230)≈0.266$ → pleasant surprise (“AAH”).

Desire. Pre-reveal I preferred lower prices (fairness tilt), $s=-1$ , $m=0.3$ . $D=E e^{-0.3}≈170.4$ . If I clung: $K=ln(170.4/300)≈-0.565$ (energy ≈0.16).

Flip to design. Accept A; retain motive by negotiating volume or redesigning materials; update seasonal keys in P; leave $λ$ modest until volatility normalizes.

5) Mini-essay example — Navigation

Claim. My Friday commute ran long; the sting dissolved when I redeployed motive to scheduling.

Setup. Context EMA gives P=26 min (Fri 4:15); ideal free-flow I=18; $λ=0.4$ (the template has value).

Reveal. A=32. $S=ln(32/22.4)≈0.355$ .

Desire. Beforehand, I wanted “earlier” (fairness toward others’ time): $s=-1,m=0.5$ . $D=E e^{-0.5}≈13.6$ . Clinging yields $K≈-0.856$ (energy ≈0.37).

Flip to design. Accept A; shift motive into departure-time policy (+10 min), and watch the predictor’s EMA for the new bridge schedule.

6) Common pitfalls & precise fixes

Unit mismatch. Fix: ensure numerator/denominator share units before ratio.
Hiding behind ideas. Fix: state A first; ideas only steer E.
Desire after reveal. Fix: compute $K$ , then set it to 0 and describe the design action.
Over-confident ideal. Fix: justify $λ$ with recent performance; learn it with $λ ← λ + 2η S (ln I − ln P)$ .
Ignoring context. Fix: declare keys (weekday, hour, season) or state “no known seasonality.”

7) Glossary (symbols you actually use)

8) Instructor notes (fast grading workflow)

Scan for units and context keys first; if missing, deduct early.
Circle the four numbers: A, P, I, $λ$ . If any is implied but not stated, ask for a revision.
If $K$ is computed but no design flip is proposed, return with a one-line prompt: “Redeploy desire into the next denominator.”
Reward succinctness: good work fits in 150–250 words with one or two clear equations.

With a shared rubric and compact template, students can argue precisely with numbers without losing the thread of meaning. Next: a short appendix of “board macros” (ready-to-copy snippets) and a one-page checklist for students to tape inside their notebooks.

Appendix — Board Macros & One-Page Checklist

Paste-ready snippets for the chalkboard and a single-page checklist students can tape inside their notebooks. Keep the vertical (surprise) clean, keep the horizontal (clinging) honest, let ideas serve, then flip them into design.

A) Board Macros (copy these lines as-is)

Core ratios
Reality: $A/E$ Surprise: $S=ln(A/E)$
Desire: $D=E e^{s m}$ Clinging: $K=ln(D/A)$
Energy: $(1/2)K^2$

Denominator build
Predictor (log-EMA): $p_t=(1-α)p_{t-1}+α ln A_{t-1}$ , $P_t=exp(p_t)$
Ideal blend: $ln E=(1-λ)ln P + λ ln I$

Many ideas → one bearing
$θ̄ = atan2(Σ w_k sinθ_k, Σ w_k cosθ_k)$
Coherence: $R = √((Σ w_k cosθ_k)^2 + (Σ w_k sinθ_k)^2) / Σ w_k$

Regime detectors (quick)
Volatility EMA: $v_t=(1-β)v_{t-1}+β S_t^2$ → $z_t=S_t/√(v_t+ε)$
EWMA mean: $Y_t=(1-γ)Y_{t-1}+γS_t$ (limits $±Lσ_Y$ )

Learning the ideal weight
Let $a=lnA, x=lnP, y=lnI$ ,
$S=(a-x)-λ(y-x)$ ,
$λ ← clip(λ + 2η S (y-x), 0, 1)$

OO–AAH / Left–Right
Vertical feeling: $S=ln(A/E)$
Horizontal load (post-reveal): $K=ln(D/A)$ .
Design flip: set $K→0$ , move motive to next $E$ .

B) Mental Math Aids (no calculator)

Ratio $A/E$	Log approx $S≈ln(A/E)$	Feel
1.10	≈ 0.095	small “AAH”
1.20 ≈ 6/5	≈ 0.182	clear “AAH”
1.33 ≈ 4/3	≈ 0.288	strong “AAH”
0.90	≈ −0.105	small “OO”
0.80 = 4/5	≈ −0.223	clear “OO”
0.67 ≈ 2/3	≈ −0.405	strong “OO”

Use ln(1+x)≈x for small x; e.g., 1.2 → x=0.2 → ~0.2 but subtract a bit (0.018) → 0.182.

C) One-Page Student Checklist

1. Context first. Name units, timing, and keys (e.g., “Fri 4:15, school pickup”).

2. State the four numbers.A (factual), P (from a log-EMA), I (template), $λ$ (weight).

3. Build the denominator. $lnE=(1-λ)lnP+λlnI$ ⇒ E.

4. Feel the vertical. $S=ln(A/E)$ → label OO (−) or AAH (+) and its strength.

5. Name the motive (pre-actual). Choose direction $s∈{-1,0,+1}$ (fairness lower, hierarchy higher, neutral), intensity $m$ .

6. Compute desire. $D=E e^{s m}$ .

7. After the reveal, check clinging. $K=ln(D/A)$ and optional energy $(1/2)K^2$ .

8. Flip to design. Set $K→0$ in words: change route/time/supplier/assumptions for the next denominator.

9. Learn gently. Update $λ$ with the gradient rule; pause during shocks; note regime changes.

10. Keep ideas as servants. They shape I and model portfolios, never the numerator.

D) Two Rapid Demos (fill-in blanks on the board)

Price
P = $ ____ , I = $ ____ , $λ$ = ____ ⇒ $lnE$ ⇒ E ≈ $ ____
Reveal A = $ ____ → $S=ln(A/E)$ = ____ (OO/AAH?)
Desire: s = −1/0/+1, m = ____ → D = ____ ; K = ln(D/A) = ____ → flip plan: _______

Navigation
P = ____ min, I = ____ min, $λ$ = ____ ⇒ E ≈ ____ min
Reveal A = ____ min → $S$ = ____ (OO/AAH?)
Desire: s = −1/0/+1, m = ____ → D = ____ ; K = ln(D/A) = ____ → flip plan: _______

E) Common Tripwires (and fixes)

Using different units in a ratio. Fix: convert first (Celsius/Celsius, dollars/dollars, minutes/minutes).
Estimating with desire instead of expected. Fix: build E from P and I; compute D separately.
Arguing with the reveal. Fix: compute $K$ , then write one sentence that zeroes it and reallocates motive.
Over-trusting the ideal. Fix: justify $λ$ with recent performance; let it learn slowly.
Ignoring regime change. Fix: watch $z_t, Y_t$ ; fork predictors when alarms persist.

F) Tiny Reference — Symbols

These macros and the checklist let you move quickly on the board without sacrificing rigor. Next appendix: a compact “design flip catalog” (how common motives translate into denominator edits across domains).

$A$	Actual (revealed scalar, factual).
$P$	Predictor (log-EMA of past actuals; context-aware).
$I$	Ideal template (ideas’ bearing; perfect-conditions baseline).
$λ$	Blend weight of ideal into expected (0–1).
$E$	Expected outcome: $ln E=(1-λ)ln P + λ ln I$
$S$	Surprise (vertical): $S=ln(A/E)$ .
$D$	Desire (pre-actual): $D=E e^{s m}$ ; $s∈{-1,0,+1}$ (direction), $m≥0$ (intensity).
$K$	Clinging after reveal: $K=ln(D/A)$ ;
$θ̄,R$	Circular mean and coherence of multiple ideas.

A	Actual (revealed scalar, factual).
P	Predictor (evidence, log-EMA of past actuals in context).
I	Ideal (ideas’ perfect-conditions template, oriented by bearings).
$λ$	Blend weight of ideal into expected.
E	Expected outcome, geometric blend of P and I.
$S$	Surprise (vertical): $ln(A/E)$ .
D	Desire (pre-actual): $E e^{s m}$ .
$K$	Clinging after reveal: $ln(D/A)$ .
$θ̄,R$	Bearing of ideas and their coherence.

Sequences, Learning Denominators, and the Two Axes of Experience

1) Sequences, and why time points forward

2) How the denominator learns: Predictor ⊕ Ideal

Horizon forecasts without breaking the one-step rule

3) Ideas in their right place: circle bearings, cooperative servants

4) Desire as design: keep the energy, lose the sting

After the reveal: sting vs. design

5) The Witness/Participant protocol (R → U → A)

6) Two curtains, revisited

Price-is-Right

Navigation

7) Ignorance, precisely

8) Three compact theorems

9) Pocket summary

Updating the Denominator, Learning Influence, and Composing Ideas

1) Making the predictor P concrete

1.1 Exponential moving average (EMA)

1.2 A volatility-aware EMA (“Kalman-lite”)

1.3 Seasonality and regime changes

2) Blending with the ideal I and learning

3) Composing many ideas into a single bearing

4) Diagnostics from sequences

5) Edge conditions and hygiene

6) Mapping emotions to axes (operational)

7) From sting to design (the flip)

8) Pocket algorithms

9) The three theorems, operational form

10) What to carry forward

Worked Demonstrations & Chalkboard Numerics

0) Quick Card (copy to the corner of the board)

1) Demonstration A — Price-is-Right (seller’s table)

Setup

Reveal & vertical read

Desire & sideways load

2) Demonstration B — Navigation (Friday 4:15pm)

Setup

Reveal & vertical read

Desire & sideways load

3) Learning the blend weight (on the fly)

4) Composing many ideas into one bearing

5) Diagnostics you can read at a glance

6) The six clear sentences (restated)

7) One more curtain call (fast mental arithmetic)

Regime Switches, Model Selection, and Robustness

1) What is a regime change?

2) Three chalkboard detectors

2.1 Z-threshold on normalized surprise

2.2 Page–Hinkley (cumulative mean shift)

2.3 EWMA control chart

3) What to do when a shift hits

3.1 Reset, fork, or blend

3.2 Bayesian model averaging (one-line weight update)

3.3 Adaptive ideal weight

4) Two demos under regime stress

4.1 Price-is-Right after a tariff change

4.2 Navigation after a new bridge schedule

5) Hygiene: keep ideas as servants during shifts

6) Parameter cheat sheet (chalkboard values)

7) Micro-lab (ten points, felt on the board)

8) FAQ (for students in the room)

Grading Rubrics & Math-as-Argument Practice

1) Why “math-as-argument” for philosophy & theology?

2) The rubric (100 points)

3) Submission template (≤ 250 words)

4) Mini-essay example — Price

5) Mini-essay example — Navigation

6) Common pitfalls & precise fixes

7) Glossary (symbols you actually use)

8) Instructor notes (fast grading workflow)

Appendix — Board Macros & One-Page Checklist

A) Board Macros (copy these lines as-is)

B) Mental Math Aids (no calculator)

C) One-Page Student Checklist

D) Two Rapid Demos (fill-in blanks on the board)

E) Common Tripwires (and fixes)

F) Tiny Reference — Symbols

Share this:

Like this:

Leave a ReplyCancel reply

Discover more from John Rector

2) Blending with the ideal I and learning $λ$

3) Learning the blend weight $λ$ (on the fly)

3.3 Adaptive ideal weight $λ$