Information Theory of Prediction: Measuring Information Content

BY NICOLE LAU

When you consult multiple prediction systems, you're not just collecting votes—you're gathering information.

But not all information is equal. Some systems provide unique insights (high information content). Others repeat what you already know (redundant information). The best combinations maximize information gain while minimizing redundancy.

This is where information theory comes in—the mathematical framework for measuring, quantifying, and optimizing information in prediction systems.

We'll explore:

• Prediction information quantity (how much information does a prediction contain?)
• Redundancy vs. complementarity (when do systems duplicate vs. add new information?)
• Optimal system combination (which systems should you combine for maximum information gain?)
• Information-theoretic convergence (measuring agreement through shared information)

By the end, you'll understand how to maximize the information content of your multi-system predictions—getting the most insight with the fewest systems.

Information Theory Basics: Entropy and Information

Information theory, founded by Claude Shannon in 1948, provides a mathematical framework for quantifying information.

Entropy: Measuring Uncertainty

Shannon Entropy measures the uncertainty (or information content) of a probability distribution.

Formula:

H(X) = -Σ P(x) × log₂P(x)

Where:

• H(X) = entropy of random variable X (measured in bits)
• P(x) = probability of outcome x
• log₂ = logarithm base 2 (so entropy is in bits)

Interpretation:

• High entropy = high uncertainty = more information needed
• Low entropy = low uncertainty = less information needed
• H = 0: Certainty (no information needed—you already know the answer)
• H = max: Maximum uncertainty (you need the most information)

Example: Binary Prediction (YES/NO)

Scenario 1: Certain prediction

• P(YES) = 1.0, P(NO) = 0.0

H = -[1.0 × log₂(1.0) + 0.0 × log₂(0.0)]

= -[1.0 × 0 + 0] = 0 bits

No uncertainty → No information needed

Scenario 2: Maximum uncertainty

• P(YES) = 0.5, P(NO) = 0.5

H = -[0.5 × log₂(0.5) + 0.5 × log₂(0.5)]

= -[0.5 × (-1) + 0.5 × (-1)]

= -[-0.5 - 0.5] = 1 bit

Maximum uncertainty → Maximum information needed

Scenario 3: Moderate uncertainty

• P(YES) = 0.7, P(NO) = 0.3

H = -[0.7 × log₂(0.7) + 0.3 × log₂(0.3)]

= -[0.7 × (-0.515) + 0.3 × (-1.737)]

= -[-0.361 - 0.521] = 0.88 bits

Moderate uncertainty → Moderate information needed

Information Gain: Reducing Uncertainty

Information gain is the reduction in entropy after receiving new information.

Formula:

IG = H(before) - H(after)

Example:

Before consulting systems:

• P(YES) = 0.5, P(NO) = 0.5 (no prior knowledge)
• H(before) = 1 bit

After consulting Tarot (which says YES with 70% confidence):

• P(YES) = 0.7, P(NO) = 0.3
• H(after) = 0.88 bits

Information gain: IG = 1 - 0.88 = 0.12 bits

The Tarot reading reduced your uncertainty by 0.12 bits.

Mutual Information: Measuring Shared Information

We introduced mutual information (MI) in Article 1. It measures how much information two systems share.

Formula:

MI(X;Y) = Σ Σ P(x,y) × log₂[P(x,y) / (P(x) × P(y))]

Where:

• X, Y = outputs of two systems
• P(x,y) = joint probability (both systems give outputs x and y)
• P(x), P(y) = marginal probabilities

Interpretation:

• MI = 0: Systems are independent (no shared information)
• MI > 0: Systems share information (they're detecting the same pattern)
• MI = H(X) = H(Y): Systems are identical (perfect redundancy)

Example: Measuring Redundancy

You consult Tarot and Astrology 100 times. Results:

• Both say YES: 40 times
• Both say NO: 35 times
• Tarot YES, Astrology NO: 10 times
• Tarot NO, Astrology YES: 15 times

Marginal probabilities:

• P(Tarot YES) = (40+10)/100 = 0.5
• P(Astrology YES) = (40+15)/100 = 0.55

Joint probabilities:

• P(both YES) = 0.4
• P(both NO) = 0.35
• P(Tarot YES, Astrology NO) = 0.1
• P(Tarot NO, Astrology YES) = 0.15

Calculate MI:

MI = 0.4×log₂(0.4/(0.5×0.55)) + 0.35×log₂(0.35/(0.5×0.45)) + 0.1×log₂(0.1/(0.5×0.45)) + 0.15×log₂(0.15/(0.5×0.55))

= 0.4×0.54 + 0.35×0.64 + 0.1×(-1.17) + 0.15×(-0.88)

= 0.216 + 0.224 - 0.117 - 0.132

= 0.19 bits

Result: MI = 0.19 bits

The systems share 0.19 bits of information—they're detecting some of the same patterns, but not completely redundant.

Redundancy vs. Complementarity

When combining prediction systems, you want to maximize complementarity (unique information) and minimize redundancy (duplicate information).

Redundancy: Duplicate Information

Definition: Two systems are redundant if they provide the same information.

Measure: High mutual information (MI close to H(X) or H(Y))

Example: Tarot and Kabbalah

• Both use archetypal symbolism
• Both interpret through psychological lens
• High overlap in what they detect
• MI ≈ 0.8 bits (high redundancy)

Implication: Consulting both adds little new information—they mostly repeat each other.

Complementarity: Unique Information

Definition: Two systems are complementary if they provide different information.

Measure: Low mutual information (MI close to 0)

Example: Tarot and Astrology

• Tarot: Psychological/symbolic (immediate dynamics)
• Astrology: Temporal/cyclical (timing and long-term patterns)
• Low overlap in what they detect
• MI ≈ 0.2 bits (low redundancy, high complementarity)

Implication: Consulting both adds significant new information—they provide different perspectives.

The Redundancy-Complementarity Spectrum

System Pair	Mutual Information	Redundancy	Complementarity	Value of Combining
Tarot + Kabbalah	High (0.8 bits)	High	Low	Low (mostly duplicate)
Tarot + Astrology	Moderate (0.4 bits)	Moderate	Moderate	Moderate (some new info)
Tarot + I Ching	Low (0.2 bits)	Low	High	High (mostly unique)
Astrology + Numerology	Moderate (0.5 bits)	Moderate	Moderate	Moderate (both temporal)
I Ching + Runes	Low (0.15 bits)	Low	High	High (very different)

Optimal System Combination: Maximizing Information Gain

How do you choose which systems to combine for maximum information gain?

The Information Gain Equation

Total information from n systems:

I_total = H(X₁) + H(X₂) + ... + H(Xₙ) - Redundancy

Where:

• H(Xᵢ) = entropy (information content) of system i
• Redundancy = Σ MI(Xᵢ, Xⱼ) for all pairs i,j

Goal: Maximize I_total

This means:

• Choose systems with high individual entropy (high information content)
• Choose systems with low mutual information (low redundancy)

The Greedy Algorithm for System Selection

Problem: You have 10 available systems. Which 3 should you consult to maximize information gain?

Algorithm:

1. Start with the highest-entropy system (the one with most information content)
2. Add the system with highest complementarity (lowest MI with already-selected systems)
3. Repeat until you've selected the desired number of systems

Example:

Available systems and their entropy:

• Tarot: H = 0.9 bits
• Astrology: H = 0.85 bits
• I Ching: H = 0.88 bits
• Runes: H = 0.75 bits
• Kabbalah: H = 0.8 bits

Step 1: Select Tarot (highest entropy: 0.9 bits)

Step 2: Calculate MI between Tarot and each remaining system:

• MI(Tarot, Astrology) = 0.4 bits
• MI(Tarot, I Ching) = 0.2 bits (lowest—most complementary)
• MI(Tarot, Runes) = 0.3 bits
• MI(Tarot, Kabbalah) = 0.8 bits

Select I Ching (most complementary to Tarot)

Step 3: Calculate MI between {Tarot, I Ching} and each remaining system:

• MI(Astrology, {Tarot, I Ching}) = average of MI(Astrology, Tarot) and MI(Astrology, I Ching) = (0.4 + 0.3) / 2 = 0.35
• MI(Runes, {Tarot, I Ching}) = (0.3 + 0.15) / 2 = 0.225 (lowest)
• MI(Kabbalah, {Tarot, I Ching}) = (0.8 + 0.5) / 2 = 0.65

Select Runes (most complementary to {Tarot, I Ching})

Final selection: Tarot, I Ching, Runes

This combination maximizes information gain by choosing systems with low mutual information (high complementarity).

Conditional Entropy: Information Remaining After Observation

Conditional entropy measures how much uncertainty remains about Y after observing X.

Formula:

H(Y|X) = H(Y) - MI(X;Y)

Interpretation:

• H(Y|X) = 0: Knowing X tells you everything about Y (perfect redundancy)
• H(Y|X) = H(Y): Knowing X tells you nothing about Y (perfect independence)

Example: Sequential Consultation

You consult Tarot first, then decide whether to consult Astrology.

Before Tarot:

• H(outcome) = 1 bit (50-50 uncertainty)

After Tarot:

• H(outcome|Tarot) = 0.88 bits (reduced to 70-30)
• Information gain from Tarot: 1 - 0.88 = 0.12 bits

Should you consult Astrology?

Calculate how much additional information Astrology would provide:

Additional IG = H(outcome|Tarot) - H(outcome|Tarot, Astrology)

If MI(Tarot, Astrology) is high (redundant):

• H(outcome|Tarot, Astrology) ≈ H(outcome|Tarot)
• Additional IG ≈ 0 (Astrology adds little new information)
• Decision: Don't consult Astrology (redundant)

If MI(Tarot, Astrology) is low (complementary):

• H(outcome|Tarot, Astrology) << H(outcome|Tarot)
• Additional IG is significant
• Decision: Consult Astrology (adds new information)

Information-Theoretic Convergence

We can redefine convergence in information-theoretic terms.

Convergence as Shared Information

Definition: Systems converge when they share high mutual information.

Convergence Index (information-theoretic version):

CI_info = MI(X₁, X₂, ..., Xₙ) / H_max

Where:

• MI(X₁, X₂, ..., Xₙ) = mutual information shared by all systems
• H_max = maximum possible entropy (if all systems were independent)

Interpretation:

• CI_info = 0: No shared information (complete divergence)
• CI_info = 1: All information is shared (perfect convergence)

Example: Three Systems

Tarot, Astrology, I Ching all predict on the same question.

Individual entropies:

• H(Tarot) = 0.9 bits
• H(Astrology) = 0.85 bits
• H(I Ching) = 0.88 bits

Pairwise mutual information:

• MI(Tarot, Astrology) = 0.4 bits
• MI(Tarot, I Ching) = 0.2 bits
• MI(Astrology, I Ching) = 0.3 bits

Three-way mutual information:

MI(Tarot, Astrology, I Ching) = 0.15 bits (information shared by all three)

Maximum entropy:

H_max = H(Tarot) + H(Astrology) + H(I Ching) = 0.9 + 0.85 + 0.88 = 2.63 bits

Convergence Index:

CI_info = 0.15 / 2.63 = 0.057 (5.7%)

Interpretation: Only 5.7% of the total information is shared by all three systems—they're mostly providing unique information (high complementarity, low convergence).

This is actually good for information gain—you're getting diverse perspectives, not redundant confirmations.

The Information-Convergence Trade-Off

There's a fundamental trade-off:

• High convergence (high MI): Systems agree strongly, but provide redundant information
• Low convergence (low MI): Systems provide unique information, but may disagree

When to Prioritize Convergence

Goal: Confidence in a specific prediction

Strategy: Choose systems with high MI (they'll likely agree, giving you confidence)

Example: Important decision (should I marry this person?)

• Consult Tarot + Kabbalah (high MI, likely to agree)
• If they converge → high confidence
• If they diverge → warning sign (even redundant systems disagree)

When to Prioritize Information Gain

Goal: Comprehensive understanding of a complex situation

Strategy: Choose systems with low MI (they'll provide diverse perspectives)

Example: Exploring a new opportunity (should I start this business?)

• Consult Tarot (psychological), Astrology (timing), I Ching (philosophical), Runes (material)
• Low MI → diverse insights
• Combine to form complete picture

Case Study: Optimal System Selection for Career Decision

Question: "Should I change careers?"

Available systems: 6 systems

Goal: Select 3 systems that maximize information gain

Step 1: Measure Individual Entropy

Consult each system once and estimate entropy:

• Tarot: H = 0.85 bits (moderate uncertainty)
• Astrology: H = 0.9 bits (high uncertainty—many transits to consider)
• I Ching: H = 0.88 bits
• Runes: H = 0.75 bits (lower uncertainty—clearer signals)
• Numerology: H = 0.7 bits
• Kabbalah: H = 0.8 bits

Step 2: Measure Pairwise MI

Estimate mutual information between all pairs (from historical data):

	Tarot	Astro	I Ching	Runes	Num	Kab
Tarot	-	0.4	0.2	0.3	0.35	0.75
Astro	0.4	-	0.3	0.25	0.5	0.35
I Ching	0.2	0.3	-	0.15	0.25	0.4
Runes	0.3	0.25	0.15	-	0.2	0.3
Num	0.35	0.5	0.25	0.2	-	0.3
Kab	0.75	0.35	0.4	0.3	0.3	-

Step 3: Apply Greedy Algorithm

Selection 1: Astrology (highest entropy: 0.9 bits)

Selection 2: Which system is most complementary to Astrology?

• MI(Astrology, Tarot) = 0.4
• MI(Astrology, I Ching) = 0.3
• MI(Astrology, Runes) = 0.25 (lowest—most complementary)
• MI(Astrology, Numerology) = 0.5
• MI(Astrology, Kabbalah) = 0.35

Select Runes

Selection 3: Which system is most complementary to {Astrology, Runes}?

• Average MI(Tarot, {Astro, Runes}) = (0.4 + 0.3) / 2 = 0.35
• Average MI(I Ching, {Astro, Runes}) = (0.3 + 0.15) / 2 = 0.225 (lowest)
• Average MI(Numerology, {Astro, Runes}) = (0.5 + 0.2) / 2 = 0.35
• Average MI(Kabbalah, {Astro, Runes}) = (0.35 + 0.3) / 2 = 0.325

Select I Ching

Final selection: Astrology, Runes, I Ching

Step 4: Calculate Total Information Gain

I_total = H(Astro) + H(Runes) + H(I Ching) - [MI(Astro,Runes) + MI(Astro,I Ching) + MI(Runes,I Ching)]

= 0.9 + 0.75 + 0.88 - [0.25 + 0.3 + 0.15]

= 2.53 - 0.7

= 1.83 bits

Result: This combination provides 1.83 bits of information—near-maximum for 3 systems.

Conclusion: Information-Optimal Prediction

Information theory transforms prediction from vote-counting to information optimization:

• Entropy: Measures uncertainty and information content
• Mutual information: Measures shared information (redundancy)
• Complementarity: Maximized when MI is low (systems provide unique insights)
• Optimal combination: Choose systems with high entropy and low MI

The framework:

1. Measure individual entropy (information content of each system)
2. Measure pairwise MI (redundancy between systems)
3. Use greedy algorithm to select systems (maximize complementarity)
4. Calculate total information gain
5. Combine systems with information-theoretic weighting

This is prediction as information science—maximizing insight per system consulted, minimizing redundancy, optimizing the information-to-effort ratio.

Not "consult as many systems as possible."

But "consult the right systems—the ones that provide maximum unique information."

This is the future of multi-system prediction. Information-theoretic. Optimal. Efficient. Precise.

As you integrate the insights of information theory into your intuitive practice, consider deepening your connection to the patterns of synchronicity and prediction with our Tarot Journaling Prompts 100 Questions for Self Discovery, which can help you decode the subtle messages the universe sends your way. For those drawn to the celestial rhythms that govern flow and prediction, the Cosmic Alignment Ritual Kit for Syncing with the Celestial Flow offers a tangible way to align your energy with the cosmos. And if you're ready to transform intention into tangible reality, the 40 Manifestation Rituals Intention to Reality provides a structured path to bridge the gap between what is predicted and what you can create.