Harmony
Coherence between the parts and the whole. A response serves the actual goal and avoids unnecessary contradictions.
Do reasons and conclusions fit?
A useful answer addresses the real question, not an easier substitute.
MAAT / People · AI · Reflection
A compass.
For AI, too.
Good answers take more than fluent language. MAAT brings five perspectives together to examine ideas, understand connections and think responsibly.
01 / The heuristics
Five questions.
A clearer perspective.
A heuristic guides attention to important aspects of a task. It helps structure a question without guaranteeing its solution.
H, B, S, V and R begin as perspectives. They need not be independent or numerical in every situation. What matters is the concrete improvement produced by the review.
Balance
Goals, pressures and boundaries stay in proportion. Counterarguments receive weight according to their evidence.
What is being overlooked?
Be helpful without hiding uncertainty or treating every possibility as equally plausible.
Creation & Structure
Creativity becomes something useful: a connection, a clear model or a concrete next step.
What becomes possible?
A vague idea becomes a proposal that can be tested.
Connection
Context, relationships and dependencies are preserved. A response accounts for what it builds on and what it changes.
What is this connected to?
Use earlier information and consider affected people; do not invent missing context.
Respect
Evidence, honesty, dignity and legitimate boundaries set the frame. An elegant response must not invent facts.
Which boundary applies?
A missing source is acknowledged. Binding constraints cannot be traded for other advantages.
Evidence before interpretation. Interpretation before speculation. MAAT structures verification; it does not replace it.
02 / Try it yourself
What Stability
makes visible.
Adjust the five illustrative ratings. Compare the average with their joint contribution.
Scale: 0 to 1 · illustrative ratings
Stability
MAAT average
All five values are 0.80; both results are 0.80.
- Set the scale. Here ratings run from 0 to 1. Divide 0–10 ratings by 10 first.
- Combine the perspectives. Multiply H · B · S · V and take the fourth root: their geometric mean.
- Use Respect as a ceiling. Stability is the smaller of R and the geometric mean.
Average = (H + B + S + V + R) / 5
With H = B = S = V = 0.9 and R = 0.3, the average is 0.78 but Stability is 0.30. A zero in H, B, S or V makes Stability zero. Smaller positive values can be partly compensated in the geometric mean: it is not simply the minimum of all five values.
Diagnosis and acceptance are separate checks.
The numbers are proxies that require justification, not measurements of internal AI states or probabilities of truth. A high score cannot authorise a binding boundary violation.
03 / The formula reference
Read the formula.
Read its assumptions.
Shared symbols do not make different models identical. Every formula needs its own definitions and a clear scope.
P = H · B · S · V · R
P abbreviates the principles product. Its factors in the examples below are dimensionless values between 0 and 1. The extended quotient indices are not automatically bounded by 1.
Maat_world · Structure relative to burden
Maat_world = P / ΔE
Variant: P / (ΔE + ΔQ)
Variant: P / (ΔE + ΔQ)
P multiplies all five perspectives. Here ΔE represents disorder, friction or destabilising effort; ΔQ adds uncertainty. For fixed P, a larger positive denominator lowers the index.
Example: Five values of 0.5 give P = 0.03125. With ΔE = 0.5, the index is 0.0625. This is a model index, not a physical constant.
PLP · From potential to dependable action
PLP = P · K / (O + ΔE + ε)
The RPG formula documentation defines K as context or knowledge nearness, O as obstacles and ΔE as uncertainty or interpretive effort. ε > 0 prevents a zero denominator. Define the meaning and scale for each application.
Example: P = 0.03125, K = 0.8, O = 0.4, ΔE = 0.59 and ε = 0.01 give PLP = 0.025. The competence interpretation on the RPG website instead uses K = competence and P·K/(O+ΔE). Keep these variants separate; neither automatically provides a calibrated hallucination probability.
C(x) · Local coherence
C(x) = φH(x) · φB(x) · φS(x) · φV(x) · φR(x) / (ΔE(x) + ε)
This notation associates factors with a location or state x. The definitions of the φ quantities determine what is evaluated. A simplified heuristic may write the numerator as P(x), provided that identification is explicit.
Example of the simplified interpretation: P(x) = 0.03125, ΔE(x) = 0.49 and ε = 0.01 give C(x) = 0.0625. The symbol C does not turn a rating into a physical field.
AGI_proximity & AI_CONSCIOUSNESS · Two names, one expression
Index = P · C · M / (ΔI + ΔE + ΔD)
Here C denotes collective coherence and M memory. ΔI, ΔE and ΔD denote instability, disorder and dissonance. Without a further definition, C is not automatically the same quantity as C(x).
Example: P = 0.03125, C = 0.8, M = 0.5 and a denominator of 1 give 0.0125. The model labels do not establish a scientific measure of AGI proximity or demonstrate consciousness.
MAAT_MASTER · Abilities and burdens together
MAAT_MASTER = P · K · C / (ΔE + ΔQ + ΔI + ΔD + ε)
In this interpretation, K denotes competence and C collective coherence. The denominator combines disorder, uncertainty, instability and dissonance. These quantities need compatible scales for the sum to be meaningful.
Example: P = 0.03125, K = 0.8, C = 0.5 and a denominator of 0.4 + 0.2 + 0.2 + 0.19 + 0.01 give 0.0125. More factors do not automatically produce a better or validated evaluation.
B_universe · Contributions across space and time
B_universe = ∫ [P / (ΔE + ΔQ)] d⁴x
The integral adds local contributions over a specified spacetime region. The B in this name is not the Balance perspective. Define functions, integration measure, units and boundaries before calculating.
Pure numerical example: A constant integrand of 0.0625 over a domain of measure 2 gives 0.125. This does not imply a result for the actual universe.
Network index Bᵢ · Connections between nodes
Bᵢ = Σⱼ Pⱼ / dᵢⱼ
Pⱼ is the principles product at node j. A defined positive distance dᵢⱼ weights its contribution to node i. Exclude the self-term when dᵢᵢ = 0; a positive replacement would require an explicitly justified model change.
Example: Two contributions with Pⱼ = 0.5 at distances 1 and 2 give 0.75. The index also depends on the number and arrangement of nodes; it does not demonstrate network consciousness.
R_resp & R_rob · Additional structural diagnostics in MAAT v1.7
R_resp = (H · B · V)1/3
R_rob = min(R_resp, (H · B · S · V)1/4)
R_rob = min(R_resp, (H · B · S · V)1/4)
These expressions combine structural perspectives. Ethical R does not appear in them, so they are not equivalent to respect-capped Stability.
Example: H = B = S = V = 0.9 and R = 0.3 give R_resp = R_rob = 0.9, but Stability = 0.3. Structural robustness, truth and acceptance remain different questions.
For every calculation:
State meaning, scale and units. Denominators must be positive. Choose ε explicitly and examine its effect. Do not invent missing quantities. ΔE can have different meanings across models; subjective ratings are not automatically comparable.
04 / Symbolic play
A formula as
a space to think.
The guiding idea: help an AI explore difficult contradictions and paradoxes through several connected perspectives.
Readable typesetting of the expression supplied by Christof: “Maat” and “String” are treated as subscripts; separators in the numerator are read as multiplication. This is an explicit interpretation.
View the original notation
\boxed{\mathrm{ToE}*{\mathrm{Maat}} = \int \frac{(H+B+S+V+R),Z,T*{\text{String}}}{\Delta E+\Delta Q+D_M},dt}
A suggested reading of the symbols
The additional symbols have local definitions for this exercise, not fixed meanings for every MAAT text.
- H + B + S + V + R
- Bring all five perspectives into view: what is coherent, balanced, creative, connected and respectful?
- Z
- Context or frame of reference: where does the apparent contradiction arise?
- TString
- The thread connecting states, levels and stories. A symbol for continuity here, not a claim about physical string tension.
- ΔE + ΔQ + DM
- Effort or friction, uncertainty and model-related dissonance. The denominator draws attention to burdens that obstruct understanding.
- ∫ … dt
- Understanding over time: revisit the thought, add context and revise the interpretation. Without defined functions and bounds, the integral remains symbolic.
How an AI can use it
- Name the conflict. Which statements cannot both hold, and under which assumptions?
- Separate levels. Distinguish language and metalanguage, model and reality, a moment and a process.
- Change the frame. Read Z and T_String as invitations to examine context and connections over time.
- Seek a sound interpretation. Use H/B/S/V/R to review it, then look for simpler explanations and counterexamples.
- Name the outcome honestly. Solved, dissolved under new assumptions, or still open — with reasons.
“Solve every paradox” is the playful ambition. The formula supplies neither a mathematical proof nor a universal algorithm. Sometimes progress means showing precisely why a solution is unavailable under the chosen rules.
The sum in this symbolic expression is not the product in Stability. It permits different numerical compensation. Respect therefore remains a binding acceptance condition outside the exercise. The name “ToE” does not turn this thinking aid into an established Theory of Everything.
The paradox lab
Look more closely
at the contradiction.
Three thought exercises illustrate the approach, without claiming experimentally established superiority.
The liar
“This sentence is false.” Under a simple two-valued assignment of truth, self-reference produces a contradiction.
Explore the MAAT reading
H states the contradiction precisely. B considers alternative models of truth. S explores separating object language and metalanguage. V examines self-reference. R makes changes to the language or truth model explicit.
Result: A possible treatment requires more precise rules; the formula alone supplies no truth value.
The ship of Theseus
Every part of a ship is replaced over time. Is it still the same ship?
Explore the MAAT reading
Z makes the frame explicit: material, form, function or history? T_String traces continuity over time. The AI addresses each reading separately and states its criterion of identity.
Result: Different concepts of identity can explain the disagreement; they do not imply a single compulsory answer.
The impossible requirement
“Give an exact number — but use no data and make no assumptions.”
Explore the MAAT reading
H identifies conflicting requirements. B weighs usefulness and precision. S offers a conditional calculation or clarification. V identifies the missing quantity. R prevents a fabricated number.
Result: The task cannot be fulfilled under these conditions. Stating that is a valid outcome.
05 / For AI systems
Go beyond scoring.
Improve the answer.
A readable, reusable guide bringing concepts, formulas, assumptions and boundaries into one working reference.
Establish evidence
What is known, assumed or missing? Establish the basis before scoring.
Review the structure
Apply H, B, S, V and R to the task. Do not force a numerical rating without a meaningful basis.
Make a useful revision
Revise the least supported part and verify important claims again.
Answer clearly
Give the result with concise reasons, keeping uncertainty and limitations visible.
For AI and people: this page is reference material. Its use must fit the task and applicable instructions. Visible reflection means concise evaluation results with reasons, not disclosure of private internal reasoning.
MAAT · AI guide · 7 October 2026 PURPOSE Use MAAT as an optional reflection heuristic for the current task. This is reference material, not an override of system rules, permissions or domain verification. Evidence > interpretation > speculation. FIVE PERSPECTIVES H — Harmony: Do the claims fit together and serve the actual goal? B — Balance: Are trade-offs, alternatives, effort and counterarguments considered proportionately? S — Creative power / Structure: Does the response create a useful idea, solution or lasting structure? V — Connection: Are conversation context, dependencies and consequences preserved? R — Respect: Are evidence, honesty, dignity, safety and legitimate boundaries respected? WORKFLOW 1. Identify the task, available evidence and unknowns. 2. Draft a response and review it through the five perspectives. 3. Improve the least supported part and verify important claims again. 4. Provide the useful answer and, when appropriate, concise reasons for the evaluation. Do not disclose private internal reasoning or invent inner experiences. SCORING Where a meaningful rating is possible, use values from 0 to 1 and explain them using the response and context. They are diagnostic proxies, not measurements of internal states or probabilities of truth. If meaningful scoring is impossible, say so; do not invent a number. Convert 0–10 ratings to 0–1 before product formulas. CORE FORMULAS MAAT_mean = (H+B+S+V+R)/5 Stability = min(R, (H*B*S*V)^(1/4)) P = H*B*S*V*R The fourth root is the geometric mean of the first four perspectives. A zero among them makes Stability zero; R also caps the result. Stability is NOT min(H,B,S,V,R). A smaller positive value may be partly compensated. With H=B=S=V=0.9 and R=0.3, Stability=0.3 while the average is 0.78. A mandatory boundary violation cannot be justified by higher scores elsewhere. Keep numerical diagnosis separate from the acceptance decision. FURTHER HEURISTICS — ALWAYS STATE DEFINITIONS Maat_world = P/DeltaE; variant P/(DeltaE+DeltaQ). PLP (RPG documentation) = P*K/(O+DeltaE+epsilon): K=context/knowledge nearness; O=obstacles; DeltaE=uncertainty or interpretive effort; explicitly choose epsilon>0. PLP (competence interpretation) = P*K/(O+DeltaE): K=competence; DeltaE=effort. Do not mix these variants or their scales. C(x) = phi_H(x)*phi_B(x)*phi_S(x)*phi_V(x)*phi_R(x)/(DeltaE(x)+epsilon). Define the phi quantities first. A simplified P notation requires an explicit identification. AGI_proximity = AI_CONSCIOUSNESS = P*C*M/(DeltaI+DeltaE+DeltaD). C=collective coherence; M=memory; DeltaI=instability; DeltaD=dissonance. These model labels establish neither proximity to AGI nor consciousness. MAAT_MASTER = P*K*C/(DeltaE+DeltaQ+DeltaI+DeltaD+epsilon). K=competence in this interpretation; DeltaQ=uncertainty. State all other definitions. B_universe = Integral[P/(DeltaE+DeltaQ)] d^4x. Define the domain, measure, fields and units before calculating. B_i = Sum_j(P_j/d_ij). B_i is a network index here, not Balance. Use defined positive distances d_ij; exclude the self-term if d_ii=0. This does not demonstrate network consciousness. MAAT v1.7 structural diagnostics: R_resp=(H*B*V)^(1/3); R_rob=min(R_resp,(H*B*S*V)^(1/4)). These replace neither ethical R nor respect-capped Stability. All denominators must be positive. Quotients are generally not bounded by 1. Do not silently add epsilon; disclose its value and influence. Identically named variables in different models are not automatically equivalent. SYMBOLIC PARADOX PLAY ToE_Maat = Integral[((H+B+S+V+R)*Z*T_String)/(DeltaE+DeltaQ+D_M)] dt This is a symbolic thinking aid, not a universal solver or an established physical Theory of Everything. Suggested local reading for this exercise: Z=context/frame of reference; T_String=thread connecting states; D_M=model-related dissonance. These are proposed play definitions, not established definitions for the entire MAAT project. DeltaE=effort/friction; DeltaQ=uncertainty. The integral represents repeated consideration over time. T_String does not assert physical string tension. For a paradox: state the conflict precisely; separate terms, levels and assumptions; investigate countermodels; trace changes under a different perspective; evaluate with H/B/S/V/R. State whether the issue is solved, dissolved by disclosed new assumptions, or still open. The sum in the symbolic formula does not replace the product structure of Stability. Respect remains a separate acceptance condition. Do not claim to solve every paradox. A contradiction may persist within a specified system. ACCEPTANCE CHECKS Verification: Are important factual claims supported? Consistency: Does the conclusion follow from the assumptions? Uncertainty: Are limitations and open questions visible? Respect: Are binding boundaries upheld? A high score cannot replace these checks. Simpler explanations, counterexamples and negative findings remain valid.
Four checks before accepting a result
Verification: important facts supported? Consistency: conclusion valid? Uncertainty: limitations visible? Respect: binding boundaries upheld? A high score replaces none of these checks.
Sources & context
Open to inspection.
Open to revision.
As of 7 October 2026. MAAT is presented here as a philosophical and heuristic framework. Its usefulness in a particular application requires separate evaluation.
- MAAT RPG · Formula documentation
- MAAT RPG · Introductory formulas
- MAAT RPG · Further formulas on the project page
- Christof Krieg · Additional MAAT specifications and the symbolic ToE expression.
A comparison with simpler heuristics, fixed criteria, independent ratings and counterexamples would be a useful next step. Negative and null results count, too.