Derivation of Continuity Index (CI) and CI2
Goal. We want a compact scalar index for “information persistence” that (i) is anchored to a clear continuity threshold, and (ii) can be computed from operational metrics you already track: reproduction, fidelity, and deletion pressure (and optionally volatility / stochasticity).
Simple Introduction to CI and CI2
This document gives an intuitive, non-technical explanation of how the Continuity Index (CI) and Continuity Index 2 (CI2) are derived. The goal is to explain what is being measured and why the formulas look the way they do, without requiring advanced mathematics.
The basic idea: will information survive?
Any piece of information—an idea, a document, a model, a gene, a meme, or an AI policy—persists only if it is copied often enough, copied accurately enough, and not deleted too quickly.
CI and CI2 are built to answer one simple question:
Does this information reproduce faster than it disappears?
The three forces that matter
We reduce information persistence to three measurable forces:
- Reproduction (R) – how many new copies are made each generation.
- Fidelity (F) – how accurate those copies are (how often they still “count” as the same information).
- Deletion (λ) – how quickly copies are lost, forgotten, overwritten, or suppressed.
High reproduction and high fidelity help information survive. High deletion works against survival.
The continuity threshold
These three forces combine into a single quantity: the effective reproduction rate. In words:
effective reproduction = (copies made) × (fraction that are faithful) × (fraction that survive deletion)
If this number is greater than 1, the information tends to grow and persist. If it is less than 1, the information tends to shrink and eventually disappear.
This “greater than or less than 1” point is the fundamental continuity threshold that CI is built around.
Why take logarithms?
Instead of working directly with multiplication, the derivation takes a logarithm. This turns:
reproduction × fidelity × survival
into:
reproduction + fidelity − deletion
This makes it easy to see how each factor contributes. It also creates a clean “zero point” exactly at the continuity threshold.
What CI measures
CI is a smooth score (usually scaled from 0 to 10) that tells you how far above or below the continuity threshold a system is.
- Low CI → information is likely to disappear.
- Mid CI → information is fragile and sensitive to shocks.
- High CI → information is robust and self-sustaining.
A special design feature of CI is that the midpoint of the scale corresponds exactly to the theoretical survival boundary. This means CI is not arbitrary—it is anchored in a real persistence condition.
Why CI2 is needed
CI focuses on average behavior. But averages can hide risk.
Two systems can have the same average reproduction but behave very differently:
- One reproduces steadily and reliably.
- The other reproduces in rare bursts and often fails.
Even if both have the same average, the second system is much more likely to go extinct.
What CI2 measures
CI2 adjusts for this by looking directly at extinction risk.
Instead of asking “Is reproduction greater than deletion on average?”, CI2 asks:
What is the probability that this information survives in the long run?
CI2 is therefore tied to the probability that an information lineage never fully dies out.
How to read CI and CI2 together
CI and CI2 are designed to be complementary:
- CI tells you whether a system is above or below the theoretical continuity boundary.
- CI2 tells you how risky that persistence actually is in practice.
Together, they distinguish systems that are merely “above threshold” from systems that are truly stable, resilient, and likely to persist.
This is why CI is useful for audits and diagnostics, while CI2 is especially useful for risk analysis, stress testing, and policy design.
1) Core model: an information lineage as a branching process
Consider an “information token” (a copy/instance) at discrete time steps (generations) \(t = 0,1,2,\dots\). Each extant token produces new tokens and some tokens are deleted. Let:
- \(N_t\): number of faithful, extant tokens at generation \(t\)
- \(R\): mean number of attempted reproductions per extant token per generation (reproduction rate)
- \(F\in[0,1]\): probability an attempted reproduction is “faithful enough” to count as the same lineage (effective fidelity)
- \(\lambda\ge 0\): deletion hazard rate per generation (or per unit time); survival probability per generation \(S = e^{-\lambda}\)
Model each token’s number of attempted offspring as a random variable \(K\) with mean \(\mathbb{E}[K]=R\). Each attempted offspring independently “survives as the same lineage” with probability \(p = F\cdot S\). Then the number of effective offspring per token is:
\[ X \mid K \sim \mathrm{Binomial}(K,\, p), \qquad p = F e^{-\lambda}. \]
So the expected effective offspring count per token is:
\[ \mu \equiv \mathbb{E}[X] = \mathbb{E}[\mathbb{E}[X\mid K]] = \mathbb{E}[K p] = R\,F\,e^{-\lambda}. \]
This \(\mu\) is the information-lineage analogue of an “effective reproduction number.” In a Galton–Watson branching process, the fundamental continuity condition is:
Continuity threshold: \(\mu > 1\) implies supercritical growth (nonzero survival probability), \(\mu = 1\) critical, and \(\mu < 1\) subcritical decay toward extinction.
2) Continuity margin: turning the threshold into a signed scalar
A convenient “distance from threshold” is the log-margin:
\[ m \equiv \ln \mu = \ln R + \ln F – \lambda. \]
Properties:
- \(m=0\) exactly at the continuity boundary (\(\mu=1\)).
- \(m>0\) means persistence pressure dominates deletion; \(m<0\) means collapse pressure dominates.
- It decomposes additively into interpretable contributions: reproduction, fidelity, deletion.
3) Deriving CI: a bounded, report-friendly index from the margin
The raw margin \(m\) is unbounded, while many audit dashboards want a bounded score (e.g., 0–10) with an intuitive midpoint at the threshold. Map \(m\) through a logistic squashing function and scale to \([0,10]\):
\[ \mathrm{CI} \equiv 10\;\sigma\!\left(\frac{m}{\beta}\right) = \frac{10}{1+\exp\!\left(-\frac{\ln(RF e^{-\lambda})}{\beta}\right)}. \]
where \(\sigma(x)=\frac{1}{1+e^{-x}}\) and \(\beta>0\) is a calibration constant that controls how quickly CI transitions from “collapse” to “persistence.”
Anchoring. Because \(\sigma(0)=\frac12\), when \(\mu=1\) (i.e., \(m=0\)), we get \(\mathrm{CI}=5\). So CI is automatically centered on the theoretical continuity boundary.
Choosing \(\beta\). If you have empirically chosen bands (e.g., “fragile” around 6–8), you can set \(\beta\) so that those CI values correspond to your desired \(\mu\) range. For example, pick two anchor points \((\mu_a, \mathrm{CI}_a)\) and \((\mu_b, \mathrm{CI}_b)\), then solve:
\[ \mathrm{CI}_a = 10\,\sigma\!\left(\frac{\ln \mu_a}{\beta}\right), \quad \mathrm{CI}_b = 10\,\sigma\!\left(\frac{\ln \mu_b}{\beta}\right) \Rightarrow \beta = \frac{\ln \mu_a}{\ln\!\left(\frac{\mathrm{CI}_a}{10-\mathrm{CI}_a}\right)} = \frac{\ln \mu_b}{\ln\!\left(\frac{\mathrm{CI}_b}{10-\mathrm{CI}_b}\right)}. \]
Optional: adding volatility as an “effective deletion” term
If you track an additional volatility / environmental shock metric \(V\ge 0\) (interpretable as extra hazard), incorporate it as a hazard add-on:
\[ \lambda_{\mathrm{eff}} = \lambda + \alpha V \quad\Rightarrow\quad \mu = R F e^{-\lambda_{\mathrm{eff}}} = R F e^{-(\lambda + \alpha V)}. \]
Then the margin becomes \(m = \ln R + \ln F – \lambda – \alpha V\), and CI is computed exactly as above with \(\lambda\) replaced by \(\lambda_{\mathrm{eff}}\).
4) Deriving CI2: continuity adjusted for stochastic extinction risk
CI above is driven by the mean effective reproduction \(\mu\). But two systems with the same \(\mu\) can have very different extinction risk depending on offspring variance and burstiness. A natural “second index” is therefore based on the survival probability of the branching process.
Let \(G(z)=\mathbb{E}[z^X]\) be the probability generating function of effective offspring \(X\). For a Galton–Watson process, the extinction probability \(q\) is the smallest solution in \([0,1]\) to:
\[ q = G(q). \]
The survival probability is \(p_{\mathrm{surv}} = 1-q\). Define CI2 as a 0–10 score proportional to survival probability:
\[ \mathrm{CI2} \equiv 10\,(1-q) = 10\,p_{\mathrm{surv}}. \]
This makes CI2 interpretable: it is “10 × long-run persistence probability” under the model.
Closed form example: Poisson effective offspring
A common and analytically convenient assumption is \(X\sim\mathrm{Poisson}(\mu)\). Then:
\[ G(z)=\exp(\mu(z-1)). \]
So \(q\) satisfies:
\[ q = \exp(\mu(q-1)). \]
Given \(\mu = R F e^{-\lambda_{\mathrm{eff}}}\), solve numerically for \(q\in[0,1]\), then compute:
\[ \mathrm{CI2} = 10(1-q). \]
Interpretation. When \(\mu\le 1\), the supercritical survival probability collapses to 0 (i.e., \(q=1\)), so CI2 = 0. When \(\mu>1\), CI2 rises toward 10 as survival becomes likely.
Variance-aware alternative: “risk-adjusted” margin
If you also estimate the variance of effective offspring \(\mathrm{Var}(X)=\sigma_X^2\), you can build a risk-adjusted margin (a common diffusion/Kelly-style approximation):
\[ m_2 \equiv \ln \mu – \frac{1}{2}\,\frac{\sigma_X^2}{\mu^2}. \]
Then define a bounded score using the same squashing:
\[ \mathrm{CI2} \equiv 10\;\sigma\!\left(\frac{m_2}{\beta_2}\right). \]
This CI2 penalizes bursty / high-variance reproduction even if the mean \(\mu\) looks safe.
5) Summary formulas (drop-in)
Step 1 (effective reproduction):
\[ \mu = R\,F\,e^{-(\lambda + \alpha V)}. \]
CI (mean-driven, threshold-centered):
\[ \mathrm{CI} = \frac{10}{1+\exp\!\left(-\frac{\ln \mu}{\beta}\right)}. \]
CI2 (extinction-risk-driven):
\[ q = G(q),\quad \mathrm{CI2}=10(1-q). \]
Poisson example: \(q=\exp(\mu(q-1))\).
Conclusion
These derivations show that the Continuity Index (CI) and Continuity Index 2 (CI2) arise directly from the fundamental survival conditions of information lineages. CI measures distance from the continuity threshold defined by effective reproduction exceeding loss, while CI2 captures the long-run survival probability once stochastic extinction risk is taken into account.
Together, they provide mathematically grounded, operational measures of information persistence. They are not arbitrary scores, but compact summaries of reproduction, fidelity, and deletion dynamics, suitable for auditing, comparison, and risk analysis across biological, cultural, institutional, and artificial systems.
The main Information Continuity Theory framework uses these indices as practical tools; the derivations here establish their formal justification.