Compendium · Reference

Glossary

The terms of art this practice runs on, defined the way this site and the underlying papers actually use them — with pointers to where each one lives. Fifty-five terms, six families: the through-line, methods, framing, mathematics, data engineering, and climate.

ProvenanceThe through-line: Every figure traced to its source; methodology audited, results replicable. On this site it is not a feature of the work — it is the product: the discipline that makes a GHG inventory, a network graph, or a model defensible when someone challenges it.Where it lives: Services · One number, fully loaded · For enterprise & compliance · For method & verification
ReceiptThis site: The hover-or-focus citation attached to claims across these pages: the underlined phrase opens a small card naming the source. If a claim here matters, it should have one.Where it lives: any underlined claim on the Work pages · the specimen deliverable
Verification gateThis site: A checkpoint built into a workflow where AI or model output must pass verification against source material before it moves forward. The method pages put it plainly: AI output is treated as a draft, never a conclusion — if a step can’t pass its check, it doesn’t ship.Where it lives: For method & verification · How I work
Soft spotsThis site: The section of a deliverable that states its own weaknesses — which values are estimates rather than measurements, how old the data is, which rankings are less certain than the totals. Stated, not buried: uncertainty appears on the page, next to the finding it qualifies.Where it lives: the specimen deliverable · One number, layer 4
Provenance packThis site: The artifact bundle that ships with an inventory or analysis: the workbook, the calculation logic, and the source-and-method record behind every figure — runnable artifacts the client owns outright.Where it lives: Services · For enterprise & compliance
Audit trailThe through-line: The documented chain from a reported figure back through every calculation and source that produced it. When an assurer traces a figure they walk this chain backward; an undocumented link is a finding.Where it lives: For enterprise & compliance · Connectome
Systems thinkingBackbone: The habit of analyzing a problem as a set of interacting parts — stocks, flows, feedback, delay — rather than a list of separate issues. The backbone the rest of the practice hangs on.Where it lives: Services · Systems map · the Burkina Faso case
System dynamicsMethod: Forrester’s modeling tradition: stocks, flows, and feedback loops simulated through time, producing behavior-over-time rather than point forecasts. The essays put it plainly: the point of such a model is not prediction, it is behavior.Where it lives: Seventeen Goals, One SystemApplied in Zelinka & Amadei (2019) Part 2 and the 2021 IGI chapter (ch. 2).
Stock and flowSystem dynamics: The two primitives of a system dynamics model: a stock is an accumulation (water in a reservoir, trust in an institution), a flow is the rate that fills or drains it. Structure drawn this way can be simulated.Where it lives: the SDG essay’s diagram
Causal loop / reinforcing loopSystem dynamics: A circle of influence among variables. A reinforcing loop (marked R) amplifies itself each pass; a balancing loop resists change. The loop drawn on the Services page — evidence → model → decision → evidence — is reinforcing.Where it lives: Services · the Burkina Faso loop diagram
ArchetypeSystem dynamics: A recurring structure that produces a recurring behavior across many systems. The insecure-secure diffusion archetype — introduced at the System Dynamics Society conference in 2018 — models a system moving between insecure and secure states with flow allowed in both directions, and underlies the SDG model in Part 2.Where it lives: Publications (entry 7)Zelinka & Amadei (2018), 36th Int. Conf. of the System Dynamics Society.
Feedback and delaySystem dynamics: Feedback is a loop in which a system’s output feeds back into its own input — reinforcing (R) loops compound change, balancing (B) loops resist it. Delay is the time between cause and effect; delayed balancing feedback acts on old information, the classic source of oscillation in systems.Where it lives: the explorer · the Burkina Faso loop diagram
Cross-impact analysisMethod: A semi-qualitative, soft-systems method in which experts score how strongly each part of a system influences each other part; the scores assemble into a matrix. It does not measure the world directly — it makes expert belief explicit, consistent, and inspectable.Where it lives: the SDG essayZelinka & Amadei (2019), IJSDA 8(1), Part 1.
Cross-impact network analysisMethod: The Part 1 paper’s proposed term for cross-impact analysis joined with network analysis: the scored matrix read as a weighted network, so interactions can be analyzed with network tools rather than held as intuitions.Where it lives: Publications (entry 2)Term proposed in Zelinka & Amadei (2019), Part 1.
Network analysisMethod: Graph theory applied to real systems: nodes, weighted links, centrality, and community detection, used here on corruption networks, SDG interactions, and the practice’s own systems map.Where it lives: Bosnia case · Systems map · Governance & anti-corruption
CentralityNetwork analysis: A family of measures for how much a node matters in a network, each capturing a different sense of “influence”: betweenness (sitting on paths between others), PageRank (being pointed to by well-connected actors), and degree (raw connection count).Where it lives: the Bosnia case · Services
Community detectionNetwork analysis: Algorithms that find groups of nodes more densely connected inside than out — the network’s natural clusters, found rather than assumed.Where it lives: Services · Systems map
Graph data scienceMethod: The applied toolchain of network analysis: building graphs from records, then centrality, community detection, and visualization — NetworkX, SciKit-Network, Neo4j, Gephi. What the corruption-network work runs on.Where it lives: the Bosnia case · Governance & anti-corruption
Grounded theoryMethod: A qualitative synthesis method that lets categories emerge from what participants actually said rather than from a prior framework. Used to synthesize the Burkina Faso working sessions before any structure was drawn.Where it lives: the Burkina Faso case
Soft vs. hard systemsMethod: The 2019 papers’ pairing: a soft-systems method (cross-impact analysis) makes expert belief explicit and inspectable; a hard-systems method (system dynamics) simulates behavior through time. Each covers the other’s blind side.Where it lives: the SDG essay · PublicationsZelinka & Amadei (2019), IJSDA 8(1), Parts 1–2.
Complicated vs. complexFraming: The 2021 chapters’ dividing line: a complicated system (a car, a bridge) is dense with parts but ultimately knowable — a solution or optimum state exists. A complex system cannot be fully comprehended; there are too many parts interacting and doing too many things, so outcomes are understood through ranges of possibilities and average states rather than single answers.Where it lives: the dimensionality essayZelinka & Daher (2021), IGI chs. 1–2.
Dimensionality of systemsFraming: The 2021 chapter’s framework: any system can be located along three dimensions — structure (spatial anatomy), temporality (how strongly time governs it), and chaos (deviation from its mean behavior). Together they give every system an address, and the type of system tells you which tools will work.Where it lives: the interactive explorerZelinka & Daher (2021), IGI ch. 2, Figure 1.
TemporalityFraming: The time dimension of a system: feedback, cycles, delays, and timescales. The 2021 system dynamics chapter is a study of this dimension.Where it lives: the explorerZelinka & Daher (2021), IGI ch. 2.
ChaosFraming: Aperiodic behavior — a system that never settles into a repeating pattern, whether from genuine chance or from a perfectly deterministic equation. In the chapter’s terms: structure and dynamics describe a system’s average state, while chaos describes its variance and deviation from that mean. A chaotic system’s state is best described as a distribution, not a point.Where it lives: the explorer’s credible-region cloudsZelinka & Daher (2021), IGI ch. 2.
EmergenceFraming: Behavior of the whole that no single part contains, arising from interaction: 1 + 1 = 3. A murmuration of starlings is the chapter’s example.Where it lives: the explorer · the dimensionality essayZelinka & Daher (2021), IGI ch. 2.
Self-organizationFraming: Global order arising without central control: each agent follows local rules, and the pattern belongs to the collective.Where it lives: the explorer
NonlinearityFraming: Output is not proportional to input, and the effects of parts don’t simply add — superposition fails. Small causes can have large effects, and vice versa.Where it lives: the explorer
Credible regionFraming: The cloud of plausible positions a system occupies in the dimensionality space: chaos-driven deviation around its mean state, plus the epistemic uncertainty of judging its placement. Why the explorer draws systems as clouds, not dots.Where it lives: the explorer · the dimensionality essay
Logistic mapMathematics: A one-line equation — next = r · x · (1 − x) — that, as r grows, moves from settling, to oscillating, to never repeating at all. The classic demonstration that chaos needs no complexity to arise. It is drawn live in this site’s footer: the bifurcation diagram sweeping from order into chaos.Where it lives: the footer of every page · fractals field guide
Chaos gameMathematics: A generative procedure: pick a corner of a triangle at random, jump halfway toward it, repeat. Thirty thousand random jumps land on the same figure every time — the Sierpinski triangle, this site’s mark. Structure out of randomness, given the right constraint. The header logo redraws itself this way on every page load.Where it lives: the logo · the 404 page · colophon
Bifurcation diagramMathematics: The picture of a system’s routes to chaos: sweep a parameter and plot where the system settles — one value, then two, then four, then a dust of infinitely many. The footer of every page here draws the logistic map’s bifurcation diagram live.Where it lives: the footer · the explorer’s chaos lab · fractals field guide
Lyapunov exponentMathematics: The standard measure of how fast nearby starting points fly apart — λ > 0 is the signature of chaos. The explorer computes it live: push the logistic map’s growth parameter to 4.0 and λ reaches +0.69.Where it lives: the explorer’s chaos lab · the dimensionality essay
Strange attractorMathematics: The set a chaotic system’s long trajectories settle onto: state space folded and stretched until the attractor has fractional dimension — the shape of the system’s destiny.Where it lives: fractals field guide · fractal studio
Self-similarityMathematics: Structure that repeats across scales — exactly (the Sierpinski triangle in this site’s mark), approximately (the Mandelbrot set’s copies), or statistically (coastlines). The field guide separates the three kinds.Where it lives: fractals field guide · colophon
Phase spaceMathematics: The space of all states a system could occupy, one axis per variable; a system’s history is a trajectory through it. The dimensionality explorer is a phase space you can fly through.Where it lives: Atlas home · the explorer
Monte Carlo samplingMathematics: Answering a question by drawing many random samples instead of solving it in closed form. The explorer’s classifier runs 400 samples over your slider settings and reports which system families are plausible, not just the nearest box.Where it lives: the explorer’s classifier
Fuzzy classificationMathematics: Classification by degree of membership rather than crisp boxes, in the spirit of Zadeh (1965): a system can be mostly complex and partly complicated at once, and honest classification says so.Where it lives: the explorer’s classifier · the dimensionality essay
NexusFraming: From the encyclopedia chapter, verbatim:
“A nexus represents the interface of different sectors, areas, or relatively exclusive groups, like in a corporation, or disciplines in sustainable development (water, energy, etc.)”
Used here for the SDG nexus and the water-energy-food nexus: the couplings are the object of study, not the sectors alone.Where it lives: the SDG essayDaher & Zelinka (2021), Springer Encyclopedia of the UN SDGs.
Water-energy-food nexusFraming: The coupled system of water, energy, and food resources — sectors that cannot be managed in isolation because a move on one ripples into the others. Subject of the encyclopedia chapter and a running thread in the sustainability work.Where it lives: Sustainability & energy · PublicationsDaher & Zelinka (2021), Springer Encyclopedia of the UN SDGs.
Peace engineeringFraming: As practiced here: bringing systems-thinking methods to reform and conflict problems, designed with in-country partners rather than delivered at them.Where it lives: Peace engineering & conflict · the Burkina Faso case
ETLData engineering: Extract, transform, load: the pipeline work of pulling scattered source data together, reconciling identifiers and units, and landing it in one analyzable structure. Most of the invisible work in the power-plants case was this.Where it lives: the ERL case · Services · One number, layer 3
Gap-fillingData engineering: Estimating missing values with models trained on the observed ones — in the ERL work, gradient-boosting models with ten-fold cross-validation. The discipline that matters: model-filled values are estimates, not measurements, and honest data says which is which.Where it lives: the ERL case · One number, layer 4 · Climate & GHG areaGrant, Zelinka & Mitova (2021), ERL 16 094022.
GeocodingData engineering: Assigning geographic coordinates to records that lack them. In the ERL work: filling in missing facility locations across more than 30,000 plants, so every plant could be placed on the map — and joined to the policies governing it.Where it lives: the ERL case · Services
Spatial joinData engineering: Connecting datasets by location rather than by shared key: mapping each power plant to the climate and energy policies governing its coordinates, linking the facility layer to the policy layer.Where it lives: the ERL case · One number, fully loaded
Record linkageData engineering: Deciding that two records refer to the same person, company, or facility. It carries the most analytic risk in network work — it is where errors enter — so the linkage steps must stay inspectable.Where it lives: the Bosnia case
Gradient boostingData engineering: An ensemble machine-learning method that builds many small models, each correcting the last. The ERL gap-fill used Friedman’s GBM via the H2O.ai platform, fit with ten-fold cross-validation on the plants that do report.Where it lives: One number, layer 4 · the ERL caseGrant, Zelinka & Mitova (2021), ERL 16 094022.
Cross-validationData engineering: Testing a model on data it never saw during training — ten folds, each taking a turn as the held-out test set — so reported accuracy is earned, not memorized.Where it lives: One number, layer 4 · the specimen deliverable
GHG inventoryClimate: A greenhouse-gas accounting of an organization or territory — every emission source identified, quantified, and (held to this practice’s standard) traceable to its underlying data.Where it lives: Climate & GHG area · For enterprise & compliance · Services
Hyper-polluting plantsClimate: The extreme tail of the emissions distribution: the ERL paper found the world’s top 5% of polluting power plants contributed 73% of electricity-based CO₂ discharges in 2018 — 14.6 times an even distribution.Where it lives: One number, fully loaded · the ERL caseGrant, Zelinka & Mitova (2021), ERL 16 094022.
Emission factorClimate: The conversion between an activity and its emissions: a GHG number is activity data times an emission factor, repeated across every source. The provenance questions are always the same — which factor, from which published set, which version, applied to which activity total.Where it lives: For enterprise & compliance · One number, fully loaded
Activity dataClimate: The measured quantity of what an organization actually did — fuel burned, kilometers driven, kilowatt-hours purchased. One half of every GHG calculation; the emission factor is the other.Where it lives: For enterprise & compliance
Scope 1, 2, and 3Climate: The GHG Protocol’s three buckets: emissions from what you own and burn (Scope 1), from the energy you purchase (Scope 2), and from your value chain (Scope 3) — the hardest to measure and, under rules like California’s SB 253, increasingly the one that must be disclosed.Where it lives: For enterprise & compliance · Services · Climate & GHG area
Third-party assuranceClimate: Independent verification of a reported inventory. When an assurer traces a figure, they walk its chain backward — activity data, factor, version, calculation. If any link is undocumented, the whole line is a finding. Inventories here are built to survive that walk.Where it lives: For enterprise & compliance · Climate & GHG area
Emission intensityClimate: Emissions per unit of output — a plant’s CO₂ per unit of electricity generated. The ERL paper estimates what could be cut if hyper-emitting plants were held to intensity standards, switched fuels, or added carbon capture.Where it lives: One number, fully loaded · PublicationsGrant, Zelinka & Mitova (2021), ERL 16 094022.
DisproportionalityClimate: How unevenly a total is distributed across its sources. The ERL paper measured national disproportionalities in power-plant CO₂ and found they vary greatly and have mostly grown over time; the extreme case is 5% of plants producing 73% of emissions — 14.6 times an even spread.Where it lives: the specimen deliverable · PublicationsGrant, Zelinka & Mitova (2021), ERL 16 094022.