Rules-as-code · Plain-English guide
How the Conservation Regulator's encoded rules assess an application for an Authority to Control Wildlife (ATCW): who assesses, how risk is tiered, when an inspection is needed, which part of the Wildlife Act 1975 empowers the decision, and who is allowed to make it.
An Authority to Control Wildlife (ATCW) lets a landholder or other applicant control protected wildlife — for example kangaroos damaging pasture — under section 28A of the Wildlife Act 1975. This ruleset does not decide the application form itself; it encodes the Conservation Regulator's internal assessment procedure: the checks an assessing officer works through and the guard-rails around who is allowed to make the final decision.
The encoded procedure answers five questions about each application:
This version (v2.0.0 as published on staging) is a substantial expansion of the earlier encoding: risk tiers, escalation, powers of approval, refusal grounds and the final decision-maker convergence are now all present. A handful of headings remain deliberate stubs — see Observations.
How to read this guide
Rules run top to bottom, and a later rule overrides an earlier one for the same property. The ruleset uses this deliberately: it starts every assessment at low risk with the Permissions Lead permitted to decide, lets any matching risk factor escalate that, and then a final "converge" step settles who may sign off. Green means permitted, red forbidden, orange obligated / required.
Idea 1
The assessor records facts — species, numbers, region, method, history. The rules combine them into conclusions: the risk tier, whether an inspection is needed, and who may approve. Change a fact and every conclusion downstream updates.
Idea 2
Every assessment begins low risk, decidable by the Permissions Lead. Each risk factor that matches pushes it up. Because later rules win, the highest tier reached is the one that sticks.
Idea 3
Low → Permissions Lead. Medium → Manager, Regulatory Operations. High → Director, Regulatory Strategy and Permissions. The final step forbids everyone below the required level from deciding.
Idea 4
An ATCW can only be issued under a specific provision of the Wildlife Act. If the requested activities and stated reason don't map to one of the empowering provisions, recommending issue is forbidden and refusal is flagged as reasonable.
The project's shared data model has 50 types; the ones this ruleset actually works with are below. "You provide" fields are recorded by the applicant or the assessing officer; "Rules work out" fields are computed and cannot be entered by hand.
The Conservation Regulator's working record for one application — the heart of this ruleset.
The parts of the wildlife application this assessment reads.
Which DEECA region the property sits in, plus each region's kangaroo cap (entered as data, not encoded — see Observations).
A Conservation Regulator staff member, with role flags the decision rules act on.
Every assessment must name an assessor and a decision maker, by staff ID — both fields are required. The rules then compute whether the two IDs differ and make that difference an obligation: the same person must never both assess and decide an ATCW application.
A second guard closes an obvious loophole: if the assessor is the Permissions Lead, then the decision maker is forbidden from being the Permissions Lead — so a low-risk application can't be assessed and signed off by the same role either.
A phone interview with the applicant is obligated for every application, without exception (procedure S2.6.2).
A property inspection (S2.6.3) is required when any one of these triggers fires:
For Eastern or Western Grey Kangaroo applications there is an additional, kangaroo-specific trigger list (public land; a wildlife or kangaroo management plan; the property adjoins a licensed wildlife shelter or foster carer; the method is not shooting; within 1 km of suburbia; property under 10 ha; concerns about the applicant's compliance history) — plus a regional numbers cap. Ask to control more kangaroos than your region's cap and an inspection is required:
| DEECA region | Kangaroos that may be controlled without an inspection |
|---|---|
| Loddon Mallee | up to 250 |
| Hume | up to 250 |
| Gippsland | up to 100 |
| Barwon South-West | up to 100 |
| Grampians | up to 100 |
| Port Phillip | up to 20 |
Where the caps live
Each cap is stored as an input constant on the region record rather than being hard-coded in a rule or lookup table, so the comparison is only as reliable as the constants supplied with the data. See Observations.
Every assessment starts at low risk. Two banks of triggers can raise it — and with it, who must be involved.
Medium risk — escalate to the Manager, Regulatory Operations — when any of these is true: first-time applicant · area of adverse community interest · unusually high wildlife numbers · numbers high for the property size or landscape · locally restricted species affected · a wildlife management plan would need to be prepared or recognised · this is the 4th (or later) successive application without reported alleviation of damage · possible/likely/almost-certain risk to health or human safety · the control method is uncommon for the species · possible/likely/almost-certain risk to animal welfare · the applicant has received a warning or sanction · the applicant has previously had an ATCW suspended or cancelled · conflict of interest (e.g. the applicant is a DEECA staff member).
High risk — escalate to the Director, Regulatory Strategy and Permissions — when any of these is true: the wildlife is a threatened species · the species is on the Conservation Regulator Advisory List · issue would present a possible/likely/almost-certain risk of adverse public or political outcry at state or national level · a possible/likely/almost-certain risk of aggravated cruelty that standard conditions cannot control.
Common wombats get a count-based tier instead: 1–30 animals is low risk, 31–60 is medium (Manager required), 61 or more is high (Director required). The rules use three cumulative thresholds and rely on later-rule-wins to keep the highest one.
Escalation also happens at the recommendation stage (S2.11.1 / S3.1.2): if the assessor does not recommend issue, the matter becomes at least medium risk and the Manager must decide; if it also involves a health-and-safety risk — or the assessor recommends suspension or cancellation of an existing ATCW — it becomes high risk for the Director.
Whenever the tier is medium or high, a written description of the risk-mitigation options explored becomes a required field.
An ATCW must rest on a specific empowering provision of the Wildlife Act 1975. The rules start from "no power" — the delegate is not empowered and recommending issue is forbidden — then check the requested activities against the stated reason for control. Activities that take or handle wildlife (destroy, control, buy, sell, acquire, receive, dispose, keep, possess, breed, process, display) map to section 28A(1); merely disturbing wildlife maps to section 28A(1A).
| Requested activity | Reason for control | Empowering provision |
|---|---|---|
| Destroy / control / take-and-handle activities | Wildlife is damaging property (buildings, crops, pasture, habitat…) | s 28A(1)(c) |
| Management, conservation, protection or control of wildlife; education, research or study | s 28A(1)(d) | |
| Ensuring the health or safety of any person | s 28A(1)(g) | |
| Supporting a recognised wildlife management plan | s 28A(1)(h) | |
| Disturb wildlife | Wildlife is damaging property | s 28A(1A)(a) |
| Supporting a recognised wildlife management plan | s 28A(1A)(b) | |
| Any other combination | Not empowered — refusal reasonable | |
When a row matches, the rules record the provision and its statutory wording on the assessment (the "reason to issue an authorisation" text), ready for the decision document.
If a power of approval was found, the assessor is permitted to recommend issue. If none was found, recommending issue is forbidden and the assessment is marked reasonable to refuse.
Even with a power in place, refusal is flagged as reasonable when:
For the most common case — damage to property under s 28A(1)(c) — there are additional discretionary grounds. Refusal is reasonable when, in the assessor's view: the damage will not be alleviated by the proposed control · the density of wildlife present is at a tolerable level for the property · there is no actual impact (only a perceived one) · the damage is seasonal and will cease naturally · or a lethal method is proposed but appropriate non-lethal measures have not been attempted.
Note the interaction with escalation: recommending refusal is itself a medium-risk trigger, so a refusal recommendation always puts the final decision in front of the Manager (or the Director, if health and safety is involved). The Permissions Lead never refuses an application alone.
After all the risk and power rules have run, a final "converge to delegate" step settles the decision-maker question — deliberately placed last so it overrides everything before it:
| Final risk tier | Permissions Lead | Manager, Regulatory Operations | Director, Regulatory Strategy & Permissions |
|---|---|---|---|
| High | Forbidden | Forbidden | Obligated |
| Medium (not high) | Forbidden | Obligated | Permitted |
| Low only | Permitted | Permitted | Permitted |
The low-risk row carries one extra condition: the Permissions Lead is only recorded as allowed when neither a Manager escalation nor a Director escalation was required along the way. The converge step also tidies up unanswered escalation flags (treating "not yet answered" as "not required") so a clean low-risk application doesn't stall on missing data.
Coverage at a glance
8 test suites, 18 scenarios, 33 expectations. All eighteen assert an outcome; none is inputs-only. This is a large step up from the previous encoding, which shipped a single smoke test.
The suites now trace the assessment end to end and, unusually for this project, test most branches both ways — a positive and a negative case for the same rule. The table lists all eight suites and what each pins down.
| Suite | Scenarios | What it asserts |
|---|---|---|
| S0-2 setup | 2 | Shooting makes the method lethal (True); with no lethal method it stays False |
| Assessment officers | 2 | Two different officers → “different officers” True; the same person → False |
| 2.4.2 Inspections | 4 | Kangaroo inspection triggered by activity and by number (True), a non-triggering case (False), and a wombat edge case — lethal control of 0 wombats (False) |
| 2.5 Delegate determination | 3 | Low risk → Permissions Lead allowed; a medium case → Manager required & PL not allowed; a high case → Director required, Manager & PL not |
| S2.6.2 Phone interview | 1 | The phone interview is Obligated |
| S2.6.3 Property inspection | 1 | Possible mis-identification triggers an inspection (True) |
| 3.1.3 Table 5 — powers | 2 | Destroy + damaging property → empowered, power = …Section 28A(1)c; destroy without damage → not empowered |
| Whole system — converge | 3 | End-to-end: low → Permissions Lead; medium-over-low → Manager outcome; high-over-anything → Director outcome |
What the tests cover well. The highest-value behaviour — the final decision-maker convergence — is now asserted for all three tiers, including the override-ordering cases (“medium on top of low resolves to medium”, “high on top of anything resolves to high”) that the last-rule-wins design turns on. The lethal-method flag, the different-officers check, the power-of-approval mapping and the inspection triggers each have a positive and a negative scenario, and the wombat “control 0” edge case is a nice touch.
Where it is still thin. A few substantial areas remain unasserted: the medium/high risk trigger banks are only exercised through their downstream delegate/convergence effect, never by asserting isMediumRisk / isHighRisk from a specific factor (e.g. threatened species → high); the wombat count thresholds (31 → medium, 61 → high) have no scenario; the refusal grounds (insufficient evidence, seasonal damage, non-lethal not attempted) are untested; and nothing checks the isRecommendationFromAssessorComplete gate or the assessor-vs-decision-maker escalation flagged in the Observations. A “threatened species → Director” and a “45 wombats → Manager” case would close the biggest gaps.
Points a reviewer or the rule author may want to check; these describe the encoding as published and are not legal advice. Several issues noted in the previous version are now fixed: the computed isLethal flag now covers shooting, poisoning and trapping-and-euthanasia (it was shooting-only, silently weakening the “non-lethal measures attempted” refusal ground); the recorded Act citation now includes the section number — “Wildlife Act 1975 Section 28A(1)c” and its siblings (the “28” was previously missing); and test coverage has grown from a single smoke test to 8 suites / 18 scenarios (see Test cases). What remains is below. Since this guide was last refreshed (15 July) two further additive changes have appeared: the high-risk trigger bank now directly obligates the Director as decision maker (note 1), and a risk-matrix data structure has been scaffolded but not yet used (note 5). Note: the published version string (v2.0.0) did not change across any of these edits, so it alone does not identify which behaviour is live — compare the rules directly.
1 · One medium-risk region obligates the assessor, not the decision maker
The first medium-risk trigger bank — the region headed “escalation to the Manager, Regulatory Operations” — obligates the assessor (personAssessor) to be the Manager, while forbidding the Permissions Lead from deciding. Every other escalation — the S2.11.1 refusal rule, the S3.1.2 twin and the final convergence — obligates the decision maker (personDecisionMaker) instead. The latest encoding also made the high-risk bank decision-maker-consistent (it now directly obligates the Director; see the note above), leaving this single medium region as the sole place that names the assessor. If the Manager must both assess and decide, that collides with the “different officers” obligation; if a senior re-assessment is intended, it is worth confirming, since everywhere else the ruleset escalates the decision maker.
2 · “Recommendation complete” is only ever set on the refusal path
isRecommendationFromAssessorComplete — described as the gate before the record goes to the delegate — is set (to true) only inside the “assessor does not recommend issue” branch (S2.11.1). An assessor who recommends issue never gets the flag set. A region titled “Need to confirm assessor has completed before delegate can sign” exists for exactly this check but is empty, so the gate is not enforced either way.
3 · Regional kangaroo caps are supplied as data
Each region’s cap (Loddon Mallee 250, Hume 250, Gippsland 100, Barwon South-West 100, Grampians 100, Port Phillip 20) is an input constant on the region record, not an encoded value or lookup table. A missing or mistyped constant silently changes when an inspection is required.
4 · Placeholder headings and authoring notes
Eleven of the 27 rule groups are empty headings — “Why the ATCW is necessary”, “Which species and what controls”, “How long for the issue”, “Risk mitigations”, regions marked “repeated — not needed”, “To be revisited” and “No code — see S3.1”, the Chief Conservation Regulator delegate step (“not coded — a workflow question”), and the incomplete-form region the authors deliberately left out. These read as authoring notes rather than gaps in the decision logic, but they will confuse anyone browsing the rule editor.
5 · A risk-matrix data model has been scaffolded but is not yet used
The data model now carries a new Risk assessment type — running tallies of low / medium / high risks, a harm to species rating (negligible / minor–moderate / major), a land-owner-consent flag and a running total — plus a risks assessment field on the Assessment record. No rule reads or writes any of them, so they have no effect on the current assessment outcome; they appear to be groundwork for a future quantitative risk-matrix step. Until rules are wired to them, the risk tier is still driven entirely by the yes/no trigger banks described in Risk tiers.