Bycatch Aware Selective Intervention Design

When a selector catches more than its intended target, count the non-target capture, redesign the selector, and make success depend on bycatch reduction as well as target yield.

Version

v1 · 2026-08-24 · History

Solution archetype #

129

Problem family

Exclusion, Inequality & Distributional Harm

Problem subfamily

Collateral, Residual & Tail Harm

One-line summary

When a selector catches more than its intended target, count the non-target capture, redesign the selector, and make success depend on bycatch reduction as well as target yield.

When to use it

Use this archetype when a net, rule, classifier, filter, inspection program, eligibility screen, enforcement campaign, or extraction process is aimed at one target class but also captures or harms adjacent non-target classes. The defining warning sign is a target-centered scorecard: the system can look successful because it catches the intended target while non-target harm is invisible, delayed, dismissed, or externalized.

Core pattern

A selective intervention creates a boundary between target and non-target, but that boundary is imperfect in the real operating environment. Cues overlap, proxies are noisy, time and location windows are crude, and edge cases move through the same channels as true targets. If the metric only counts the target yield, operators have an incentive to widen the net rather than sharpen the selector.

Bycatch-aware design makes the collateral capture legible. It defines non-target classes, tests the selectivity window, records capture rate and harm, provides exclusion or release paths, assigns ownership, and changes the success metric so target yield cannot improve by quietly consuming non-targets.

When This Archetype Applies

Partial catalog groundingSome structural conditions are represented by existing abstractions, but no sufficient condition set is fully represented.

Diagnostic problem

A selective process is optimized for target capture, but the selector cannot perfectly distinguish target from non-target classes. Because the governing metric counts target yield more than non-target harm, the system has weak pressure to detect, release, compensate, or redesign around collateral capture.

Applicability expression6 distinct conditions

Selective target interventionandImperfect selector specificityandNon-target act exposureandTarget-only success metricandUnderreported delayed bycatchandIrreversible non-target harm

Algebraic1∧2∧3∧4∧5∧6

groundedpartly groundedopen

6 conditions, all required.

6Required in every casenumbered 1–6

These hold no matter which pattern applies.

1

Selective target intervention · grounded

A selective intervention or capture process is operating on a target class. more

primeBycatch— A selective process aimed at one target class also captures non-target classes because of the selector's finite specificity, and the harm persists because the success metric counts only the target.

and — also required

2

Imperfect selector specificity · grounded

The selector has finite specificity and cannot perfectly distinguish target from non-target at action time. more

primeBycatch— A selective process aimed at one target class also captures non-target classes because of the selector's finite specificity, and the harm persists because the success metric counts only the target.

and — also required

3

Non-target act exposure · grounded

Non-target classes are exposed to the same act as the target class. more

primeBycatch— A selective process aimed at one target class also captures non-target classes because of the selector's finite specificity, and the harm persists because the success metric counts only the target.

and — also required

4

Target-only success metric · grounded

The intervention's success metric counts target outcomes while excluding non-target burden. more

primeBycatch— A selective process aimed at one target class also captures non-target classes because of the selector's finite specificity, and the harm persists because the success metric counts only the target.

and — also required

5

Underreported delayed bycatch · open

Non-target harm is delayed or voiced too weakly to enter ordinary review. more

and — also required

6

Irreversible non-target harm · open

Non-target capture or harm is costly to reverse. more

Coverage

4 of 6 conditions grounded · 2 open.

Read the methodology · Download the trigger-logic data

Key components

Component	Description
Target Class Definition ↗	Name exactly what the selector is supposed to catch, flag, remove, admit, or affect. The definition must be operational enough to test.
Non-Target Class Register ↗	List valuable, protected, vulnerable, adjacent, invisible, or operationally important classes that may be captured despite not being the target.
Selector Specificity Profile ↗	Document the cues, proxies, thresholds, time windows, spatial bounds, data features, or procedural rules that determine what the selector actually captures.
Selectivity Window Boundary ↗	Identify where the selector is reliable and where target/non-target cue overlap makes bycatch likely.
Bycatch Rate and Harm Ledger ↗	Record non-target capture frequency, severity, reversibility, distribution, and downstream burden. Treat bycatch as data, not anecdote.
Expanded Success Metric ↗	Put non-target harm on the same scorecard as target yield. A selector that catches more targets by increasing collateral capture should not be scored as simply better.
Exclusion or Escape Pathway ↗	Provide a way for non-targets to avoid capture, be released, appeal, be restored, or be compensated. Size the pathway for real expected volume.
Selector Refinement Feedback Loop ↗	Use bycatch evidence to sharpen cues, narrow windows, change thresholds, improve training data, or replace the method.

Common mechanisms

A Non-Target Impact Pre-Mortem asks who else may be captured. A Selectivity Window Test probes selector specificity before scale-up. A False-Capture Audit samples caught cases to estimate bycatch. A Bycatch Rate Dashboard places non-target harm next to target yield. A Negative Filter or Exclusion Device keeps known non-targets out. An Escape Hatch or Release Protocol removes non-targets and feeds learning back into the selector. A Success Metric Reweighting penalizes bycatch. A Bycatch Tolerance Stop Rule suspends or narrows the intervention when harm crosses the agreed bound.

Boundary with nearby archetypes

This draft is close to accepted error_tradeoff_calibration, but the emphasis differs. Error tradeoff calibration chooses thresholds from false-positive and false-negative costs. Bycatch-aware design adds the surrounding governance pattern: non-target class ecology, target-only metric repair, release or exclusion pathways, harm ownership, and selector retuning.

It is also near hidden_type_screening, adverse_selection_filtering, and externality_internalization. Those archetypes cover screening, pool protection, and external cost responsibility. This draft focuses specifically on the bycatch structure: an imperfect target selector captures adjacent non-targets and hides that harm behind target-yield success.

Practical recipe

1. Define the target class and target action.
2. Register exposed non-target classes.
3. Map the cues, proxies, thresholds, windows, and channels used by the selector.
4. Test the selectivity window before and after changes.
5. Measure non-target capture rate, severity, reversibility, and distribution.
6. Add exclusion, escape, release, appeal, compensation, or restoration mechanisms.
7. Revise success metrics so target yield is scored together with bycatch cost.
8. Assign a harm owner and retune, suspend, or replace the selector when bycatch exceeds tolerance.

Example

A fishery targets one species, but the gear also captures turtles and juvenile fish. The original dashboard tracks target tonnage and compliance actions, so the operation appears to improve. Applying this archetype, the regulator defines non-target classes, samples catch composition, maps bycatch by gear type and season, adds exclusion devices and release protocols, and changes the scorecard so target catch is discounted by non-target harm. When the bycatch threshold is exceeded, the gear rule is retuned before the next deployment.

Failure modes

The common failures are target-yield tunnel vision, anecdote dismissal, overbroad retuning, paper escape hatches, proxy-discrimination bycatch, and compensation-only normalization. The safest posture is to assume every selector has finite specificity, then prove that non-target capture is measured, bounded, repaired, and used to improve or replace the method.

Common Mechanisms

11 documented mechanisms across 6 implementation forms.

The grouping reflects forms represented among the mechanisms currently documented for this archetype; an absent form is not necessarily an impossible implementation.

Assessment, Review & Assurance · 1 mechanism

• False-Capture Audit — An arm's-length review that samples what the selector actually caught, sorts true target from non-target, and reports a false-capture rate the operator can't self-certify away.

Control, Automation & Runtime · 3 mechanisms

• Compensation and Restoration Trigger — A standing rule that, once collateral harm is confirmed, automatically opens a route to make non-targets whole and pins the bill on a named accountable party.
• Negative Filter or Exclusion Device — A front-end cutoff, built into the selector, that admits the target while turning non-targets away before they are ever caught.
• Selector Retuning Cycle — A repeating loop that feeds observed bycatch back into the selector's settings, tightening specificity iteration by iteration and escalating to a different method when tuning stops paying off.

Experiment, Test & Rehearsal · 2 mechanisms

• Non-Target Impact Pre-Mortem — Before deployment, imagines the intervention has already caused off-target harm and works backward to name who gets caught and how — turning bycatch into a design input rather than a post-mortem finding.
• Selectivity Window Test — Sweeps the selector across its control variable to map where it separates target from non-target, locating the operating window in which selectivity holds and the edges where it collapses.

Monitoring, Sensing & Alerting · 2 mechanisms

• Bycatch Rate Dashboard — A live scoreboard that keeps off-target capture and its cumulative harm on screen next to target yield, broken out by non-target class, so bycatch can't hide behind a good headline number.
• Non-Target Sentinel Sampling — Watches a small, deliberately chosen panel of non-target classes as early-warning sentinels, sampling them directly so off-target harm surfaces in the field before it becomes systemic.

Protocol, Workflow & Routine · 1 mechanism

• Escape Hatch or Release Protocol — A built-in route by which a non-target that has already been caught can get back out before the harm is locked in.

Rule, Policy & Commitment · 2 mechanisms

• Bycatch Tolerance Stop Rule — A pre-committed limit on off-target harm that halts or forces redesign the moment bycatch crosses it — no matter how good target yield looks.
• Success Metric Reweighting — Rewrites the scorecard so a bycatch term counts against success, making off-target harm subtract from the headline number instead of sitting outside it, and names who owns that term.

Compression statement

Bycatch-Aware Selective Intervention Design applies when a net, filter, rule, classifier, campaign, enforcement program, procurement screen, or extraction process aims at one target class but also captures non-target classes because specificity is finite. The intervention expands the design from target yield to target yield plus non-target protection: define target and non-target classes, map the selectivity window, measure false capture, add escape or exclusion mechanisms, revise success metrics, and create feedback that retunes the selector before collateral harm becomes normalized.

Canonical formula: imperfect_selector + adjacent_non_target_classes + target_only_success_metric + hidden_capture_harm + bycatch_ledger + selectivity_refinement + escape_or_exclusion_path -> responsible_selective_intervention

Related Abstractions

Abstractions this archetype builds on — directly (a source ingredient) or as a related pattern. Links follow the typed catalog namespace.

Built directly on (8)

• Bycatch: A selective process aimed at one target class also captures non-target classes because of the selector's finite specificity, and the harm persists because the success metric counts only the target.
• Externality: Spillover effects.
• Feedback: Outputs influence inputs.
• Measurement: Mapping a target's attribute onto a scale via an instrument and procedure, yielding a value-plus-uncertainty tied to a unit and frame.
• Metric: A distance function on pairs obeying non-negativity, symmetry, and the triangle inequality.
• Screening: Inducing self-revelation.
• Selectivity Window: A process discriminates among targets only inside a bounded operating range of a control parameter, and loses or reverses that discrimination outside it.
• Type I & Type II Errors: False positive/negative.

Also references 15 related abstractions

• Accountability: Responsibility for actions.
• Boundedness: Values remain within limits.
• Catalysis: A facilitator lowers the barrier of a permitted-but-slow transformation on a specific pathway and returns unconsumed each cycle, so a small quantity transforms a large substrate over many turnovers.
• Classification: Sorting entities into discrete categories by explicit rules, turning unbounded variation into a finite, reusable map for downstream reasoning and action.
• Constraint: Limits possibilities to guide outcomes.
• False Positive Paradox: Under a rare base rate, most positive flags are wrong even when the detector is highly accurate.
• Funnel Analysis: Reading per-stage attrition across an ordered sequence to localize where a population is lost and which stage binds the final yield.
• Observability: Infer internal state externally.
• Path Dependence: Outcomes are shaped by the specific historical sequence of past choices, which lock in consequences and foreclose alternatives that persist despite present incentives to change.
• Resource Management: Allocation of finite assets.

▸ Show 5 more

Editorial Notes

Problem Classification

Classification: Exclusion, Inequality & Distributional Harm → Collateral, Residual & Tail Harm

Problem kernel: target yield hides non-target capture harm

Rationale: An imperfect selector optimizes desired capture while collateral cases remain weakly detected, released, compensated, or represented in its objective.

Independent corroboration: The earliest necessary condition in the frozen evidence is: A selective process is optimized for target capture, but the selector cannot perfectly distinguish target from non-target classes. That is a collateral residual and tail harm problem because Aggregate success leaves non-target, residual, or rare-case damage uncounted, unowned, uncompensated, or exposed to severe loss.

Review outcome: Independent reviewer agreement; high confidence.