Adaptive Redirection

Version

v4 · 2026-08-30 · History

Prime #

1315

Origin domain

Organizational & Management Science

Subdomain

strategy under uncertainty → Organizational & Management Science

Aliases

Redirect with Preservation, Adaptation with Memory, Evidence Driven Redirection, Preserved Learning Pivot

Related primes

Adaptation, Learning

Core Idea

Adaptive redirection is the response to evidence that turns against a course of action: the agent selects a materially different course, holds the objective fixed across the change, and deliberately makes something the failed course produced — knowledge, capability, tooling, data, relationships — do load-bearing work on the new path. ^[1] The event is discontinuous in course and continuous in accumulated value, and that pairing is what separates it from both stubbornness and starting over.

Five commitments have to hold at once, and dropping any one of them changes what the event is. A current course must be underway toward an objective. Feedback must weaken the course specifically, leaving the objective standing rather than refuting it. The option set must contain three live alternatives — keep going, redirect, abandon — so that what follows is a selection rather than the only thing available. The course actually taken must differ materially along some dimension that was carrying weight, not merely in name or narration. And the retained material must still be load-bearing afterwards: delete it, and the new course becomes slower, costlier, or impossible. ^[1]

Ordinary vocabulary is far looser. "Pivot," "course correction," and "strategic refresh" get applied to incremental tuning, to reroutes the agent had no power to refuse, to renamings, and to clean restarts, all with the same word. What the prime supplies is a refusal of all four and a decision procedure — a deletion test run against each retained asset — that settles the ambiguous cases without appeal to how the change was announced or how it turned out. ^[2]

Structural Signature

Disconfirmed course under a surviving objective → live three-way option set → material change of course → carry-forward whose removal would degrade the new course. ^[3]

Recurring features:

• An objective that outlives the course, so the event is re-routing rather than goal revision.
• Directional evidence able to discriminate "this path is wrong" from "this destination is wrong."
• A choice set in which persistence and abandonment are both genuinely open, not merely nameable.
• Material discontinuity along at least one weight-bearing dimension: mechanism, market, method, route, or counterparty.
• A durable update produced by experience on the rejected path, already existing at the moment of the change.
• A carry-forward channel with counterfactual bite, which is what divides a retained asset from ceremonial continuity.
• Post-change accounting that scores the new course as the continuation of one effort rather than the opening of a second one. ^[3]

What It Is Not

The prime is a classification, not a recommendation. It does not hold that redirecting beats continuing, that agents redirect too rarely, or that an organization capable of redirection is healthier than one that persists; persistence under weak disconfirming evidence is often correct, and the prime is entirely silent about which option a given situation warrants. Nor does it predict outcomes. A redirection that satisfies every structural commitment can fail badly, and a reckless restart with nothing carried forward can succeed. Classification here runs on the relation among course, evidence, options and retained value, not on what happened next. ^[4]

Several other things are outside its claims. It does not require drama, speed, or a public announcement — a quiet reallocation inside a research group qualifies if the roles are present, and a heavily publicized restructuring does not qualify if they are absent. It sets no threshold on how much must be preserved; one genuinely load-bearing asset is enough, and a long inventory of retained-but-inert items is not. It does not assert that preservation is beneficial, since the same channel that makes a new course cheap can import the assumptions that failed. It makes no claim that objectives are permanent, only that a change of objective is a different event from this one, to be described in its own terms rather than folded in. And it does not supply an evidence standard: how much disconfirmation should trigger reconsideration is a matter for the domain, not for the abstraction.

Broad Use

The prime's reach comes from a specific fact about its parts: the carry-forward channel is typed completely differently from field to field while the role it plays stays fixed. In pharmaceutical development the channel is a safety database, a manufacturing and stability package, and a validated assay, and a compound that fails in one indication carries all three into another. ^[5] In semiconductor and hardware engineering the channel is test fixtures, characterization data, and yield-learning from a silicon revision whose architecture has been rejected. In military operations it is reconnaissance already paid for, established lines of communication, and the tempo advantage of forces already in motion, redirected against a different objective point when an axis of advance proves untenable.

In clinical care the channel is the diagnostic workup, the tolerance profile, and the observed response pattern that a failed first-line therapy produced, all of which constrain and accelerate the choice of second line. In research programs it is instrumentation, cohort access, and negative results that eliminate a region of hypothesis space, redeployed after a central hypothesis collapses. In software platforms it is the data model, the operational tooling, and the customer relationships that survive a rejected interaction paradigm. In careers it is credentials, domain fluency, and a professional network carried from an occupation that has stopped rewarding them.

What these share is not vocabulary but the shape of the audit: in each, one can ask which items crossed the boundary and whether the new path would stand without them. Where that has a determinate answer the prime applies; where the retained items are named only in retrospect and nothing hangs on them it does not, however closely the surface narrative resembles a redirection.

Clarity

The confusion this prime dissolves is a false binary. Public argument about a struggling effort almost always organizes itself into two camps — hold the line, or cut the losses — and each camp reads any move by the other as evidence of the vice it fears, either sunk-cost stubbornness or a failure of nerve. Naming the third option as a structurally distinct thing, with its own entry conditions, converts an argument about character into a checkable question: does the evidence bear on the route or on the destination, and is there anything here that would still be worth something on a different route? ^[6]

A second confusion is an accounting one. It is correct that money and time already spent should not by itself drive a forward-looking choice, and that principle is often stretched into the claim that everything spent on a disconfirmed course was waste. The stretch does not follow. Expenditure on a failed path routinely produces assets that are indifferent to why they were built — a qualified supplier, a trained team, a measurement rig, an eliminated hypothesis — and those assets are forward-looking facts about present capability, not backward-looking claims on loyalty. Keeping the two apart is the difference between disciplined cost accounting and throwing away the only useful product of an expensive failure.

The third clarification is diagnostic rather than rhetorical. Because the objective is the invariant, the question "has this effort failed?" resolves into two much sharper ones: whether the goal is still worth wanting, and whether this particular way of pursuing it is still the best available.^[7] An effort can be abandoned because the first question was never asked separately from the second, and defended because the second was never asked separately from the first; both directions of failure are real. ^[8]

Manages Complexity

What the abstraction buys is permission to stop maintaining a narrative. A long effort that has changed direction accumulates an enormous amount of state that feels obligatory to carry: who argued for what, which assumptions were held by whom, what the original justification was, how much has been spent, and whether the whole thing should be read as a success or an embarrassment. Almost none of that is needed going forward. The prime replaces it with a three-slot ledger — the objective, unchanged and stated once; the superseded course, recorded as superseded and closed; and an explicit inventory of what crossed over, each item marked load-bearing or not.

Three specific tracking burdens drop away. The first is the live defence of the original choice: once the course is formally superseded, its authorship stops being a variable anyone has to manage, and the people who chose it stop having a stake in re-arguing it. The second is cumulative spend as an open question. Under the ledger it is closed twice over — as a forward-looking input it is irrelevant, and as a producer of assets it has already been converted into the carry-forward inventory, so there is nothing left to reason about. The third is the parallel bookkeeping that efforts under strain tend to develop, where an optimistic account and a pessimistic account of the same facts are maintained side by side because no one can say which is the real one.

The residual complexity is real but sits in one place: the carry-forward inventory, which must be enumerated honestly and re-checked as the new course develops, since an item load-bearing at the moment of change can quietly stop being so.

Abstract Reasoning

The abstraction licenses a four-step diagnostic that can be run on any candidate case, including ones described in vocabulary borrowed from a different field.

Step one: name the objective and check that it is the same before and after. If it moved, the event is goal revision or abandonment, and the redirection reading is unavailable no matter how the change was framed. Step two: locate the evidence and check what it targets. Evidence against feasibility of the route supports redirection; evidence against the value of the destination does not. Step three: establish that the option set was live. If persistence was impossible or abandonment was unthinkable, the change was a forced reroute, and the deliberate selection the prime requires was never made. Step four, the decisive one: enumerate what was retained and run a deletion test on each item — suppose it had not crossed over, and ask whether the new course would be measurably worse. Items that survive deletion unchanged are ceremonial. ^[9]

Each step catches a distinct impostor. Step one catches quiet goal-drift dressed as tactical flexibility. Step two catches the case where a program persists on a route whose destination has already been refuted. Step three catches the reroute presented as a decision, which matters because credit and blame are assigned very differently to the two. Step four catches the restart with sentimental attachments, where the retained items are named for continuity's sake and do no work.

The procedure also runs forward, which is its more useful direction. Once the carry-forward set is enumerated and tested, it predicts where the new course will be strong — precisely along the dimensions the retained assets cover — and where it will be fragile, since the new course inherits, along with each asset, whatever assumptions were baked into it on the path that failed.

Knowledge Transfer

What travels across substrates is the skeleton and the tests, not the content. The role structure transfers intact: a course, an objective that outlasts it, evidence with a target, a three-way option set, and a channel of retained value. So do the two instruments — the discrimination between evidence against the route and evidence against the destination, and the deletion counterfactual that separates load-bearing retention from ceremony. So does one substantive prediction: that the failed course's assumptions ride along inside whatever is carried forward, which is why redirections tend to fail in ways that rhyme with the original failure. ^[10]

What does not travel is nearly everything that decides a real case. "Material difference" has no substrate-independent meaning; changing a branching heuristic is a material change in a solver and a rounding error in a turnaround plan. Evidence standards are domain property, with a regulated trial, a market test and a battlefield report differing by orders of magnitude in both cost and finality. Tempo and reversibility do not transfer: some domains allow a redirection to be undone within a week, others fix it for a decade. Nor does valence — in entrepreneurial vocabulary a pivot is close to a virtue, in a clinical protocol a change of course is a deviation requiring justification, and importing one field's scoring along with the structure is a common and expensive error.

One boundary deserves separate mention. The prime requires deliberate selection, so processes that produce the same outward shape without an agent — a lineage whose structure acquires a new function, a market that reallocates capital through the exit of firms rather than the choice of any one of them — reproduce the carry-forward but not the selection. The structure is genuinely similar and the analogy is worth drawing, but the diagnostic steps that depend on an option set cannot be run on them.

Examples

Formal/abstract

Consider a branch-and-bound search for an optimal integer solution. The solver commits to a dive: a branching variable, an ordering, a region of the tree it intends to resolve. Partway down, the relaxation bound at the current node rises above the value of the best solution found so far, which is decisive evidence that nothing in this subtree can beat what is already in hand. The objective — prove an optimal solution — is untouched. Three options are genuinely open: continue deeper anyway, jump to a different open node under a different branching rule, or stop and return the incumbent with a gap.

The solver redirects, and the redirection is cheap only because of what crosses the boundary. The incumbent value carries over and immediately prunes nodes in the new region. The cutting planes generated during the failed dive are globally valid and tighten the relaxation everywhere in the tree. The pseudocost statistics accumulated while branching badly are exactly what tell the solver which variables to branch on next. Each survives a deletion test: strip the incumbent and pruning collapses, strip the cut pool and the relaxation loosens, strip the pseudocosts and branching reverts to guesswork. A cold restart on a fresh copy of the model would reach the same tree with none of it. Mapped back: the dive is the course, the bound is evidence against the route and not against optimality, the three open moves are the option set, and the incumbent, cuts and pseudocosts are a carry-forward set that is load-bearing under deletion — every role present, in a system with no strategy, no narrative, and no reputational stake in the choice.

Applied/industry

A compound misses its primary endpoint in a mid-stage trial for one indication. The sponsor's objective — bring an approved therapy from this molecule to patients — is untouched by the result, which speaks to a mechanism-population match rather than to the molecule's value. The live options are to press into a late-stage trial on a favorable subgroup, to redirect the compound to a second indication where the mechanism has independent support, or to terminate and free the capital.

Redirection to the second indication carries forward four things, each testable by deletion. The safety database from every subject dosed so far determines the starting phase in the new indication; without it, development restarts at first-in-human and loses years. The manufacturing and stability package means clinical supply exists on the day the new protocol opens; without it, nothing can be dosed. The bioanalytical assay and the pharmacokinetic model transfer directly, since they characterize the molecule rather than the disease. Investigator and site relationships shorten enrollment. All four are load-bearing, and the contrast case makes the point sharply: in-licensing a different molecule for the same second indication pursues the identical objective with no carry-forward at all, and is a restart. So is a program that keeps the failed indication, the failed endpoint and the failed population while renaming the protocol. Mapped back: indication is the course, the endpoint miss is route-directed evidence, the three-way choice is real because termination was budgeted and available, and the four transferred assets are what make this a continuation of one development effort rather than the first year of a new one. ^[5]

Structural Tensions

T1 — Preservation versus contamination. The carry-forward channel is what makes redirection cheap and is also how the failed course reaches into the new one. Tooling arrives with the workflow it was built for; a customer list arrives with the segment definition the evidence may have just refuted; a team arrives with the instincts that produced the disconfirmed plan. Nothing in the abstraction separates the reusable part of a failed course from the part that failed, because the two are often one object seen twice. Each retained item needs its own contamination check.

T2 — Redirection versus disguised persistence. Redirection is scored far more favorably than stubbornness, which gives any agent under pressure a strong reason to describe continued commitment in redirection vocabulary. A relabeling that leaves the objective, budget, staffing and operating plan untouched satisfies the language completely while failing the structure, and from outside the two are indistinguishable: the announcement reads the same either way. Only an audit of what materially changed can tell them apart, and that audit is usually available only to people with a stake in its outcome.

T3 — Evidence strength versus option availability. Waiting for evidence decisive enough to justify a change consumes the resources that make change possible. Weak early signals leave the carry-forward assets intact and the option set wide, but acting on them invites repeated half-turns. Strong late signals arrive after the budget, credibility and staff needed for a different course have been spent proving the old one wrong. The option set narrows while confidence rises, so peak warrant and peak capability rarely coincide, and that gap is where most redirections are decided.

T4 — Objective fixity versus objective elasticity. The prime requires the objective to survive, but the objective is described by the same agent making the change, at whatever level of abstraction it chooses. Stated broadly enough — serve our customers, advance the science — almost any change preserves it; stated narrowly enough — ship this product to this segment by this date — almost none does. Classification therefore rests on a level of description the abstraction does not fix, and a retrospective restatement can manufacture a redirection out of an abandonment.

T5 — The deletion test versus its unavailability. The counterfactual is the prime's sharpest instrument and the hardest to actually run. Whether the new course would have failed without a retained capability is never observed, so the judgment is reconstructed by the people who chose to retain the asset and who now have an interest in its having mattered. Successful redirections generate confident accounts of what carried them; failed ones generate equally confident accounts of what should have. Neither is evidence, and the test is reliable only before the change.

T6 — Preservation cost versus travelling light. Retention is not free. Keeping a team, a facility, a data pipeline and a set of relationships alive across a change consumes exactly the capacity the new course needs, during a period when they produce nothing. A redirection with a rich carry-forward set can be slower and costlier than a restart with almost none, while satisfying the definition far more richly. The prime requires that some retention be load-bearing; it says nothing about the optimum, and more preservation is not more adaptive redirection.

Structural–Framed Character

Adaptive Redirection sits on the framed side of the structural–framed spectrum, labeled mixed-framed at an aggregate of 0.5. All five diagnostics read at half, so the grade is an even split rather than one criterion dragging the rest.

What travels is a course-evidence-redirection-carry-forward skeleton: a current course aimed at an objective, feedback that weakens the course without necessarily weakening the objective, an option set in which persistence, redirection and abandonment are distinguishable, a deliberate move to a materially different course, and retained learning, capability or infrastructure that is load-bearing on the new path rather than ceremonial continuity. It survives across product strategy, research programs, clinical plans, engineering designs, military maneuver and organizational turnaround; domain vocabulary can be removed without changing the mechanism.

Human-practice-bound does the most work at its half weight, and it is what keeps the prime off structural ground: the definition needs an agent who deliberately selects among distinguishable options and deliberately carries specific value forward. This is change with selective memory, not drift. Institutional origin reads half — the paradigm settings are institutional, though no office or charter is required. Vocabulary travels at half, with pivot named as the entrepreneurial specialization. Evaluative weight is half: the mechanism is stated descriptively, but redirection is set against restart and abandonment in a way that carries mild favorable tone. Import-vs-recognize is half.

The grade means the prime applies wherever an option set and a real carry-forward channel exist. Check that the retained material is load-bearing before importing the strategic framing with it.

Substrate Independence

Adaptive Redirection is a highly substrate-independent prime — composite 4 / 5 on the substrate-independence scale. What travels is a four-role sequence carrying no domain vocabulary: a course disconfirmed by evidence while the objective it served survives, an option set in which persisting and abandoning are both genuinely open, a material change along some weight-bearing dimension, and a carry-forward whose removal would leave the new course slower or impossible. Product strategy, research programmes, clinical treatment plans, engineering design, military maneuver, and organizational turnaround preserve those roles intact. What holds it below the ceiling is a commitment the statement still carries: the roles presuppose an agent able to hold an objective fixed and choose among three live paths, so every demonstrated instance sits in deliberate, goal-directed territory.

• Composite substrate independence — 4 / 5
• Domain breadth — 4 / 5
• Structural abstraction — 4 / 5
• Transfer evidence — 4 / 5

Relationships to Other Abstractions

Current abstraction Adaptive Redirection Prime

Parents (2) — more general patterns this builds on

• Adaptive Redirection is a kind of Adaptation Prime

  Adaptive Redirection is the deliberate discontinuous-course species of adjustment to changed evidence, distinguished by preservation and redeployment from the old course.

• Adaptive Redirection presupposes Learning Prime

  A redirection qualifies only when experience on the rejected course produces a durable update that is retained and used to select or execute the new course.

Children (1) — more specific cases that build on this

• Pivot Domain-specific is a kind of Adaptive Redirection

  Pivot is a strict specialization of Adaptive Redirection.

Hierarchy paths (3) — routes to 2 parentless roots

• Adaptive Redirection → Adaptation

Show alternative paths (2)

Neighborhood in Abstraction Space

Adaptive Redirection sits in a sparse region of abstraction space (74^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely rather than landing on a neighbor.

Family — Trajectories, Thresholds & Path Dependence (39 primes)

Nearest neighbors

• Backtracking — 0.74
• Remediation — 0.70
• Reversibility Horizon — 0.70
• Return Path — 0.69
• Eventual Consistency — 0.69

Computed from structural-signature embeddings · 2026-09-10

Not to Be Confused With

Adaptive Redirection sits directly beneath Adaptation, its taxonomic parent, and the gap between them is the whole content of the child. Adaptation covers any modification of organization or behavior that restores fit to changed conditions, and most of what it covers is continuous: a firm reprices weekly, a control loop trims a setpoint, an organism acclimatizes. None of that involves selecting between named alternatives, and none of it requires that anything be deliberately conserved, because in continuous adjustment nothing is being left behind. Adaptive Redirection is the discontinuous species, and it adds two commitments the parent does not carry — a discrete choice among persistence, redirection and abandonment, and a conserved remainder that does work on the far side. Tuning a campaign's spend allocation is adaptation; scrapping the channel strategy while keeping the creative assets and the audience data that the failed channel produced is adaptive redirection.

The relation to Learning runs the other way: this prime presupposes learning rather than specializing it. Learning is the production of a durable update from experience, and it is complete when the update exists, whether or not anything is done with it. An analyst can learn that a supplier is unreliable and act on nothing. Adaptive Redirection is a course-selection event that consumes a learning output as an input; remove the durable update and the change becomes an arbitrary switch, because there is nothing to preserve and nothing to preserve it with. The two also differ in what fails when they fail: learning fails by not updating, redirection fails by updating and then not redeploying.

Path Dependence describes the same continuity between past and present without any of the agency. In path dependence, earlier states constrain later ones through lock-in — switching costs, installed base, accumulated convention — and the mechanism operates whether or not anyone wants it to and whether or not the result is good. Adaptive Redirection requires that the carry-forward be selected and made load-bearing on purpose. The sharpest way to hold them apart is to ask whether the retained material could have been left behind. If it could not, the case is lock-in wearing a strategy's clothes; if it could and was not, the retention is a decision and the prime applies. The two can coexist, and often the same asset is both: a database that is genuinely useful on the new course and also the reason a cleaner course was never seriously considered.

Escalation of Commitment is the near-neighbour that resolves the same evidence in the opposite direction. Both begin identically — a course under way, feedback that weakens it, prior investment on the table — and escalation answers by increasing commitment to the failing course, typically justifying the increase by the size of what has already been spent. Adaptive Redirection answers by leaving the course while extracting what the spending produced. The confusion is practical, not theoretical: escalation is routinely narrated in redirection language ("doubling down on a refined approach"), and the discriminating question is simply whether the course itself changed materially or only its intensity and its description.

Sunk Cost and Irreversible Commitment covers the reasoning error that drives escalation, and it is worth separating from this prime because the two are easy to read as opposites when they are actually orthogonal. Sunk-cost reasoning says past expenditure should not enter a forward-looking choice, and that is correct. Adaptive Redirection makes no claim about past expenditure as a claim on loyalty; it concerns present assets that happen to have been produced by past expenditure. A trained team is not a sunk cost, it is a current capability, and honoring it is not the fallacy. Confusing the two produces the characteristic mistake of over-corrected organizations, which discard usable capability in order to demonstrate that they are not attached to it.

Decision is the general act of selecting from an option set: present in every case of this prime, sufficient for none. A decision can select a new course and burn everything associated with the old one, which is often fully rational and is simply a restart. What this prime adds is a constraint on the relation between the option taken and the option left.

Exaptation shares the redeployment move and drops everything else: a structure shaped for one function comes to serve another, with no disconfirming signal, no option set, and no agent choosing between named courses. It is the useful limiting case, showing that reuse alone is not redirection and that this prime's addition to bare reuse is the whole evidence-and-selection apparatus around it.

Solution Archetypes

No catalogued solution archetypes reference this prime yet.

Notes

Two points that did not fit the body. The prime is scale-free in its agent but not in its option set: a solver, a person and a state can all instantiate it, but only where abandonment is genuinely possible, which is why redirection is hardest to identify in institutions that cannot terminate anything. And the load-bearing test is far weaker retrospectively than prospectively, so classifications of historical cases should be held more loosely than classifications made at the moment of decision.

References

[1] Ries, Eric. The Lean Startup: How Today's Entrepreneurs Use Continuous Innovation to Create Radically Successful Businesses. Crown Business, 2011. Defines the pivot as a change of strategy under an overarching vision that rarely changes, keeping validated learning from the abandoned strategy in play. registry ↩a ↩b

[2] Bajwa, Sohaib Shahid, Xiaofeng Wang, Anh Nguyen Duc, and Pekka Abrahamsson. "'Failures' to be celebrated: an analysis of major pivots of software startups". Empirical Software Engineering, 2017. Case survey of 49 startups distinguishing multiple pivot types and triggering factors, documenting the heterogeneous set of changes practitioners all label a pivot. registry ↩

[3] Kirtley, Jacqueline, and Siobhan O'Mahony. "What is a pivot? Explaining when and how entrepreneurial firms decide to make strategic change and pivot". Strategic Management Journal 44(1), 2023. Longitudinal study of 93 strategic decisions finding that firms change course only after new information conflicts with or expands prior beliefs, and that a pivot is an accumulated series of decisions altering some strategic elements while others are carried forward, so the post-change course reads as one continuing effort. registry ↩a ↩b

[4] Baron, Jonathan, and John C. Hershey. "Outcome bias in decision evaluation". Journal of Personality and Social Psychology 54(4), 1988. Establishes experimentally that judgements of decision quality are contaminated by knowledge of the outcome, so the quality of a decision and the quality of its result are separate objects of evaluation. registry ↩

[5] Pushpakom, Sudeep, et al. "Drug repurposing: progress, challenges and recommendations". Nature Reviews Drug Discovery 18(1), 2019. Documents that repurposed compounds are de-risked by existing safety, pharmacokinetic and formulation/manufacturing packages and can therefore enter development at a later phase, with shorter timelines and lower cost than a fresh programme. registry ↩a ↩b

[6] Staw, Barry M. "Knee-deep in the Big Muddy: A study of escalating commitment to a chosen course of action". Organizational Behavior and Human Performance 16(1), 1976. Shows commitment to a chosen course rising after negative feedback, establishing the persist-versus-withdraw framing that dominates argument about a failing effort. registry ↩

[7] Brandstätter, Veronika. "Holding On or Letting Go? Persistence and Disengagement in Goal Striving." In Emerging Trends in the Social and Behavioral Sciences. Wiley, 2018: 1–14. Supports only the separability of the goal question from the means question — an action crisis over the route "does not have to result in the abandonment of the goal, but may likewise be overcome by a renewal of the commitment." It does not say how often, or why, the two questions get conflated. registry ↩

[8] Kappes, Cathleen, and Kaspar Schattke. "You Have to Let Go Sometimes: Advances in Understanding Goal Disengagement." Motivation and Emotion, vol. 46, no. 6 (2022): 735–751. Supports that failure runs in both directions: "healthy or functional goal regulation consists of persistent goal pursuit … and flexible goal adjustment," so over-persistence and premature disengagement are both regulation failures. It does not establish that either is routine, nor that either arises from failing to ask the goal question separately from the route question; that diagnosis is this entry's own. registry ↩

[9] Lewis, David. "Causation". The Journal of Philosophy 70(17), 1973. Establishes counterfactual dependence – would the outcome have been different had it not occurred – as the test of whether a factor made a difference, which is the inference the deletion test runs against each retained asset. registry ↩

[10] Leonard-Barton, Dorothy. "Core capabilities and core rigidities: A paradox in managing new product development". Strategic Management Journal 13(S1), 1992. Shows that capabilities carried into a new development effort bring with them the skills, systems and values that made them work before, and so import the assumptions of the effort that produced them. registry ↩