Skip to content
AFTER CERTAINTY

Nonfiction

Coupling

Cohesion, Consequence, and the Architecture of Responsibility

By Kevin Steffensen

A practical framework for seeing where systems are tightly joined, where they drift apart, and how those coupling choices shape accountability over time.

Offered under open publishing terms (CC BY-SA). Download formats are free where available; purchase links support print editions.

Type
Nonfiction · monograph
Author
Kevin Steffensen
Published
2026
Chapters
37 chapters
Formats
EPUB, PDF, DOCX

Also in Core After Certainty

View all 6

The question this book explores

How do connection and separation in systems shape responsibility over time?

Why this book exists

Coupling offers a practical architecture for seeing where systems are tightly joined, where they drift, and how those choices distribute consequence and accountability.

Who it is for

Systems thinkers, architects, and practitioners who need a concrete lens on cohesion, drift, and designed friction.

What it is not

  • It is not a software-only coupling guide, though software examples appear.
  • It does not treat decoupling as always virtuous.

Central ideas

A curated handful of concepts and patterns this book develops — not the full inventory.

  • Concept

    Boundary

    In this book: Architectural hinge: where joining and separation redistribute responsibility and consequence across systems.

    Explore the concept
  • Concept

    Accountability

    In this book: Follows how tight or loose coupling changes who can see and answer for effects over time.

    Explore the concept
  • Concept

    Agency

    In this book: Clarifies practical capacity inside coupled systems—where action can still matter when links constrain options.

    Explore the concept
  • Concept

    Attention

    In this book: Names what must be watched at seams and interfaces so drift and exception-paths stay visible.

    Explore the concept
  • Concept

    Authority

    In this book: Tracks steering rights across coupled units—who may change the join when consequences cross boundaries.

    Explore the concept
  • Pattern

    Exceptions are Forever

    In this book: Applies how coupling workarounds become permanent architecture—provisional joins that never uncouple.

    Explore the pattern
  • Pattern

    Feedback Drives Change

    In this book: Shows when outcome-signal still reaches the join in time to shift direction—healthy coupling.

    Explore the pattern
  • Pattern

    Learning Collapses

    In this book: Marks failure when coupling hides consequences until the group can no longer read what its choices are doing.

    Explore the pattern

Explore all concepts and patterns for this book

Inside this book

How the book is organized — parts, chapters, and what each section investigates.

About 204 min of reading across 37 sections in 6 parts.

Part I - The Structural Grammar

5 sections
  1. Show summary

    Cohesion moves from software components to people and institutions — where responsibility lives, whether boundaries are clear enough to learn when something goes wrong, and how fragmentation begins before anyone formally notices.

    Central questionWhere does responsibility actually live — and is that boundary clear enough to learn?

Part II - Software as Early Laboratory

8 sections

Part III - AI and Structural Entropy

6 sections

Part IV - Institutions Under Drift

5 sections

Part V - Oscillation and Design

6 sections

Other sections

7 sections
  1. Preface3 min
  2. IntroductionIntroduction2 min
  3. Coherence Under ScaleInterlude7 min

Where it fits

Suggested reading order and nearby volumes — not a mandatory sequence.

Continue exploring

Reading trails featuring this book

Editorial paths that include this book as one stop in a longer sequence.

Browse all reading trails

Full related inventoryShow

Complete concept, pattern, and thinker lists for readers who want the full semantic map.

Major concepts

Concept

Accountability

Accountability names whether consequence still reaches the people who can respond. It becomes important when scale, narrative, or abstraction thins contact between action and effect. It helps preserve correction and proportionate answerability, but can fail when moral sorting replaces sustained contact with outcomes. It differs from responsibility because accountability emphasizes exposure to consequence; responsibility includes the ongoing obligation to account even when control is partial.

View Concept →

Concept

Agency

Agency names whether a person or institution can actually move the scene, not only be named on the org chart. It becomes important when speech, responsibility, or moral weight outrun real leverage. It helps preserve honest coordination by distinguishing who can act from who must only absorb consequence, but can fail when low agency is mistaken for bad faith or when structural limits are narrated as personal failure. It differs from responsibility because responsibility can remain assigned when agency has thinned; it differs from judgment because agency is capacity to act, not the evaluative practice of deciding well.

View Concept →

Concept

Attention

Attention names the scarce gate between signal and response. It becomes important whenever more reaches a person or institution than can be fully tracked. It helps preserve focus and timely response, but can fail when visibility metrics are mistaken for judgments about importance. In When Incentives Become the Moral Language, attention marks the domain where newsrooms still know some stories matter even if they will not perform—yet operational language shifted toward clicks, shares, and time on page. Audience behavior reflects habit, outrage, and placement, not the same thing as what the public needs for democratic judgment. It differs from bias because attention names the capacity itself; bias names the selective work performed through it.

View Concept →

Concept

Authority

Authority names who can coordinate action when full agreement is impossible. It becomes important whenever groups must move before every perspective aligns. It helps preserve direction and collective capacity, but can fail when operational power outruns justified recognition or visible consequence. It differs from legitimacy because authority is the capacity to steer; legitimacy is whether that capacity reads as earned and revisable enough to coordinate trust rather than performative compliance.

View Concept →

Concept

Boundary

Boundary names where one kind of obligation ends and another begins—between roles, institutions, scales, or interpretive frames. It becomes important when coordination must clarify who can decide, who must answer, and what counts as shared ground. It helps preserve legible responsibility and evaluative contact, but can fail when boundaries harden into walls that block correction. In the post-interpretive authority context, a boundary is where external understanding ceases to govern legitimacy while power continues. It differs from coupling because a boundary marks a shift in shape, not the strength of attachment between action and consequence.

View Concept →

Concept

Coherence Maintenance

The ongoing work required to keep a system governable: enough agreement about who decides, what happened, and what must change next for learning to remain possible. Coherence maintenance becomes expensive as independently evolving boundaries multiply. Systems sometimes pay for it through clear ownership and timely feedback; they sometimes pay through coordination substitutes that preserve throughput while thinning corrective capacity.

View Concept →

Concept

Cohesion

Cohesion names whether it is clear who owns a decision, role, or outcome inside a boundary. It becomes important when diffuse assignment lets harm travel without redesign authority. It helps preserve legible ownership and coordinated response, but can fail when cohesion hardens into exclusion or when group binding is mistaken for clear responsibility. It differs from coupling because cohesion concerns ownership clarity within a boundary; coupling concerns how tightly action remains attached to consequence across a chain.

View Concept →

Concept

Consequence Architecture

Consequence architecture names professional literacy for systems where generation or scale outruns informal caution. It becomes important when guardrails exist on paper but production paths bypass them, or when postmortems produce action items detached from redesign authority. It helps preserve corrigibility under acceleration, but can fail when organizations prove they cared while the same failure class repeats. It differs from coupling because consequence-architecture is designed structure; coupling is the resulting attachment strength. It differs from cohesion because architecture designs the whole path; cohesion names ownership clarity within one boundary.

View Concept →

Concept

Constraints

Constraints names the present-tense architecture of what can be done—rules, budgets, timelines, approvals, and technical ceilings. Good constraints channel behavior toward learning and accountability; poor constraints hide tradeoffs or displace consequence. It differs from constraint because constraints describes limits inside an operating system; constraint describes the originating pressure that once made a moral order necessary.

View Concept →

Concept

Context Collapse

Context collapse names what happens when boundaries that should separate concerns, audiences, or interpretive frames dissolve into a single shared surface—when distinct contexts that require different handling are compressed together in ways that look coherent but hide important distinctions. It becomes important when coordination depends on maintaining separate concerns but structure or speed causes them to merge. In software and AI systems, context collapse appears as low cohesion at speed: modules, prompts, corpora, or tools become entangled until ownership and redesign authority become unclear, where output looks syntactically valid or rhetorically fluent even though meaning mixtures govern action undetected. In communication and social media, context collapse occurs when messages intended for one audience are received and evaluated by another, when professional and personal contexts merge on a single platform. Context collapse makes interpretation harder because the appropriate frame is unclear; it makes redesign harder because concerns have become so entangled that there is no clean cut point. The question context collapse raises is: where did separate concerns collapse into one apparently coherent surface? Which contexts that should stay separate are now entangled?

View Concept →

Concept

Contract

Contract is used in a software engineering sense. A contract is the explicit behavioral agreement attached to an interface: what inputs are valid, what outputs are expected, what errors can occur, and what invariants are guaranteed. Strong contracts reduce ambiguity between components and make failures easier to diagnose.

View Concept →

Concept

Coordination

Coordination names the problem of acting together when lives are fragmented—when people must move collectively across distance, time, strangers, and difference without complete information or full agreement. It becomes important when individual judgment alone cannot sustain the group, when behavior must hold across contexts that no one person can see completely. Coordination helps preserve collective capacity by making expectations predictable, by enabling strangers to act together under shared rules, by allowing large-scale projects that no single person could accomplish. Roles, rituals, law, canon, procedure, and bureaucracy are all coordination technologies—tools for making tomorrow recognizable from yesterday so action can continue. But coordination can become brittle when it suppresses the feedback and variation it needs to stay responsive, when inherited patterns keep moving after fit has thinned, when dissent is treated as disruption rather than signal. Coordination purchased at the cost of revision can continue functioning long after it has stopped serving the purposes that gave it authority. The question coordination raises is: what coordination problem is this tool solving? What breaks if behavior stops being predictable? Coordination differs from alignment in that coordination is about collective movement, while alignment is about shared direction or values.

View Concept →

Concept

Coordination Pressure

The structural strain that appears when many actors must stay aligned under scale while holding partial information and changing on different timelines. Coordination pressure is not the same as everyday busywork. It names the cost of maintaining a workable shared picture of responsibility, facts, and next actions when cohesion and coupling weaken. It often appears as approvals, queues, handoffs, and reporting layers that synchronize activity without returning consequence to a redesign-capable boundary.

View Concept →

Concept

Correction

Correction names the lived work of narrowing the gap between what was assumed and what is actually happening. It becomes important whenever systems, beliefs, or coordination can drift while still feeling operational. It helps preserve shared reality and learning, but can fail when every correction spawns new framing instead of updated contact with conditions. It differs from repair because correction updates belief and behavior; repair addresses damaged standing and trust. It differs from revisability because correction is the act; revisability is the designed capacity to keep correcting.

View Concept →

Concept

Coupling

Coupling names the strength of the link between what was decided and what later arrives. It becomes important when scale, abstraction, or delay lengthen the chain between decision and lived harm. It helps preserve learning and legitimacy, but can fail when dashboards stay green while frontline frustration grows. It differs from cohesion because coupling tracks action-to-effect attachment; cohesion tracks whether responsibility is clearly held. It differs from consequence-architecture because coupling describes the attachment; consequence-architecture describes intentional design of boundaries, feedback, and escalation.

View Concept →

Concept

DevOps

DevOps is used in a software engineering sense. DevOps refers to practices and operating norms that reduce the distance between software implementation and operations so that delivery, reliability, and learning stay structurally connected.

View Concept →

Concept

Distance

Distance names separation between decision and consequence, between action and its effects, between authority and those it governs. It becomes important as coordination extends beyond direct relational encounter—as scale grows and abstraction layers multiply. Distance makes harm harder to trace, feedback slower to arrive, and responsibility harder to locate. It helps coordination reach across space and time, enabling larger and more complex forms of collective action, but it thins the connection between choosers and those who live with the outcomes. As distance opens, action and reflection no longer meet in the same place; decisions travel while bodies receive the consequences; institutions record outcomes in ways that differ from how people experience them. Distance becomes built into the architecture of moral life at scale—not as failure but as structure. The question it raises is: how far does decision travel before consequence is felt? Who can see the full chain between policy and harm? Distance differs from mediation (routing through intermediaries) in that distance is the gap itself, while mediation is what fills or crosses the gap.

View Concept →

Concept

DORA

DORA is used in a software engineering sense. DORA refers to research-backed software delivery metrics and practices (for example, lead time for changes and change failure rate) used to evaluate learning-loop quality between development and operations.

View Concept →

Concept

Drift

Drift names slow misalignment—meaning drifts as it travels, role compliance replaces discernment, systems continue without admitting conditions have shifted. It becomes important when continuity masks thinning fit and inherited tools coordinate long after they stop persuading. It helps recognize strain before rupture feels sudden, but can fail when drift is mistaken for stability because the machinery still turns. It differs from inheritance because drift emphasizes loss of fit over time; inheritance emphasizes what persists. It differs from normalization because drift is the movement; normalization is drift settling into expected background.

View Concept →

Concept

Feedback

Feedback names how systems stay in contact with their own effects. It becomes important whenever action and consequence are separated by time, scale, or abstraction. It helps preserve learning, correction, and trust in ongoing coordination, but can fail when dashboards stay green while lived harm accumulates off the ledger. It differs from trust because feedback is the returning signal itself; trust is the bridge that lets people act before every signal arrives. It differs from correction because feedback is what returns; correction is the work of updating in response.

View Concept →

Concept

Finite perspective

Finite perspective names the condition under which all sight in this project begins. It becomes important whenever someone speaks as if their view were complete, neutral, or uninfluenced by position. It helps preserve humility, integration, and honest disagreement, but can fail when partial sight is treated as moral failure or when institutions demand omniscience from individuals. It differs from bias because finite perspective names the structural condition of partiality; bias names the selective work performed within it. It differs from trust because trust extends action beyond firsthand sight without pretending the extension erases partiality.

View Concept →

Concept

Governance Coupling

The degree to which formal oversight structures preserve real consequence return to decision-capable authority. Governance coupling weakens when signal is filtered, delayed, or politically insulated, even if nominal accountability mechanisms still exist.

View Concept →

Concept

Harm

Harm names where cost lands when influence is exercised—who absorbs it, who is kept in view, and what displacement is treated as acceptable. It marks the lived consequence of action, authority, design, or coordination. The visibility dimension matters—who is harmed in particular, not on average? Can they name it? Can they respond? Does the system register it or treat it as acceptable loss? Harm continues while interpretation expands, briefings make harm legible without smaller, people can still tell when something is doing harm. The question harm raises—is harm unchanged outside while we analyze inside? When does naming harm substitute for reducing it? The failure mode—harm persists while moral work feels complete, delays continue after excellent postmortems, harm becomes statistical at scale. The restoration path—return to particular harms and people, check effects outside the room after briefings, stop when outcomes for people are unchanged, pair analysis with named remedy attempts, accept harm as inevitable without abandoning response.

View Concept →

Concept

Hexagonal Architecture

Hexagonal architecture is used in a software engineering sense. Hexagonal architecture (also called ports and adapters) is a boundary pattern that keeps domain logic insulated from external technologies by routing integration concerns through explicit interface ports and adapter implementations.

View Concept →

Concept

Humility

Humility—naming tradeoffs, uncertainty, distributional conflict—is epistemically virtuous and politically costly at scale. Certainty orients anxious publics and protects institutions from looking paralyzed. The outperforming of humility is not proof audiences are childish; it is proof that orientation is scarce and fear is abundant. Under interpretive stress, humility also means acknowledging what summaries omit and who pays for resilience.

View Concept →

Concept

Incentives

Incentives are the reward and penalty patterns that influence behavior over time. Incentives include formal metrics, promotion logic, political pressures, funding structures, and social recognition patterns. Incentives do not determine behavior by themselves, but they strongly bias what actors optimize for.

View Concept →

Concept

Interface

Interface is used in a software engineering sense. An interface is the defined boundary through which one component interacts with another (for example, function signatures, API endpoints, message schemas, or protocol expectations). A good interface makes dependencies explicit while minimizing exposure to internal implementation details.

View Concept →

Concept

Judgment

Judgment names the capacity to decide when understanding is incomplete but harm, coordination, or responsibility will not wait. It becomes important whenever action must proceed under partial information and real consequence. It helps preserve proportionality, answerability, and room for revision, but can fail when judgment hardens into dogma, paralysis, or punitive finality. It differs from certainty because judgment accepts incomplete sight; it differs from preference because judgment weighs consequence beyond taste; it differs from optimization because judgment keeps contestability and human proportion in view; it differs from rule-following because judgment must still respond when rules lag reality.

View Concept →

Concept

Legibility

Legibility names the degree to which work, contribution, harm, or signal can be perceived and evaluated by those at distance—especially by those who grade, fund, allocate resources, or make promotion decisions. It becomes important because visibility shapes what gets valued, rewarded, and sustained. What becomes legible can be measured, managed, and integrated into formal systems; what remains illegible often goes unrecognized even when it is essential. Legibility helps coordination at scale by making expectations clear, progress trackable, and accountability visible. But uneven legibility is not the same as uneven value; legibility is a visibility structure, not a measure of worth. Legibility becomes dangerous when it substitutes for trust rather than serving it, when dashboards calm funders while field coordination narrows, when metrics answer anxiety rather than shared questions, when legibility demands consume the work they were meant to track. The question legibility raises is: is legibility serving trust or substituting for it? What work becomes invisible when only counted activity feels real?

View Concept →

Concept

Legitimacy

Legitimacy names the standing that lets authority coordinate without constant coercion. It becomes important when scale, abstraction, or time separate decision-makers from lived consequence. It helps preserve trust and willing participation, but can fail when smooth operation is mistaken for moral standing or when scrutiny is treated as disloyalty. It differs from alignment because legitimacy still invites proportionate inquiry; alignment can bind through belonging signals alone.

View Concept →

Concept

Operational Coupling

The degree to which the people who design or decide also remain close to execution and runtime consequence. Operational coupling is stronger when builders, operators, and reviewers share consequence exposure. It is weaker when handoffs separate decision authority from incident impact.

View Concept →

Concept

Reciprocity

Reciprocity names mutual exposure, obligation, and constraint made visible—the condition where both sides remain reachable by consequence and answerable to shared expectations. It becomes important when trust depends not only on rules but on mutual vulnerability, when accountability works through shared exposure rather than only through enforcement. Reciprocity helps preserve relationality and answerability in coordination; it makes obligation concrete and visible rather than abstract. When obligations are reciprocal, personal, and explicit, deviation threatens the relationship itself, not merely a principle. Trust is stronger when both parties can see and feel what happens if expectations break. But reciprocity thins as scale grows and abstraction replaces relationships—as categories replace persons, as responsibility shifts from mutual constraint to role compliance, as power travels beyond the reach of mutual consequence. Asymmetry grows when obligation flows only one way, when those who decide are insulated from the effects of their decisions. The question reciprocity raises is: who still shares exposure if this arrangement fails? Where has reciprocity thinned into one-way compliance? Reciprocity differs from accountability in that reciprocity is mutual, while accountability can be unidirectional.

View Concept →

Concept

Responsibility

Responsibility names what continues after action—the obligation to remain responsive to what one's choices set in motion. It becomes important when systems scale beyond any one person's control but moral weight still accumulates. It helps preserve sustained answerability, but can fail when responsibility collapses into futility or when symbolic gestures replace contact with effects. It differs from accountability because responsibility names the enduring relation; accountability names whether consequence still reaches those who can respond.

View Concept →

Concept

Scale

Scale names what happens when coordination extends beyond the range of direct relationship—when groups, institutions, systems, or problems grow large enough that no one person can see the whole picture, when distance opens between decisions and consequences, when strangers must act together without kinship or shared memory. Scale becomes important because it enables reach, continuity, repeatability, and collective capacity that small groups cannot achieve; it makes large projects, long-term infrastructure, and complex coordination possible. But scale can also thin what holds moral life together at human range: context becomes harder to preserve, consequence becomes harder to trace, feedback slows and weakens, answerability diffuses across roles and distance. At scale, systems often operate on aggregates rather than particulars, on categories rather than persons, on procedures rather than judgment. Responsibility can shift from relational accountability to role compliance, where doing the job correctly feels sufficient even when outcomes are troubling. The question scale raises is: what changes when no one sees the whole picture? Where is scale eroding the capacity to see people rather than just patterns? Scale differs from size in that scale is about the structural changes coordination requires, not just numerical growth.

View Concept →

Concept

Stale Representation

A compressed summary of system state that was accurate enough when produced and misleading by the time it governs action. Stale representation appears in dashboards, compliance reports, model cards, filings, and risk metrics. It is a coupling failure: consequence changes faster than the return path that would force revision at a redesign-capable boundary.

View Concept →

Concept

System

A system is an organized set of people, processes, rules, and structures that produces repeatable outcomes over time. The term applies across domains: software architectures, teams, firms, agencies, governance structures, and media institutions. The shared criterion is not technology; it is patterned behavior under constraints, incentives, and feedback.

View Concept →

Concept

Temporal Coupling

The degree to which consequences return quickly enough to shape near-term decisions. Strong temporal coupling shortens the time between action and correction. Weak temporal coupling introduces lag, memory loss, and interpretive noise that can protect failing designs from timely revision.

View Concept →

Concept

Trust

Trust names the bridge that lets people act together when no one holds the whole scene. It becomes important when verification must extend beyond firsthand sight—through specialists, records, institutions, or repeated experience. It helps preserve participation across partial perspectives, but can fail when trust stops tracking reality and closure feels safer than checking. It differs from certainty because trust accepts incomplete verification; it differs from blind belief because it still depends on signals that can be tested, strained, or withdrawn.

View Concept →

Concept

Uncertainty

Uncertainty names the condition moral and social systems are built to navigate—the space where coordination must happen before truth is settled, where lives are fragmented and behavior must hold together under incomplete information. It becomes important because human action cannot wait for perfect knowledge; groups must decide, move, and coordinate even when the future is unclear, when disagreement is costly, when explanation cannot travel fast enough to settle disputes before consequences arrive. Uncertainty is not a problem to eliminate but a condition to navigate—the ground on which moral life operates. Certainty emerges as an adaptive response to uncertainty, compressing ambiguity into stable expectations so action can proceed. But uncertainty itself is neither failure nor freedom; it is the default state that requires coordination tools. The After Certainty project explores how to practice judgment, accountability, and responsiveness inside uncertainty rather than demanding that uncertainty be resolved before moral life can begin. The question uncertainty raises is: what uncertainty is this system trying to reduce? Is ambiguity dangerous because the stakes are genuinely high, or because failure now scales? Uncertainty differs from ignorance in that uncertainty is about incomplete information in contexts requiring action, while ignorance is simply lack of knowledge.

View Concept →

Concept

Waterfall

Waterfall is used in a software engineering sense. Waterfall refers to a phase-sequenced development model where requirements, design, implementation, testing, and release are treated as largely distinct stages with formal handoffs and review gates.

View Concept →

Major patterns

Supporting · Corrective

Bad News Reaches Power

Hierarchy, success, identity, fear, and incentives filter bad news as it moves upward. - Correction requires uncomfortable evidence - Hierarchy filters what reaches authority Systems learn and correct only when uncomfortable evidence can reach people with the authority and capacity to act.

View Pattern →

Pattern

Dissent is Welcomed

Problems, risks, or failures show up inside a group or setting. Timely correction depends on whether bad news can reach those who decide. - Speaking up can cost people socially or at work - Reporting lines filter or delay information - Good news travels easier than bad news - People learn what is safe to say Problems show up early. Reality reaches those who decide in time. If you see a problem, say it now. Dissent is welcomed here—we can't fix what we don't raise while it's still small.

View Pattern →

Supporting · Obscuring

Distance Conceals Consequence

Those with the greatest authority may have the least direct contact with what their decisions produce. - Authority often sits far from effects - Distance makes consequence optional in practice The farther decisions are separated from their effects by time, place, role, abstraction, or organizational layers, the easier it becomes for consequences to disappear.

View Pattern →

Pattern

Examples Accumulate

A leader reacts to a mistake, competing pressures, or unclear ground in a shared setting. People need to know what counts as OK. Clear rules are missing or thin. - People watch others when things are unclear - What people see moves faster than written rules - Early reactions set what feels allowed - Silence and emphasis both send a signal What the leader does—especially under pressure—quickly sets what others treat as acceptable. Careful what we model in the room; examples accumulate faster than anything we post on the wiki.

View Pattern →

Pattern

Exceptions are Forever

A group adopts emergency measures or exceptions during a crisis or rush. Moves made for speed or survival do not always roll back when things calm down. - Urgency justifies skipping normal limits - Short-term wins reinforce the exception - Rolling back takes effort and coordination - New ways of working turn into habit fast Workarounds and shortcuts turn into how the group works day to day. Let's be careful carving out another exception—exceptions are forever, and this one will outlive the emergency we made it for.

View Pattern →

Pattern

Feedback Drives Change

New facts, visible risk, or bad assumptions show up after direction is set. The group must choose: change course or protect the public story and what people already signed up for. - Going public makes reversal costly - Momentum favors staying the course - Saying we were wrong can threaten legitimacy - Small fixes are easier than full reversals The group shifts direction based on new facts or new risk. We're not wed to yesterday's plan—feedback drives change, and what we learned this morning changes what we ship.

View Pattern →

Pattern

Learning Collapses

Those who decide sit far from where effects show up. News from the ground does not reliably reach them in time. - Size adds distance from impact - Formal reporting filters or delays signals - Bad news is costly to raise - Good news travels easier than bad news They do not get a clear read in time on what their choices are doing. We're steering off summaries again. Learning collapses when the ground story never reaches this room.

View Pattern →

Pattern

Meaning Gets Distorted

A provisional interpretation is repeated without re-exposure to source context. Coherence is preserved at the expense of accuracy. - Confirmation bias filters incoming data - Social reinforcement rewards consistency - Identity defense resists revision Interpretation can harden in ways that no longer track reality. Meaning gets distorted when certainty outlives contact with what was actually said.

View Pattern →

Supporting · Corrective

Responsibility Follows Leverage

Collective explanations can dissolve responsibility, while individual blame can conceal unequal power and structural production. - Leverage is uneven - Blame and collectivity can both obscure capacity The greater a person's or institution's capacity to shape outcomes, prevent harm, benefit from a system, or repair consequences, the greater its responsibility.

View Pattern →

Major thinkers

Research sources

article

Escaping Capability Traps Through Problem Driven Iterative Adaptation (PDIA).

Andrews, Matthew, Lant Pritchett, and Michael Woolcock. "Escaping Capability Traps Through Problem Driven Iterative Adaptation (PDIA)." World Development 51 (2013): 234-244. https://doi.org/10.1016/j.worlddev.2013.05.011.

View Source →

article

Index Funds and the Future of Corporate Governance: Theory, Evidence, and Policy.

Bebchuk, Lucian A., and Scott Hirst. "Index Funds and the Future of Corporate Governance: Theory, Evidence, and Policy." Columbia Law Review 119, no. 8 (2019): 2029-2146.

View Source →

article

Manifesto for Agile Software Development.

Beck, Kent, Mike Beedle, Arie van Bennekum, Alistair Cockburn, Ward Cunningham, Martin Fowler, James Grenning, Jim Highsmith, Andrew Hunt, Ron Jeffries, Jon Kern, Brian Marick, Robert C. Martin, Steve Mellor, Ken Schwaber, Jeff Sutherland, and Dave Thomas. "Manifesto for Agile Software Development." 2001. https://agilemanifesto.org/.

View Source →

article

On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?

Bender, Emily M., Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?" Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT), 610-623. https://doi.org/10.1145/3442188.3445922.

View Source →

book

Site Reliability Engineering: How Google Runs Production Systems

Beyer, Betsy, Chris Jones, Jennifer Petoff, and Niall Richard Murphy, eds. Site Reliability Engineering: How Google Runs Production Systems. Sebastopol, CA: O'Reilly Media, 2016.

View Source →

article

Challenges and Success Factors for Large-Scale Agile Transformations: A Systematic Literature Review.

Dikert, Karina, Maria Paasivaara, and Casper Lassenius. "Challenges and Success Factors for Large-Scale Agile Transformations: A Systematic Literature Review." Journal of Systems and Software 119 (2016): 87-108.

View Source →

book

Accelerate: The Science of Lean Software and DevOps: Building and Scaling High Performing Technology Organizations

Forsgren, Nicole, Jez Humble, and Gene Kim. Accelerate: The Science of Lean Software and DevOps: Building and Scaling High Performing Technology Organizations. Portland, OR: IT Revolution Press, 2018.

View Source →

book

Continuous Delivery: Reliable Software Releases through Build, Test, and Deployment Automation

Humble, Jez, and David Farley. Continuous Delivery: Reliable Software Releases through Build, Test, and Deployment Automation. Boston: Addison-Wesley, 2010.

View Source →

institutional document

ISO/IEC 42001:2023 — Information Technology — Artificial Intelligence — Management System. Geneva: ISO, 2023

International Organization for Standardization. ISO/IEC 42001:2023 — Information Technology — Artificial Intelligence — Management System. Geneva: ISO, 2023.

View Source →

article

Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure.

Jensen, Michael C., and William H. Meckling. "Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure." Journal of Financial Economics 3, no. 4 (1976): 305-360.

View Source →

book

From Higher Aims to Hired Hands: The Social Transformation of American Business Schools and the Unfulfilled Promise of Management as a Profession

Khurana, Rakesh. From Higher Aims to Hired Hands: The Social Transformation of American Business Schools and the Unfulfilled Promise of Management as a Profession. Princeton: Princeton University Press, 2007.

View Source →

book

The DevOps Handbook: How to Create World-Class Agility, Reliability, and Security in Technology Organizations

Kim, Gene, Jez Humble, Patrick Debois, and John Willis. The DevOps Handbook: How to Create World-Class Agility, Reliability, and Security in Technology Organizations. 2nd ed. Portland, OR: IT Revolution Press, 2021.

View Source →

article

How Should Research Contribute to Instructional Improvement? The Case of Lesson Study.

Lewis, Catherine C., Rebecca R. Perry, and Akihiko Murata. "How Should Research Contribute to Instructional Improvement? The Case of Lesson Study." Educational Researcher 35, no. 3 (2006): 3-14. https://doi.org/10.3102/0013189X035003003.

View Source →

book

Clean Architecture: A Craftsman's Guide to Software Structure and the Design of Professional Systems

Martin, Robert C. Clean Architecture: A Craftsman's Guide to Software Structure and the Design of Professional Systems. Upper Saddle River, NJ: Prentice Hall, 2017.

View Source →

article

Model Cards for Model Reporting.

Mitchell, Margaret, Simone Wu, Andrew Zaldivar, Parker Barnes, Lucy Vasserman, Ben Hutchinson, Elena Spitzer, Inioluwa Deborah Raji, and Timnit Gebru. "Model Cards for Model Reporting." Proceedings of the Conference on Fairness, Accountability, and Transparency (FAccT), 220-229. https://doi.org/10.1145/3287560.3287596.

View Source →

institutional document

Artificial Intelligence Risk Management Framework (AI RMF 1.0)

National Institute of Standards and Technology. Artificial Intelligence Risk Management Framework (AI RMF 1.0). NIST AI 100-1. 2023. https://doi.org/10.6028/NIST.AI.100-1.

View Source →

institutional document

Secure Software Development Framework (SSDF) Version 1.1: Recommendations for Mitigating the Risk of Software Vulnerabilities

National Institute of Standards and Technology. Secure Software Development Framework (SSDF) Version 1.1: Recommendations for Mitigating the Risk of Software Vulnerabilities. NIST SP 800-218. 2022. https://csrc.nist.gov/pubs/sp/800/218/final.

View Source →

article

The Impact of AI on Developer Productivity: Evidence from GitHub Copilot.

Peng, Sida, et al. "The Impact of AI on Developer Productivity: Evidence from GitHub Copilot." arXiv preprint arXiv:2302.06590 (2023). https://doi.org/10.48550/arXiv.2302.06590.

View Source →

article

Do Users Write More Insecure Code with AI Assistants?

Perry, Neil, Megha Srivastava, Deepak Kumar, and Dan Boneh. "Do Users Write More Insecure Code with AI Assistants?" Proceedings of the 2023 ACM SIGSAC Conference on Computer and Communications Security (CCS). https://doi.org/10.1145/3576915.3623157.

View Source →

book

Administrative Behavior: A Study of Decision-Making Processes in Administrative Organizations

Simon, Herbert A. Administrative Behavior: A Study of Decision-Making Processes in Administrative Organizations. 4th ed. New York: Free Press, 1997.

View Source →