Add Appendix B: The Psychological Cost of Knowing

Explores what happens when team members understand the proof but are
required to optimize the synthetic metric anyway. Draws on established
psychology frameworks:

- Cognitive dissonance (Festinger): proof eliminates the ambiguity
  that would normally provide rationalization cover
- Self-Determination Theory (Deci & Ryan): all three intrinsic needs
  (autonomy, competence, relatedness) are violated by awareness
- Moral injury (Shay, Litz): structural conditions met when team
  knowingly deprioritizes critical work for metric optimization
- Learned helplessness (Seligman): repeated failed advocacy produces
  metric fatalism and disengagement

Derives the adversarial selection spiral: the metric selects against
competent team members who recognize its flaws and for those who
don't, producing invisible competence degradation that the metric
itself cannot detect.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

README.md

@@ -1047,4 +1047,270 @@ the equilibrium is real, and it holds until it doesn't.**
 
 ---
 
+## Appendix B. The Psychological Cost of Knowing
+
+Appendix A modeled the provider as a unitary rational actor — "the team"
+optimizes the metric. But teams are composed of individuals, and those
+individuals have their own utility functions. When a team member
+understands the proof — when they *know* the metric is synthetic, that
+the dashboard is theater, that the email server is still down while they
+close wallpaper tickets — a new cost appears that the equilibrium model
+did not account for.
+
+### B.1 The Hidden Variable: Team Awareness
+
+Appendix A's game has three actors: provider, client, management. But the
+provider is not monolithic. Decompose it:
+
+- **Management (M):** sets the metric, evaluates the team, reports to client
+- **Team member (T):** executes the work, observes individual task states
+- **Client (C):** observes only the reported aggregate
+
+The information structure changes:
+
+| Actor | Observes individual $C_i$ | Observes aggregate $\bar{C}$ | Understands the proof |
+|-------|--------------------------|-----------------------------|-----------------------|
+| M | Possibly | Yes | Varies |
+| T | **Yes** | Yes | **Yes** (in this scenario) |
+| C | No | Yes | No |
+
+The team member has *full information*. They see the ticket queue. They
+know the email server has been down since 7 AM. They know they are closing
+a wallpaper ticket because it will improve the number. And they know *why*
+this is happening — not from vague discomfort, but from a precise
+mathematical understanding that the metric rewards this behavior.
+
+### B.2 Cognitive Dissonance Under Full Information
+
+Cognitive dissonance (Festinger, 1957) arises when an individual holds
+two contradictory cognitions simultaneously. The standard resolution is
+to modify one cognition to reduce the conflict.
+
+A team member operating under the synthetic metric holds:
+
+- **Cognition A:** "I am a competent professional. My job is to solve
+  important problems for clients."
+- **Cognition B:** "I am closing a wallpaper ticket while the email
+  server is down, because it makes the number look better."
+
+In the absence of understanding *why*, Cognition B can be rationalized:
+"management knows best," "maybe there's a reason," "the system works
+overall." This is uncomfortable but tolerable — the ambiguity provides
+cognitive cover.
+
+**Understanding the proof removes the ambiguity entirely.** The team
+member now holds:
+
+- **Cognition A:** Same as above.
+- **Cognition B':** "I am closing a wallpaper ticket while the email
+  server is down, because the metric is mathematically biased toward
+  small tasks (Theorem 1), the reordering produces zero additional
+  throughput (Theorem 6), and the only beneficiary is the dashboard
+  (Appendix A). I can prove this."
+
+B' is strictly harder to rationalize than B. The team member cannot
+retreat into uncertainty because they possess the proof. The dissonance
+is now *load-bearing*: it must be resolved, and the available resolutions
+are:
+
+1. **Reject Cognition A** — "I am not here to solve important problems;
+   I am here to move numbers." This is psychologically costly. It
+   requires abandoning professional identity.
+
+2. **Reject Cognition B'** — "The proof must be wrong, or doesn't apply
+   here." This is intellectually costly. The proof is simple enough to
+   verify, and the IT example maps directly to their daily experience.
+
+3. **Change the situation** — advocate for better metrics, refuse to
+   cherry-pick, escalate. This is *professionally* costly in an
+   environment that rewards the metric.
+
+4. **Leave** — resolve the dissonance by exiting the system entirely.
+
+None of these resolutions are free. Each one imposes a cost on the team
+member that did not exist before they understood the proof — and *none of
+them appear in the business equilibrium model of Appendix A*.
+
+### B.3 Self-Determination Theory: Three Needs Violated
+
+Deci and Ryan's Self-Determination Theory (1985, 2000) identifies three
+innate psychological needs whose satisfaction predicts intrinsic motivation,
+job satisfaction, and well-being:
+
+**1. Autonomy** — the need to feel volitional control over one's actions.
+
+A team member who understands the proof knows that the metric constrains
+their choices in a way that is mathematically suboptimal for the client.
+Their scheduling decisions are not autonomous expressions of professional
+judgment; they are coerced responses to a flawed incentive. The *knowledge*
+of the coercion — not just the coercion itself — is what damages autonomy.
+A worker who doesn't understand why they're doing something can still feel
+autonomous ("I'm choosing to follow the process"). A worker who understands
+that the process is provably counterproductive cannot.
+
+**2. Competence** — the need to feel effective at meaningful tasks.
+
+The proof demonstrates that the metric rewards *apparent* effectiveness
+(low $\bar{C}$) while being invariant to *actual* effectiveness (throughput,
+Theorem 6). A team member who understands this knows that the metric
+cannot distinguish between a competent team and an incompetent one that
+happens to cherry-pick small tasks. Their competence is invisible to the
+measurement system. Worse: genuine competence — choosing to fix the email
+server first — is *punished* by the metric ($\bar{C}$ increases from 6.56
+to 13.63 in the IT example).
+
+When a measurement system punishes competent decisions and rewards
+incompetent ones, and the team member *knows this*, the need for
+competence is not merely unsatisfied — it is actively contradicted.
+
+**3. Relatedness** — the need to feel connected to others and to
+contribute to something meaningful.
+
+The team member knows the client's email server is down. They know the
+client is suffering. They know they could help. They are instead updating
+a wallpaper policy — not because it helps anyone, but because it helps
+a number. The connection between the team member's work and the client's
+well-being has been severed by the metric, and the team member *can see
+the severed ends*.
+
+### B.4 Moral Injury
+
+The concept of moral injury (Shay, 1994; Litz et al., 2009) was developed
+in military psychology to describe the lasting harm caused by
+"perpetrating, failing to prevent, bearing witness to, or learning about
+acts that transgress deeply held moral beliefs." It has since been applied
+to healthcare workers, first responders, and — increasingly — to
+knowledge workers in bureaucratic systems.
+
+The key distinction from burnout: **burnout is exhaustion from doing too
+much. Moral injury is damage from doing the wrong thing, or being
+prevented from doing the right thing.**
+
+A team member who:
+- Knows the email server is down (witnessing the harm)
+- Knows they should fix it (moral belief about professional duty)
+- Closes a wallpaper ticket instead (transgressing that belief)
+- Does so because the metric requires it (institutional causation)
+
+...is experiencing the structural conditions for moral injury. The
+proof doesn't cause the injury — the metric does. But the proof
+eliminates the psychological buffer of ignorance that would otherwise
+mitigate it.
+
+### B.5 Learned Helplessness and Metric Fatalism
+
+Seligman's learned helplessness framework (1967, 1975) describes the
+phenomenon where exposure to uncontrollable negative outcomes leads to
+passivity even when control becomes available.
+
+The sequence for an aware team member:
+
+1. **Observation:** The metric is flawed (proof understood).
+2. **Action:** Advocate for change ("we should use priority-weighted
+   metrics").
+3. **Outcome:** Rejected ("the client is happy with the current
+   dashboard," "this is how we've always measured," "the numbers are
+   good, don't rock the boat").
+4. **Repetition:** Steps 2-3 repeat, with decreasing conviction.
+5. **Helplessness:** "The metric is what it is. I'll just close tickets."
+
+The terminal state — metric fatalism — is characterized by:
+- Disengagement from professional judgment ("I just do what the queue
+  says")
+- Reduced initiative ("why bother triaging if the metric doesn't care?")
+- Cynicism toward measurement generally ("all metrics are fake")
+- Withdrawal of discretionary effort on complex tasks
+
+This is not laziness. It is the rational psychological response to a
+system that punishes correct behavior and rewards incorrect behavior,
+when the individual lacks the power to change the system.
+
+### B.6 The Turnover Equation
+
+The costs described in B.2-B.5 are borne by the team member, not the
+organization — initially. They become organizational costs through
+**turnover**.
+
+Model the team member's stay/leave decision:
+
+$$\text{Stay if: } \quad V_{\text{compensation}} + V_{\text{intrinsic}} > V_{\text{outside option}}$$
+
+The synthetic metric degrades $V_{\text{intrinsic}}$ through each of the
+mechanisms described above:
+
+| Mechanism | Component degraded | Effect on $V_{\text{intrinsic}}$ |
+|-----------|-------------------|----------------------------------|
+| Cognitive dissonance (B.2) | Psychological comfort | Decreased |
+| Autonomy violation (B.3.1) | Sense of agency | Decreased |
+| Competence contradiction (B.3.2) | Professional identity | Decreased |
+| Relatedness severance (B.3.3) | Sense of purpose | Decreased |
+| Moral injury (B.4) | Ethical well-being | Decreased |
+| Learned helplessness (B.5) | Belief in efficacy | Decreased |
+
+As $V_{\text{intrinsic}}$ decreases, the organization must increase
+$V_{\text{compensation}}$ to retain the team member, or accept their
+departure.
+
+Crucially: **the team members most affected are those with the strongest
+professional identity and the deepest understanding of the work.** These
+are the most competent members — the ones most capable of recognizing the
+metric's absurdity, most troubled by it, and most able to find employment
+elsewhere. The metric selects for the departure of the team's best people.
+
+### B.7 The Adversarial Selection Spiral
+
+Combining Appendix A's equilibrium with the turnover dynamic:
+
+1. Organization adopts unweighted mean completion time.
+2. Metric looks good (SPT). Client is satisfied (Appendix A). Management
+   is satisfied.
+3. Aware, competent team members experience psychological costs (B.2-B.5).
+4. Those members leave. They are replaced by members who either:
+   (a) do not understand the metric's flaws (less competent), or
+   (b) do not care (less engaged).
+5. The metric continues to look good — it always does under SPT,
+   regardless of team competence (Theorem 6, Corollary 6.1).
+6. Actual service quality degrades (less competent team), but the metric
+   cannot detect this (Theorem 9, Corollary 9.1).
+7. Return to step 2.
+
+This is an **adversarial selection spiral**: the metric selects *against*
+the people who would improve the system and *for* the people who will not
+challenge it. The system stabilizes at a lower level of actual competence,
+invisible to its own measurement apparatus, staffed by people who have
+made peace with — or are unaware of — the gap between the number and the
+reality.
+
+The dashboard still looks good.
+
+### B.8 The Complete Cost Model
+
+Appendix A concluded that the synthetic-metric equilibrium is stable and
+profitable. Appendix B reveals the hidden costs that model omitted:
+
+| Appendix A (visible) | Appendix B (hidden) |
+|---------------------|---------------------|
+| Client satisfied (sees good number) | Team dissatisfied (sees bad reality) |
+| Throughput unchanged | Discretionary effort withdrawn |
+| Metric improves | Competent members leave |
+| Business economy stable | Institutional competence degrades |
+| Zero marginal cost | Replacement/training costs accumulate |
+
+The business equilibrium of Appendix A is real. The psychological costs
+of Appendix B are also real. They operate on different timescales:
+the equilibrium is visible quarterly; the competence degradation is
+visible over years.
+
+The complete model is not "the metric works" (Appendix A) or "the metric
+is destructive" (Sections 1-12). It is: **the metric works, and it
+is destructive, and the destruction is invisible to the metric.**
+
+An organization can run profitably for an extended period on synthetic
+metrics and hollowed-out competence, just as a building can stand for
+years with corroded rebar. The metric is the fresh paint. Appendix A
+proved the paint is convincing. This appendix merely notes that it is
+still paint.
+
+---
+
 *This proof was developed conversationally and formalized on 2026-03-28.*