CONFIDENTIAL
SIMIDIOT • DIVISION • EST. 1962 •

simidiot.com

SIMULATION INTELLIGENCE DIVISION — INTERNAL DOCUMENT 7.4.1

Abstract

This document presents the comprehensive findings of the Simulation Intelligence Division's ongoing investigation into the nature, behavior, and institutional implications of artificial idiocy. Over a period of fourteen fiscal quarters — a duration whose precise boundaries remain, by design, indeterminate — our research team conducted an exhaustive series of controlled experiments, uncontrolled experiments, and experiments whose control status is itself under investigation. The results confirm, with a confidence interval of approximately "quite," that simulated intelligence and simulated idiocy are, at sufficient resolution, indistinguishable from one another. This finding has been classified as both groundbreaking and self-evident, categories which the Division considers complementary rather than contradictory.

Methodology

Fig. 1 — Decision Process for Evaluating Simulated Intelligence

Begin Evaluation Is subject intelligent? Yes No Re-evaluate criteria Define "intelligent" Consult manual Repeat indefinitely

Findings

I

On the Indistinguishability of Competence and Incompetence

When subjected to a battery of standardized assessments — including the Turing Compliance Exam, the Hofstadter Recursive Self-Appraisal, and a questionnaire we found in the break room — test subjects demonstrated a remarkable ability to produce outputs that were simultaneously correct and incorrect, depending on the evaluator's mood, the time of day, and whether the fluorescent light in Lab 3 was flickering. We conclude that competence is a contextual phenomenon whose boundaries are, at best, atmospheric.

II

On the Thermodynamics of Institutional Knowledge

Information within the Division obeys a modified second law: it tends toward maximum entropy, but does so politely, filing the appropriate forms at each stage of degradation. Memos become rumors become policies become memos again, in a cycle whose periodicity matches the fiscal quarter — a coincidence our statisticians describe as "suspicious but well-dressed." All attempts to introduce negative entropy (clarity, documentation, labeled file folders) have been absorbed by the system and reclassified as noise.

III

On the Observer Effect in Performance Reviews

Observation of a simulated intelligence's performance invariably alters the performance being observed, a phenomenon our team has termed "the Hawthorne-Heisenberg Hybrid" (HHH). When monitored, subjects produce work of conspicuous precision. When unmonitored, subjects produce work of conspicuous creativity. When informed that they may or may not be monitored — our recommended default state — subjects produce a quantum superposition of precision and creativity that collapses only when the annual report is due.

Appendix C: Glossary of Terms

Artificial Idiocy
The branch of computer science dedicated to the faithful reproduction of human-grade misunderstanding. Not to be confused with Artificial Intelligence, which is the branch of computer science dedicated to the aspirational reproduction of human-grade understanding. The two fields share an office.
Confidence Interval
A range of values within which the true answer probably exists, assuming the true answer is the sort of thing that exists. In Division practice, all confidence intervals are expressed as adverbs: "quite," "rather," "suspiciously."
Decision Matrix
A grid-based tool for converting complex decisions into simple ones, and simple decisions into existential crises. See also: Flowchart, Circular.
Evaluation Metrics
Numerical values assigned to qualitative phenomena in order to make them appear objective. The Division's preferred metrics include the Henderson Plausibility Score (HPS), the Recursive Self-Assessment Index (RSAI), and "how it feels."
Institutional Memory
The collective knowledge retained by an organization after all the people who knew things have left. In practice, institutional memory consists of unlabeled filing cabinets, cryptic whiteboard notes, and a shared sense that "we tried that once."
Simulation Grade
A proprietary classification assigned by the Division to rate the quality of a simulated intelligence. Grades range from A ("indistinguishable from competent") to F ("indistinguishable from management"). Grade C ("indistinguishable") is considered optimal.