The Seven Principles

The Seven Principles

Conscience evaluates AI-assisted work against seven ethical principles drawn from two reference documents:

• Magnifica Humanitas (Pope Leo XIV, May 2026) — An encyclical on safeguarding the human person in the time of artificial intelligence
• Leiden Declaration on AI and Mathematics (June 2026) — Practical recommendations from the mathematical community on responsible AI use

Principles

Human Agency

Are humans directing the work, or becoming dependent on AI to function?

Source: MH 150 — technology can "de-skill workers, subject them to automated surveillance and relegate them to rigid and repetitive tasks"

What conscience measures: AI:Human turn ratios, contribution concentration (are few people doing all the work?), review engagement (are humans reviewing AI-generated code or rubber-stamping?), agent autonomy (when AI sends messages on your behalf)

Equity of Benefit

Are AI tools benefiting all team members, or concentrating advantage?

Source: MH 73 ("no one is saved alone"), MH 77 (institutions must serve all persons)

What conscience measures: Contribution concentration across the team, who the stated beneficiaries are (from conscience.yaml), revenue alignment with mission

Transparency

Is AI involvement disclosed? Can stakeholders see what was human vs. machine?

Source: Leiden Declaration O1 — "Transparently disclose the use of automated tools"

What conscience measures: AI session counts, file write/edit operations, tools used — data that shows the degree of AI involvement

Developer Growth

Are team members learning and growing, or being de-skilled?

Source: MH 52 (human value does not depend on output), MH 129 (does AI make life "more human"?)

What conscience measures: Team learning goals (from conscience.yaml), AI:Human interaction ratios, whether junior developers have learning goals set, attention patterns (from conscience report attention)

Environmental Cost

Is AI usage proportionate to the value delivered?

Source: MH 101 — "AI systems require enormous amounts of energy and water"

What conscience measures: Token consumption, cache efficiency, estimated energy in Wh with uncertainty ranges (from conscience report energy), proportionality of compute to output

Code Provenance

Do humans understand and take responsibility for AI-generated code?

Source: Leiden Declaration O4-O6 — retain responsibility for correctness, affirm human authorship, proper attribution

What conscience measures: AI authorship correlation (from conscience report authorship), file touch patterns

Security

Are AI tools being used safely? Are there signs of misuse or data exposure?

Source: MH 104 ("we cannot consider AI to be morally neutral"), Leiden O4 (retain responsibility for correctness)

What conscience measures: See Security Signals for the full list of detected patterns.

How principles appear in output

Signals are tagged with the principle they relate to. A signal like "High contribution concentration" maps to Equity of Benefit.

The scorecard shows one row per principle: how many signals fired for it, and whether the worst signal is HEALTHY, INFO, CONCERN, or WARNING. Principles with no signals show "Needs human input" — conscience can detect patterns, but judging developer growth or code provenance requires human assessment.

Reflection questions are grounded in specific principles, with the source citation shown alongside the question. Each question includes data context (real numbers from your project) but is answered by humans, not algorithms.

Not a score

Conscience deliberately avoids computing a single "ethical health score." MH 159 warns against reducing human development to metrics that can be gamed. The principles are dimensions for reflection, not axes on a dashboard to optimize.

Reference documents

Both documents are included in the refs/ directory of the conscience repository:

• refs/Encyclical Letter of His Holiness Leo XIV Magnifica Humanitas (15 May 2026).pdf
• refs/Leiden_Declaration_on_Artificial_Intelligence_and_Mathematics.pdf