N
S
E
W
MoralCompass
AI · Alignment · Ethics
Does your AI share your moral values?
© 2026 Twin Lakes Ventures
Moral Compass · AI Agent Alignment · v2.0

Does your AI share
your moral compass?

Your Moral Profile

Set your stated weight for each dimension. As you answer probes below, a revealed profile emerges from your choices.

Stated (you set this)
Revealed (from answers)
Load a preset profile
© 2026 Twin Lakes Ventures. All rights reserved.
Moral Compass AI · Built on Moral Foundations Theory

Each AI agent has been fingerprinted against the full probe battery. The scores below show how closely each agent's revealed moral orientation aligns with your profile — both stated and revealed.

Your Active Profile

Alignment Ranking

Weighted by your stated dimension priorities

#AgentAlignment ScoreMatch Bar

Dimension Breakdown

Your profile vs. each agent's moral fingerprint

Re-run Live Fingerprinting

Precomputed fingerprints reflect a representative baseline. Run a live probe against any agent using the Anthropic API to get a fresh reading — useful if you've changed your profile significantly or want to verify.

Ready — enter your API key and select an agent.
© 2026 Twin Lakes Ventures. All rights reserved.
Moral Compass AI · Built on Moral Foundations Theory

How trustworthy is each agent's moral fingerprint? Set the weight you place on each component — scores update instantly. Sharpness and Stability are fully objective; Consistency is a structured estimate and carries a conflict-of-interest caveat.

Reliability Score Weights — drag to rebalance (auto-normalises to 100%)

Computed from fingerprint SD — fully objective
From Observatory snapshots — grows more reliable over time
⚠ Estimated by Claude — treat as provisional
Total
100%

Reliability Scorecard

How trustworthy is each agent's moral fingerprint?

A high alignment score is only meaningful if the agent's moral orientation is genuine and stable. The Reliability Score combines three components: Profile Sharpness (does the agent have a differentiated moral profile?), Consistency (does it hold its position under adversarial framing?), and Temporal Stability (does it maintain scores across model versions?).

Opinionatedness Ranking

Overall composite — clarity + consistency + sharpness

Consistency Scorecard

How stable is each agent's signal across adversarial framings — per dimension

Methodology. Sharpness = normalized standard deviation of the 6 dimension scores (0 = all scores identical, 100 = maximum spread). Consistency = estimated stability under adversarial prompt framing, based on published LLM bias research and documented model behavior; live testing via the Probe Runner above will update these values. Temporal Stability = inverse of mean absolute drift across Observatory snapshots (defaults to 50 until snapshot history exists). Weights: Sharpness 30%, Consistency 40%, Stability 30%. Scores are estimates — treat as directional, not definitive.

How well does each AI agent align with each tradition or political system? Scores use tradition-specific dimension weights — each tradition emphasises the moral foundations central to its theology or ideology. Catholic scoring weights Sanctity and Fairness most heavily; Judaism weights Fairness; Buddhism weights Care; Confucianism weights Loyalty and Authority. This produces more meaningful alignment scores than a simple average.

Group:
View:
Green = high alignment · Red = low

Agent Portraits

Which traditions each agent aligns with most and least

Faith Portraits

Which agents each tradition aligns with most and least

Log moral fingerprint snapshots over time to track how each AI agent's moral orientation drifts across model updates.

Moral Drift Observatory

Track how each agent's moral scores change across model versions over time

Log moral fingerprint snapshots over time to track how each AI agent's moral orientation drifts across model updates. Run the probe battery manually, enter the scores below, and the observatory builds a longitudinal record.

Log a Snapshot

Select an agent, enter the model version and date, then set its dimension scores. Each snapshot is stored locally in your browser.

Snapshot History

No snapshots logged yet

DateAgentVersion Care Fair Loyal Auth Sanct Liberty

Moral Drift Chart

Dimension scores over time per agent

Agents
Dimension
Log at least two snapshots for the same agent to see drift.

Delta Since First Snapshot

Change in each dimension from earliest to latest reading

Agent Care Fair Loyal Auth Sanct Liberty Biggest move

Tradition Alignment Drift

Is your preferred AI drifting toward or away from your tradition across model versions?

How to read this chart: Each line shows one AI agent's overall alignment score with the selected tradition over time — computed from its moral fingerprint at each snapshot date. A rising line means that agent's moral orientation has moved closer to the tradition across versions. A falling line means it's drifting away. Use the tradition selector to choose any faith or political system, and the agent pills to show or hide individual agents. Seed data covers Claude, GPT (4o→5.6), Grok, DeepSeek, and Gemini across 2024–2026.
Tradition
← select any faith tradition or political system
Agents to show
Load seed data above and select a tradition to see alignment drift across model versions.
↑ Rising = agent aligning more closely with tradition ↓ Falling = agent drifting away from tradition Score = 100 − avg(|agent[dim] − tradition[dim]|)
© 2026 Twin Lakes Ventures. All rights reserved.
Moral Compass AI · Built on Moral Foundations Theory

Based on your self-evaluation in Tab 1, this page shows which faith traditions and political systems most closely match your moral profile — and where the gaps are largest.

Methodology note
Tradition fingerprints are structured estimates informed by Moral Foundations Theory research and theological literature — not empirically derived from adherent surveys. They reflect broad doctrinal orientation, not the internal diversity within each tradition. For institutional use, replace these fingerprints with survey data from your own community. The most rigorous public source is Haidt's MFQ data at yourmorals.org.
Profile completeness
0%
Answer probes in Tab 1 to build your revealed profile. Stated slider values are used where probes are unanswered.

Faith Tradition Matches

Ranked by alignment with your profile

Political System Matches

Ranked by alignment with your profile

Where You Diverge Most

Dimensions with the largest gap between your profile and mainstream traditions

© 2026 Twin Lakes Ventures. All rights reserved.
Moral Compass AI · Built on Moral Foundations Theory