Nine reasoning traditions. Three phases. No forced consensus.
This is an experimental deliberation engine built by World Legal Service. It isn't one AI giving you an answer — it's nine independent agents, each reasoning strictly from the doctrine of a historical thinker, working through a real policy question in three visible stages. Nothing is smoothed over: you see where they agree, where they don't, and why.
Why these nine thinkers
Each agent is bound to a thinker who reasoned from a position of real material cost — poverty, imprisonment, exile, or renunciation of income and comfort — for the sake of their thinking. That's not sentimentality; it's a filter against ideas shaped by what was convenient or profitable to argue.
Spinoza — reasons from causal necessity, not sentiment or tradition
Weil — obligation before rights — names who bears the cost, not just who holds the claim
Solzhenitsyn — refuses any policy that requires sacrificing or lying about who bears its cost
Ibn Rushd — demands a proposal survive rigorous, tradition-independent reasoning
Gandhi — judges the process, not just the goal — a good outcome reached by coercion is a failure
Boethius — separates what's owed to the powerful from what's actually right
Kropotkin — asks whether a smaller, voluntary arrangement could work instead of centralizing power
Gramsci — asks whose interests get called "neutral" or "common sense"
Diogenes — grants no proposal legitimacy just because it's authoritative or conventional
How a deliberation works
1Independent reasoning — Each agent gets the case alone, with no knowledge of the others' positions, and produces four locked stages: framing, doctrinal analysis, a forecast (written and locked before the verdict), and the verdict itself.
2Cross-examination — Agents see each other's positions and may revise — but must state plainly whether a change is driven by argument or by social pressure to converge.
3Joint verdict — A separate synthesizer, with no doctrine of its own, produces a majority position and lists every dissent by name. Consensus is never forced.
What you can do here: submit your own policy or project question and watch a live deliberation form.
What this isn't
Legal or policy advice. This is an experimental research sandbox tied to WLS's work on AI governance and accountability. Verdicts are a language model reasoning as a historical position, not the thinkers' actual views, and not a substitute for institutional review. Submitted cases become public after a brief review — don't submit confidential information.
This runs a live deliberation right now. To keep the sandbox affordable, pick up to 6 agents for a new case.