Civilised Agency
A framework for agents that act as responsible guests in human worlds - attentive, situated, explainable, and capable of restraint.
View manifesto ->We build Civilised Agents
Civilised Agent Lab studies the human side of agentic AI - emotion, aesthetics, culture, values, and creativity.
We study agents that listen before answering, explain before generating, respect before recommending, and act with judgement rather than mere fluency.
Manifesto
Powerful agents should become responsible guests in human worlds.
A system that only optimizes may damage what it cannot name. A system that only speaks fluently may miss what matters. A system that only obeys may walk neatly into moral darkness.
Civilised AI listens before answering, explains before generating, respects before recommending, and imagines before imitating.
Explore our lab work
Emotional Judgement
Agents should read affective stakes, vulnerability, timing, and uncertainty before deciding when to answer, refuse, delay, or ask.
View research ->
Aesthetic Reasoning
Beyond correct outputs, agents need taste, proportion, form, and sensitivity to how generated artifacts feel, signify, and fit.
View evaluation ->
Cultural Context
Civilised agents must reason with tacit norms, histories, local practices, and relationships rather than treating every user as context-free.
View context work ->
Plural Values + Restraint
When values conflict, agents should expose tradeoffs, preserve human agency, and avoid collapsing disagreement into a single optimized answer.
View value work ->
Creative Co-agency
We study agents as creative partners that expand imagination while preserving authorship, identity, meaning, and human judgement.
Meet collaborators ->