Reasonable Machines: A Research Manifesto

Transparency alone isn't enough: The paper argues that simply demanding AI systems be "transparent" falls short for autonomous systems making moral and legal decisions (like self-driving cars or healthcare applications). Hidden decision-making processes require more than just visibility—they need systems that can actively explain and justify their reasoning.

Machines need to learn moral reasoning: Rather than pre-programming ethical rules, the paper proposes that intelligent autonomous systems should be designed to autonomously justify their actions by "learning to give and take reasons" for decisions, similar to how humans engage in moral reasoning and dialogue.

Trust requires two-way communication: Building genuine trust with AI systems requires "normative communication"—machines that can not only explain their decisions but also engage in reasoning about ethical and legal questions, creating a more meaningful human-machine interaction than current one-directional explanations.

New technical tools are necessary: The manifesto calls for developing hybrid reasoning approaches, ethico-legal ontologies (structured frameworks for ethical and legal concepts), and argumentation technology—suggesting that existing AI methods need fundamental redesign to handle moral and legal decision-making responsibly.

Governance frameworks must evolve: The paper implies that current governance approaches focused on accountability and transparency need to be supplemented with technical systems that enable machines to participate in normative reasoning, creating a more robust "ecosystem of trust."