New AI Framework Lets Machines Negotiate Like Humans—With Clear Rules
Researchers have created a formal system that allows AI agents to argue strategically with each other, complete with measurable outcomes and predictable stopping points. The work matters because it could underpin more trustworthy negotiation systems, from automated contract talks to policy deliberation platforms.
Originaltitel: Equilibria in quantitative bipolar argumentation dialogues
<p>We introduce MQBAFs, multi-agent extensions of quantitative bipolar argumentation frameworks (QBAFs). In MQBAFs, agents have objectives to maximise or minimise the strengths of some arguments, and establish preferences over these objectives. Agents make utterances by adding or removing arguments or attacks to/from a QBAF, or by changing arguments' initial strengths. We then define equilibria for MQBAFs, in which no rational agent would make any additional utterance. The notion of MQBAF opens quantitative bipolar argumentation to principled strategic analysis.</p>