The Athenian agora was at once a marketplace, a bargaining floor and a voting assembly. This one revisits the core problems of my Intelligent Agents MSc module (automated negotiation, auctions, voting and distributed optimisation) as interactive demonstrations. A real Python engine computes every result, and five of the six demos run that engine live in your browser via Pyodide.
Two agents haggle over a job offer under a deadline. Watch the bidding dance against the Pareto frontier, the concession curves cross and a frequency model learn the opponent's hidden priorities round by round.
English, Dutch, first-price and Vickrey auctions: bid shading, who wins and the revenue-equivalence theorem converging live, the mechanism-design route to allocating scarce supply.
The same ballots, different rules, different winners: plurality, Borda, instant-runoff and the Condorcet cycle that breaks them all.
DCOP: independent agents coordinate to optimise a shared objective under constraints, the multi-agent backbone of resource-allocation and scheduling engines.
Logistics supply as a transportation problem, solved three ways (a central LP optimum, a decentralised market where demand bids for capacity and a greedy baseline) and measured on the efficiency gap. The market reaches the optimum and rediscovers its prices by bidding.
Two-sided markets with no prices at all: candidates and companies (or residents and hospitals with quotas) pair off by Gale–Shapley deferred acceptance. Tentative offers, rejection cascades, and a stable match no pair wants to leave; whoever proposes gets the best stable deal going.