Decision-theoretic sequential decision making (SDM) is concerned with endowing an intelligent agent with the capability to choose the 'best' actions – those that that optimise the agent's task performance. SDM techniques have the potential to revolutionise many aspects of society, and recent successes have sparked renewed interest in this field, such as agents that learn to play Atari games and which can beat master Go players.
Despite these successes, fundamental problems of scalability prevent SDM methods from addressing other problems with hundreds or thousands of state variables. INFLUENCE seeks to overcome this barrier, by developing a new class of ‘influence based SDM methods’ that address scalability issues by using novel ways of abstraction. This will be an important step towards realising the promise of autonomous agent technology.