The Ministry of Love is interested in using reinforcement learning agents and serious gaming

to ensure ideological consistency amongst the nation's increasingly online youth. These agents

must interact by means of large conversation trees. The ministry wishes to train these agents in

acceptable conversational pathways through a punishment and reward$-$based system. Choices

in the tree can be grouped into three broad categories namely badthink $($i$.$e$.$ those inconsistent

with current party policy$),$ neutral, and goodthink $($i$.$e$.$ ideologically correct opinions$).$

ARTIFICIAL INTELLIGENCE

Describe how Q$-$Learning can be applied to this problem in terms of the following:

a$.$ How rewards and punishments could be modelled $(3$ marks$)$

b$.$ How expected values are updated $(5$ marks$)$

c$.$ The two modes of operation of such a system $(2$ marks$)$