Question $2$

The INGSOC political party has expressed an interest in automating aspects of their state$-$wide

thoughtcrime detection apparatus. However, at present, the technology for the remote

monitoring and meaningful interpretation of brain activity is not mature enough for this

purpose. As an interim measure the party has commissioned an upgrade to its system of widely

deployed networked surveillance cameras to automatically identify any changes in body

language which are associated with ideological deviance.

Due to the high network usage of the Ministry of Truth's $($Minitrue$)$ propaganda broadcasts it

is not possible for the video footage to be streamed back to the Ministry of Love's $($Miniluv$)$

cloud for analysis. As such each camera needs to be realised as an autonomous agent and be

capable of making limited decisions on its own. The camera contains a kinematic model of

known deviant behaviours and must decide whether to raise the alarm or not based on the

degree to which a person's movements over time match the expected behaviours of potential

troublemakers. Due to computational constraints in the cameras the model isn't capable of

improvement directly although the firmware containing the model is frequently updated.

a$)$ Perform a PEAS analysis of an agent suitable for use in such a scenario. Be sure

to clearly indicate what each letter in the acronym means.

b$)$ Draw a diagram to illustrate a suitable agent architecture for this scenario