Machine learning for control of a personal video recorder ("PVR"), including providing records of transmissions of messages; identifying a message point in operation of a PVR; selecting, in dependence upon the message point, a message to be transmitted; and determining in dependence upon the records of transmissions of messages whether to transmit the selected message. Embodiments include determining whether to transmit the selected message in dependence upon a frequency with which the message is transmitted to a user. Embodiments include transmitting to users messages prompting for responses identifying commands; storing records of transmissions of messages, including storing records of responses; and selecting a message to be transmitted in dependence upon the records of responses to messages. Embodiments include structuring the selected message in dependence upon the records of responses to messages, including inserting in the selected message choices for a set of sequenced commands.