Human agents are capable of giving advice to other humans through the use of their natural intelligence. Some ‘AI assistants’, such as Alexa, are examples of Intelligent Agents (IA) as they use sensors to receive requests and then accomplish their goals by returning results. These IAs are also capable of providing advice to a human, which results in the question: Are humans more likely to accept or refuse an IA’s advice in comparison to a human agent?

The aim of this project is to identify what physical and interactive attributes are crucial for an IA to possess in order to be considered trustworthy for providing behaviour change advice. Identifying and cataloguing the impact of these attributes will allow developers to implement IA systems that are deemed trustworthy by a user. This study involved the development of an IA, with variable attributes, for a week-long study with six participants who would be given advice by the IA. The verdict of the advice (if the advice was followed) would then be used to determine if the IA was deemed trustworthy.

The results of the study concluded that participants would prefer to receive advice twice per day and would prefer the gender of an IA to be neutral. Participants also preferred to receive a varied amount of advice and receiving the same advice everyday caused participants feel unlikely to follow the advice given. There was some evidence that receiving the same advice everyday caused participants to be more likely to question the Analysis method of the IA. However, further testing on this should be done to corroborate it.