There are probably a number of examples of chat bots / virtual assistants being used in government websites such as Alex at ATO
Just wondering if there are guidelines being used to develop these assets, in terms of:
the types of customer/client/user interactions designed to deal with
the ‘persona’ or ‘character’ that the chat bot or virtual assistant embodies
the nature of the ‘conversation’ or interaction that it provides
Unfortunately, I don’t think there is a simple guideline due to the breadth of options and what you’re hoping the CUI to do. I know that we’ve been looking at the Microsoft conversational style guides to assist with the way are looking at our interactions with users and looking at a ‘dumb’ chatbot, where it is essentially a Q/A style conversation due to privacy and legal reasons, we don’t want provide bad advice (if we went with a proper AI style CUI)
Maybe it is not so much a guideline on what to do (since it would, as you say, be rather context sensitive), but I would say that a ‘what definitely not to do’ checklist would be valuable (e.g. importance of considering uncanny valley in both visual and behavioural characteristics).
I know there has been some success in setting the expectations from the get go (i.e. in the introduction/greetings “Hi i’m a digital assistant here to help you XYZ”. Some areas have also dabbled in certain trigger words that immediately results in an X response (i.e. could be around mental health issues). I would definitely match the conversation with your style (e.g. we’re not having a ‘overly friendly’ one due to the legal terminology/complexity of what we are answering but due to the overwhelming number of chatbots providers and users, what is bad for one group is good for another - there could be people wanting users to think their chatbot is real! I think as long as you set those expectations, the can/vs can’t say is up to you (within legal reason haha)
