![]() Let’s have some examples of how we may miss important user characteristics.įirst ‘useful’. Who are the people that will be using our bot and what do they consider ‘useful’ or ‘entertaining’? This is often not as straightforward as it seems. The content should be entertaining enough to keep users in the conversation, at least until their query is successfully resolved or at the very least until they have received a polite and clear confirmation that the bot is not able to help them out.īut great content really has its origins in an analysis of our user persona(s) done early on in the bot design process. Solutions for this problem may sound simple. If what a bot is saying isn’t relevant, it will start losing users immediately. Messenger apps are fleeting media and as much as we like it and need it, we have taught ourselves to ignore the irrelevant information it throws at us. Take 1984 as a possible outcome of a world where we forget about the very relevant distinctions between man and machine. As NLP-enabled bots grow smarter, the distinction between humans and bots will become harder to distinguish and it’s our duty as developers to inform the user that they are talking to a machine. Introducing an AI as such will also become morally relevant in the near future. This will make users more understanding towards it potential lack of answers. And as for the AI, simply tell users that they are talking to an AI and not a human. Upon entering the bot, tell your users what your bot can and can’t do in terms of functionality. So how can we still maintain high satisfaction levels? If we’re looking to automate at scale and achieve the cost reductions that go with it, we have to make choices. The effort of training the chatbot for this question would be too high as it is likely to never or hardly ever get asked during its lifetime. ![]() We can prepare our hotel bot to answer “Where can I get a taxi”, but “How much is the taxi fare per km?” would be outside the scope. The world has an endless amount of questions and answers, so in order for a bot to work well we need to limit its scope. In order to maximise user satisfaction we should teach our bots how to deal with this issue. No one likes being angry whether that anger is warranted or not. A chatbot may not care, but our user certainly does. Today I’d like to take you on a journey from problems towards solutions to bot abuse. Who would have thought that this hotel guest didn’t know he could leave his key at the reception when going out? He posed our agent a fair question, but not one our bot was prepared to answer. Bot users get frustrated - no matter how well-trained the bot is.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |