Not all chatbot conversations are same. So, it is crucial to train chatbots in a way that respects the users' sentiments. When it comes to conversations, there are three distinct sentiments that we feel. There are the conversations we love to have such as when we're ordering a delicious pizza for that big game, and conversations we hate to have like when we're calling for tech support. Our sentiment towards conversations shapes our expectations and tolerances.
You're going to be much more receptive to humor when ordering a pizza versus addressing a fraudulent credit card charge. You need to identify what conversation type your chatbot is dealing with and shape the user experience (UX) to fit the sentiments of the user best.
If your chatbot is addressing conversations that users hate to have then you need to give careful consideration to its personality and tone of voice. Users dealing with these conversations are going to have short fuses and less tolerance for unnecessary jokes. Examples of these types of conversations could include tech support, fraudulent bank charges, or even paying bills.
In these cases, it is probably better to err on the side of caution and design your chatbot with a blander personality that is more direct and professional in tone. The last thing you want is to prolong an already undesirable conversation by inserting humor the user doesn't want in the first place. Getting the user to the resolution they're seeking, as quickly as possible, should be the primary goal of chatbots built to handle these hated conversations.
When your chatbot is fortunate enough to handle conversations that users love to have, you should be drooling at the mouth for the opportunity to craft a well-thought-out chatbot personality. Users engaging in these conversations are going to be far more receptive to unique personalities and some witty banter from the chatbots. In fact, failing to capitalize on this opportunity to give your chatbot a strong personality could be the reason you lose out to the competition.
While you have a lot more leeway when it comes to your chatbot's UX with these lovable conversations, you should also be wary about venturing too far into left field. For example, if your chatbot is supposed to recommend wine and dinner pairings, then you may not want to give it the personality of a mad scientist. Instead, have the chatbot act like an actual wine sommelier would at a vineyard. Personable, knowledgeable, and with a little bit of wine-appropriate humor mixed in. Your users will be more apt to take your chatbot's recommendations seriously if it emulates the typical real-life experience they're accustomed to.
This is probably the biggest challenge when determining how to mold your chatbot's personality and tone of voice. Unfortunately, it's going to be far more apparent when dealing with a conversation that is on one of the far ends of the love/hate spectrum. Examples of conversations in the middle of this spectrum include basic account inquiries, checking flight statuses, gathering sales information, or even finding out how and when to contact a business.
In these conversation scenarios, you should still err on the side of caution when in doubt. Because users aren't exactly thrilled to have these types of conversations they may still be prone to impatience and have a low tolerance for inappropriate humor. Ultimately, users are looking to get some sort of value from your chatbot, and you should aim to provide it as simply and quickly as possible.
If you can find opportunities to mix in some personality and joking around without disrupting your value proposition and natural conversation flow then, by all means, go for it and take advantage. Your chatbot may even cause a user to learn to love one of these conversations if you're lucky enough.