What manners should you teach your chatbot?


By Jess Thoms. 

If we assume that eventually bots will replicate human interactions so effectively that a user cannot tell the difference — then bots need manners. So, what etiquette do bots need to know? Since it’s our job to teach them how to behave, what principles do we need to follow?

Even mundane, menu driven bots need basic etiquette. If a bot is an extension of a brand voice, then any negative experiences directly reflect the brand. As a general rule, bots should be polite, helpful, and efficient.


As simple as it sounds, make sure a bot always replies to hello, goodbye, and general greetings. Even past the initial formalities a user can be expected to use these phrases again,and this should not trigger the bot’s response to a first ‘hello’. How do you greet someone you just met compared to a friend you’ve known for years? After the first set of greetings, the user moves into a more personal context with the bot. Remember, it’s the job of the human to teach bots how to interact like humans. Formal introductions are standard, but as you move further into a conversation more informal gestures should be explored.

General rules.

How would you expect a polite human being to interact with you? Please, thank you, and excuse me are all simplicities that should not be overlooked. Most bots are here to help. It’s fair to say helpful mannerisms should be standard.

Bots may have the ability to pull data from a user’s profile, but exposing this in a conversation without explicitly asking for permission is invading someone’s privacy — and it’s creepy. When meeting someone for the first time, you wouldn’t bring up information only accessed by combing through their Facebook profile the night before. Don’t be that person. Or bot.

Also — be courteous. Would you call your mom at three in the morning to say hello? Well, be aware of user’s time zones if you are using broadcast messages.

Onboard effectively.

Setting up the context for a conversation is critical, and will set the tone for an entire interaction. Clarify what the user should expect to give and receive in the conversation, the purpose and limitations of the bot, and the end goal for the bot experience. To an extent, on-boarding should be stripped of personality if it will help set up an effective conversation flow. Of course, personalization is crucial. But if the bot is too casual in it’s personality or vague in the on-boarding flow — then you are setting users up for frustration.

Don’t ignore.

Problematic content should be addressed. You can expect a myriad of random, and at times inappropriate, content input from users. Instead of ignoring this content — address it. Depending on your bot’s context and purpose, it may be appropriate to call the user out on their actions. Otherwise, acknowledge the user said something irrelevant, and direct the conversation back on track. Ignoring this behaviour will just encourage inappropriate content in an attempt to ‘break’ the bot by a user.


At the very least disclose that your bot is not quite human. You can subtly expose a bot without making it completely depleted of humanity. This also plays a role in the naming of a bot. At this stage of the bot ecosystem, users are not aware of signs that they are talking to an intelligent assistant rather that a human.

“Hi, my name is StudentBot. I can help you better manage your timetable this semester.”

is clearer than

“Hey! It’s James, I can help manage your timetable better.”

Deception may soon become a problem if bots are being designed in a way that they could pass as a human. Using inbuilt features like buttons, cards, and quick replies will also hint to the user that they are not speaking to a person.

“Sorry, I don’t understand.”

To make interactions more personal and human, bots need to employ empathy, apologies, and offer solutions. Fallbacks that only apologize don’t achieve anything. In reality, to drive a conversation forward you would rephrase what you meant, and check to see if your counterpart now understood. Ultimately, if a user is confused a bot needs to step back and explain it’s purpose, limitations, and ask how it can help.

Apologies matter. If a user is pushing the boundaries of a bot interaction, and is becoming frustrated, apologize. Take the high road. Think about the tone of the interaction — make the user feel like they are heard and always offer a solution.

The challenge of designing human-like bots is that they will need to be created for a diverse range of contexts. However, retaining basic humanity, etiquette and manners will help set a standard of conversation no matter who you’re talking to.

This article was originally published on Medium.