Designing the personality of a bot
Given that personal assistants are here to stay, it might be a good time to stop and think about how to design the personality of a bot, making them credible in their relationships with users.
We must not forget that a bot will participate in conversations with users who, through speech interfaces, will want them to be as human as possible. To achieve this, we need to give our bot a personality so that conversations are coherent and consistent, as well as responsive to the needs of our users.
This is important: do users demand it, and will it bring any benefit to our experience?
These are the first questions you should ask yourself when you start designing a bot.
Who is the target audience?
Targeting young users is not the same as targeting older users, just as designing a bot for an online shopping application is not the same as creating a bot for an online shopping application or a medical application.
It is crucial that we obtain the most reliable data possible from our audience, in order to be able to use this data to design the personality. The age range or the language they use are essential to get the personality of our bot correctly.
What role will the bot have?
Bots such as Alexa or Siri are generalists and therefore have to answer a wide variety of questions, something that can be clearly seen in the TV commercials in which they are featured.
This is a huge technical challenge and we advise you to start small and work your way up. As with any good product, it is enough that the bot does one thing and does it well, so turning it into a specialist will allow us to simplify not only the technical part of its implementation, but also its language and personality.
Knowing what the bot’s ultimate purpose is will allow you to assign it a personality in an easier and more efficient way. Customer service? Your bot should be empathetic with users in order to solve their problems and avoid frustration. Sales? Your bot can be more enthusiastic or even curious in order to find the best product for your customer.
Methodologies for designing the personality of a bot
Based on the above, we can consider two possible purposes for our bot:
- Customer service: it will be empathetic with users to solve the problems that affect them. In this way, it will be able to lead users to a solution, even if they are wrong at some point. Important tip: never, under any circumstances, put your bot in a defensive position; that will only increase the problem, as well as increase users’ frustrations.
- Motivational bot: it will guide users in the objectives they have set, helping and motivating them to achieve their goals. It is more of a coach than a helper, its personality will help them to achieve actions and help them to overcome possible frustrations.
Interactions will fit one or the other profile depending on many criteria, such as the moment the user is in, like onboarding. However, your bot should have a fixed personality. Let’s take a look at different options with which we can design it.
The bot persona
This methodology by Austin Beer adapts the creation of personas to the personalities of the bots and has a very interesting starting point: the need to empathize not only with the user but with the bot as well, putting team members in front of each other to create a conversation in which some adopt the role of the user and others the role of the bot.
The parts of this diagram include:
- Bot definition: the name and objectives of the bot, which are placed in the central part.
- The user’s objectives: separating them from those of the bot helps to better understand the role it will play with the user and how the two can be aligned.
- On the right side, there are the motivations of the bot: what it thinks and what it feels.
- Finally, on the left side, we will find its actions (Do), in other words, what it does.
How many features should be included in each part of Think, Feel and Do? Austin Beer recommends no more than three in each, allowing the teams participating in the session to concentrate on truly meaningful aspects, without losing focus on unimportant issues at the moment.
The Myers-Briggs types
With this method we can create personalities based on behaviors in response to certain circumstances, setting the bot’s personality according to the actions it has to perform at any given moment.
We can define up to 16 different personality types, based on the 4 dichotomies used: extraversion/introversion, sensing/intuition, thinking/feeling and judging/perceiving.
As it is based on opposites, certain features may end up being strange when they are implemented in practice. Despite this, it is one of the most widely used methodologies, due to its wide use in human resources and the ease of its implementation.
The Big Five model or personality spectrums
This methodology, common in psychology, allows the creation of personalities using five major factors:
- Openness to experience
- Conscientiousness
- Extraversion
- Agreeableness
- Neuroticism
These traits are graduated on a scale, with which we can create infinite personalities with multiple traits depending on the type of user, the moment of the experience or evolution.
Ari Zilnik defined the personality of the Emoji Salad chatbot using this methodology. To arrive at the final result, he assigned to each of these traits characteristics of famous characters (real or fictional), resulting in a unique personality with multiple dimensions, exactly as it happens to any of us.
Although it may be somewhat more challenging, the multitude of possibilities that this methodology opens up will allow us to create a more complex and adaptable chatbot, in other words, a more human one.
If we add the possibilities that open up to us thanks to the application of learning models and artificial intelligence, we will have a personality that evolves and matures, exactly as happens with humans.
Side note: addressing diversity and gender
Although for voice assistants the choice of a male or female voice may be a distinguishing feature (Alexa and Siri are clearly identified with the female gender), in a chatbot gender is not so decisive.
In fact, it is quite advisable for chatbots to be just that, robots. Avoiding racial or gender traits is more than advisable, since users know they are interacting with a machine. And if necessary, you can always interact with a human. This will also avoid the appearance of biases that can affect both the design and the user experience. Even though there may be business reasons that may lead us to believe otherwise.
Communicating with users
Once the personality of the bot has been defined using some of the techniques seen above, it is time to start talking to users.
And, just like any other part of your company or brand, it must use the same tone of voice as your brand. If your brand is serious, you can’t give your bot exaggerated humorous features, because it may annoy users, especially if they are looking for support.
Each word here should be carefully selected to help users achieve their objectives:
“I pass your request to a human operator because it is beyond my capabilities. If you want to add something, write it down and I will send it to them” or “Contact support as I can’t help you”.
What is the best way to address the user? While both are valid, the first is much more empathetic and avoids the frustration of taking an action.
Choose your words wisely, because a lot will depend on them, especially in areas such as sales or support.
And remember: research, design, test, review data and improve.
This article is a translation of the following one published on our corporate website:
