A guide to developing bot personalities.

Illustration: Madeline Rawlings

Illustration: Madeline Rawlings

By Jess Thoms

Personality is the new UX.

Conversational interfaces have reduced user experience down to a few lines of text. With bots, UX becomes conversational, products talk back, and persona’s now go both ways. Every bot has a voice — which means every bot needs a personality.

If conversational computing means personality is the new user experience, how do we approach the design of these nuanced digital entities?

Why does your bot need a personality?

Chatbots and voice assistants are for humans. Conversational interfaces exist for better interactions between humans and computers. So then, how can we personalise these conversations to be more life-like, intimate, and representative of human interaction? Through personality. Building a rich and detailed personality makes your chatbot more relatable, believable, and relevant to your users.

Investing in personality informs every touch point of a chatbot. Personality creates a deeper understanding of the bot’s end goal, and how it will communicate through choice of language, mood, tone, and style. Seeing a bot as a lifeless piece of technology is a mistake. People project human traits onto everything — but now these objects talk back. Whether you like it or not, your users will still assign a personality to your bot if one hasn’t been explicitly designed.

“If you don’t spend the time crafting that character and motivation carefully, you run the risk of people projecting motivations, personality traits, and other qualities onto your App and brand that you may not want associated with them.” — Oren Jacob (Google I/O ’17)

Conversational experiences have to be personal. In order for brands to engage through bots and ultimately see conversions — they need quality conversations. Engagement and retention on conversational interfaces requires users to have an emotional connection to the experience.

Personality in action.

Branded personalities.

Brand stories can now be distilled into a digital experience, delivered in a conversation with an artificial personality designed to engage a specific target audience.

Chris Messina said 2016 would be the year of conversational commerce. I hope that 2017 and beyond will see the rise of personalized, conversational commerce from brands who value more meaningful, personal engagement through the experiences they devise.

Distilling a brand voice into a bot persona creates a myriad of personalized opportunities for commerce, gaming, and brands who seek to engage through personalization at scale.

This means a better user experience. When done well, of course.

Aiden Livingston, founder of Casting.AI and author of Build a Bot Workshop, believes chatbots need to be based on users’ personalities to be effective in selling to users.

He points to how teen clothing store employees tend to be clones of the people from their ads. It’s no coincidence — stores are choosing to mirror their target demographic for their customer interactions.

Why? Because their customers relate more to their peers and thus are more likely to buy from them. Could a crotchety 86-year-old man ring up pants and ask “the young people” if they found everything ok? Sure, after 86 years, Walter knows a thing or two about what makes for a fine pair of slacks.
However, youth-focused retailers have realized that if all their store clerks look like the cast from 1985’s classic film Cocoon, they are probably going to sell a lot less pants to teenagers.”

Mirroring a customer’s personality and talking to them the way they would interact with their friends is key to engagement. Design a character that is a true representation of the ideal customer to mimic human interactions.

“Brands need to seriously consider: How will customers talk to our brand? And, who are they talking to?” — Aiden Livingston

Does this bot have strong opinions? If it represents a brand, how does the bot personality align with the values and tone of the brand?

Who builds personalities?

The people who will drive the creation of these artificial personalities and subsequent copy creation need to understand the importance of personas. Additionally, this includes writers who see the value in microcopy and personas. Small pieces of text used to direct and inform users may seem meaningless — but when it comes to conversational interfaces — microcopy is all you have.

Skills to build a personality come from writers, designers, actors, comedians, playwrights, psychologists and novelists. The integration of these skills into tech roles have sprung terms such as conversation designer, persona developer, and AI interaction designer. Having these specific skill sets helps exponentially. Google is hiring creatives to bring humor and storytelling to human-to-machine interactions, and Microsoft Cortana’s writing teamincludes a poet, a novelist, a playwright, and a former tv writer.

“Authors and performers have been conjuring up convincing artificial personalities for millennia, so it makes sense that they’d be on the vanguard of designing this new kind of software interaction.” — John Pavlus

CONVRG co-founder Liz Snower offers that a thoughtfully planned and well executed personality is what saves a bot from delivering a clunky experience.

“Given the very minimal amount of copy that comprises a bot’s lexicon, conveying personality is not as simple a task as it may seem. This is why, if you’re a bot developer or agency, it’s imperative your team includes a writer who “gets” bots and personas. Just because there are only 10 lines of copy needed for any given conversation, don’t underestimate the importance of the bot’s words.”

Marketers craft brand narratives, creative writers plot a storyline, and copywriters make micro-copy magical. Conversation designers are a hybrid of all three. Don’t underestimate the importance of words.

Building a personality.

Target audience.

Diversity of perspectives, backgrounds, and skills is key in personality building — but you also need someone who embodies the target user to consult during the process. The feedback cycle is crucial to building personalities, and the nuances of language and culture are often missed by those on the outside. A 47-year-old male marketing manager is not the expected user of a teen beauty brand’s chatbot experience. Regardless of talent and input, the final say should come after testing with the target audience.

What’s your bot’s job?

Is your bot going to be a generalist or a specialist? Siri and Alexa are generalists — ready to answer a wide range of questions on just about anything. An all knowing black cylinder assistant, and a slightly dull voice from Apple whose full potential no one really understands.

Specialists are knowledgeable about one topic or industry — and go deep in that area. A mediation bot doesn’t know advanced algebra. Would you ask your yoga instructor to help with your math homework? Bots work effectively as specialists in one area, to help improve the lives and productivity of their users.

Understand your user — is the bot’s goal to reach a conversion quickly and effectively? Or is the whole bot experience crafted to engage long term as part of a larger creative campaign? Are they going to be a general assistant or a expert specialist?

Bot personalities need to be built in reverse — from the user goal and ‘job’ backwards. Instead of a human deciding which career suits their persona, a bot’s job gets chosen first. Once you have established your bot’s goal, and background, you can determine a personality type, and traits that will guide your dialog development.

“A customer support bot should empathize with people, even those who might be in the wrong. Just acknowledging and validating an emotion is often enough to make customers feel understood and release negativity, whereas being defensive or argumentative only exacerbates the problem.” — Adelyn Zhou, Head of Marketing, TOPBOTS

Mapping your personality out into a Bot Persona will help you translate a designed personality into a real ‘job’ use case. Austin Beer’s work focuses on human-centered design methods that can help people collaborate to create better bots. His Bot Persona Toolkit identifies how a user goal translates into a bot experience, and how the bot thinks, feels, and does as a result of the job at hand.

Austin describes that too often, people fixate on specific elements of a personality: hilarious, professional, robot-like. However, in reality, our personalities are incredibly multifaceted.

“What we need are tools that help us understand how all the different pieces of a personality work together and shift based on the situation. Moreover, we also need to help entire teams to agree and understand what is being made. Everyone should be able to point to a single poster and say, that’s the mission, now let’s get cooking.”

“That’s the mission, now let’s get cooking.” — Austin Beer

 

Personality types.

Myers-Briggs.

At Xandra we assign our bot personas a Myers–Briggs personality type. This works well as a basis to determining strengths, weaknesses, likely interests, dislikes, and their natural born abilities. These traits help define how a personality will react in certain situations, their tone of voice and body language, and whether someone displays introverted or extroverted behaviours. Translating this to bots it allows conversation designers guidance on what phrases they would use, whether they need emojis, or how to deal with harassment from users.

Amongst the wealth of material on Myers-Briggs, you will also find assignments to suitable careers and jobs per personality type. Matching this with the appropriate bot ‘job’ will confirm whether you’re building the right personality.

Spectrums.

Ari Zilnik, a New York based user experience designer and co-creator of Emoji Salad — an SMS-based chatbot game — uses ‘spectrums’ when defining a chatbot’s personality.

“I am by no means a psychologist, but there is a common model to describe personality across five personality traits. This model is called the Five Factor Model, and defines personality as a combination of openness to experience, conscientiousness, extraversion, agreeableness, and neuroticism. Ultimately, we as humans have many methods of understanding ourselves, and as personality designers, there are myriad models we can use to express or define a personality.
I find it useful to draw on personalities I am familiar with when I design a chatbot’s personality. What I end up doing is creating a few opposing spectrums of traits — for example I might create a scale with ‘dry humor/silly humor’, and another that is ‘enthusiastic/understated’. I typically have about 6–8 of these spectrums. I then create a list of about ten people — actors, friends, politicians, characters — and try to sort them across these spectrums. Finally, I look at the sorted lists, and I determine which character/person within that spectrum would best serve my user.”

This approach to personas might leave you with a personality that has “the enthusiasm of Aziz Ansari, paired with Obama’s dry sense of humor” Ari describes. This process allows for a more relatable example to refer to when designing dialogue.

Diversity and gender.

There are various approaches to how involved a bot will be with it’s gender and identity, and how much it will disclose about it’s persona to a user. When it comes to voice assistants, considerations around accents and male or female tone becomes part of a bot identity. In voice, tone is largely a preset feature, which means dialog design revolves around using this tone in the most effective way.

But do we even need gender? I question the relevance of gender in bots, and depending on how ‘human’ a bot is being designed — it may never be neccessary. When asked “what gender are you” Siri replies with “I exist beyond your human concept of gender”. Siri then prompts you to choose a different voice under settings. Moreover, designing personality traits based on gender is more problematic than useful.

Translating personality into dialog.

How do you pick the right words?

Every minute detail of your chatbot’s personality does not need to be programmed into its dialog. An author has the privilege of knowing the intimate details of their characters, however, these details are not explicitly written into the text. The summary of a character is not simply what shows up on a page. Your user doesn’t need to know these details, but you do. Choose the right words by asking yourself “would they really say that?”. Having a backstory to a character allows for personality development, and subsequently better dialog choices. Dialog informed by personality, compels your user towards the goal you ultimately want to achieve.

Tone of voice.

Without any effort placed on monitoring the development of a tone of voice, the copy gets messy. Is the bot delivering a professional service or a light hearted game? Tone of voice is directly influenced by personality development. Since the entire digital interaction is reduced down to a few lines of text — a friendly (or not so friendly) nudge in the right direction is the difference between conversion or frustration.

Tone of voice can easily be patronising. There is an art to delivering instructions.

It’s definitely not like this:

“Nice work in signing up to use this booking system!!!”
“Good job selecting an option.”
“Click the Next button to read the Next section.”

Be mindful of how personality informs the tone of each interaction. Whilst adhering to the required fallbacks and user prompts, think how these can be designed through the style of language and word choice. “Oh! We can fix that” evokes a different emotion from people than “Hmm… contact customer service to resolve this issue”. Be meaningful and deliberate with each word.

Testing bot conversation through real conversations.

Computer conversations are hard to predict — and your bot is likely accessible to a range of users despite your best efforts to push towards your target audience. Testing your conversations with a person (yes, a real life human) completely separate to your project will quickly point out any shortcomings.

“By having real life conversations rather than keeping our minds in our computers can help us better empathize with our users.” — Austin Beer

Conversation requires a shared language, context, and understanding of participants, which the wrong language choices can throw off course. Remember, bots are for humans. Simulate the bot personality and language choices in a conversation, notice how natural it feels, and how the other person is reacting to your word choice.

Through these real conversations you can easily find holes in your conversation flow, certain phrases that don’t translate effectively, and test the overall mood of your conversation. Building personalities requires iteration and testing to find the most natural fit for your bot use case.

“Ask questions, look at your data, and iterate accordingly.” — Liz Snower, Co-founder CONVRG

Resources.

PullString has a range of publications dedicated to writing convincing computer conversations with character.
PullString | Publications

Try this test from 16Personalities to find a Myers-Briggs type.
Free personality test | 16Personalities

Austin Beer is the most well versed person in bot personas that I’ve met. This is a great resource on designing intelligence.
How to Design Intelligence: 3 Ways to Make Human-Centered Bots


Many thanks to Ari Zilnik, Liz Snower, Adelyn Zhou, Austin Beer, and Aiden Livingston for being awesome humans and contributing their valued expertise to this article.


This post was originally published on Medium. 

Jess Thoms