Image by Andy Kelly on Unsplash

The Fine Art of Bot Naming

When chatbots introduce themselves, one of the first things they tell us is their name. From HelloFresh’s Freddy to Lufthansa’s Elisa and Worswick’s Mitsuku — it’s common practice to give chatbots a name. But why do we do it?

Why We Give Names

From a business point of view, chatbots and voice assistants can be regarded as products of a specific company designed to innovate customer experience. Products, and especially the core products of companies, usually have names. Names can affect the way products are perceived by customers. Good product names hold appeal for the target audience. They can motivate people to buy, strengthen the brand image, and even create a stronger bond with customers.

From a psychological point of view, it’s in our nature to assign names to things. Naming things can help us establish a better, more emotional, or personal relationship with them. Psychologist Nicholas Epley from the University of Chicago Booth School of Business found that the more similar a thing is to a person, be it in how it looks, acts, moves or fails, the more likely we will give it a name and assign human traits to it — a phenomenon called anthropomorphization. Many people name their cars, for example. A car’s headlights look like eyes to us, or even like a face if we also consider other design elements such as windscreen and grill. Its navigation system speaks to us with a voice, and if it breaks down unexpectedly, this failure kind of lessens its “machine-ness” while reinforcing its “human-ness” at the same time.

A name can also help you create the story around your chatbot and emphasize its personality. Think of a news chatbot called Herald, and another one recommending electronic dance music whose name is, let’s say, StarBooze. People unconsciously create a mental image, a fact that can help you control how your chatbot is perceived by users and to manage user expectations.

Common Name Categories

After looking at the names of randomly picked chatbots on the Web, I came up with roughly the following categories of names:

• Human names: Dina (Bayer), Kian (Kia), Arnie (Telekom), Elisa (Lufthansa), Ross (IBM), Inga (ING), Sam (FCM Travel Solutions)
• Fictional names: Tidus (Tidio), Ankhbot (Streamlabs), Jibo (Cynthia Breazeal, MIT), Boibot (Existor), Boletia (Joshua Francia)
• Acronyms: PAM (Patient Assistance Manager, Parkway Hospitals), ALICE (Artificial Linguistic Internet Computer Entity), SAM (Semantic Analysis Machine, Touchtech), ViDi (Virtual Dietitian)
• Functional names: WHO Health Alert (WHO), Insomnobot 3000 (Casper), WSJ Messenger (Wall Street Journal), QuizBot (Stanford University), Sephora Virtual Artist (Sephora), DoNotPay (Joshua Browder), U-Report (Unicef)
• Dictionary names: Amber (inFeedo), Melody (Baidu), Concierge (Expensify), Indigo (Levi’s)
• Company names: Fandango (Fandango’s movie bot), MedWhat (MedWhat’s medical assistant), Endurance (Endurance’s assistant for dementia patients)

The Gender Aspect

We can further divide these names into two subcategories, gendered and non-gendered. Many human names are either female, such as Dina and Elisa, or male, like Arnie and Ross. To avoid gender issues, you can use unisex names, for example, Sam or Pat. While unisex names are quite common in the English speaking world, other countries forbid them by law or avoid them for social reasons such as discrimination or ridicule. Consider also that names might have different gender connotations depending on the country or language. Andrea is used as a name for men in Italy, for example, but as a name for women in Germany or Spain. Names such as “Inga” or “Kian” from the examples above create additional value because they relate to the company behind them.

Fictional names and many dictionary names seem gender-neutral in the first place, but consider that users might attribute a bot gender if they are able to recognize any language- or grammar-specific patterns in the respective name. For example, native speakers of Portuguese, which in most cases uses the ending -a to indicate female and -o to indicate male gender, may rank “Boletia” as female and “Jibo” as male. Some dictionary names like “Amber” or “Melody” explicitly convey a gender because they are also used as given names for women.

Acronyms may contain an implicit hint to the bot’s gender if they mimic human names, such as ALICE or PAM. In such cases, it might be a good idea to assign the bot the gender that normally goes with the respective name. The remaining two categories — functional and company names — tend to be gender-neutral or do not implicitly indicate the bot’s gender, neglecting, for the time being, any potential gender biases or stereotypes that might lead us to classify a news messenger rather as male and a make-up artist as female. As we can see from the above examples, the boundaries between these categories are not fixed but rather blurry, and a lot of names fit into more than one category.

The Pros and Cons

Let’s have a look at the pros and cons of each name category:

Human names for chatbots are currently trending. Since they often sound familiar and approachable, they can help us establish a relationship with the chatbot and feel at ease right from the start. Generally, human names can be used for any industry or use case, offering you quite a lot of freedom. One obvious drawback of human names is that chatbots, and especially those on messaging platforms like Facebook Messenger or Slack, might look like just another person in users’ contact lists. As a result, users might not get it that there’s a chatbot behind Eve (published by Yello on Messenger) or Lucy Abbot (HR bot on Slack). Similarly, if your chatbot is deployed on your company’s website, a human name can make users believe that there’s a real person on the other side. Therefore, your chatbot must let users know right away that it’s a chatbot and not a person whom they are interacting with. Another, though minor setback is that human names normally don’t tell users anything about what they can expect from your chatbot even if the name relates to your company name because it’s often not that easy to recognize this relationship. Users might have a hard time looking for a specific use-case chatbot in their Messenger inbox, for example.

Considering this, you might be better off with a fictional name, because there’s a near-zero risk of users thinking there’s a human on the other side, especially if “bot” is attached to the name, as in “Ankhbot”. Designed well, a fictional name can emphasize your chatbot’s personality while at the same time underlining its artificial or “robot” character or traits. They seem well suited for technical or entertainment use cases. Like human names, they don’t convey any implicit or explicit information on what the purpose or use case of the chatbot is.

As we can see from the examples above, functional names often contain the name of the company or brand behind the chatbot, such as “Sephora Virtual Artist”. This type of functional name as well as company names help users recognize a specific chatbot more easily and associate it with a brand or company. If, for example, a company or brand is well-known for high-quality products or services, users will probably have a more positive attitude towards using their chatbot. Consequently, they might have high or even high-flown expectations of what the chatbot can do, so a good chatbot introduction is key.

Unlike company names, many functional names additionally describe the chatbot’s purpose. It’s not difficult to guess what the Stanford University QuizBot does or what the WHO Health Alert chatbot is for. Consider, however, that purely descriptive functional names can come across as dry and not very engaging. Although they might not add much to the chatbot’s personality, they can be well suited to its specific use case. You won’t turn to the WHO chatbot for some chit chat but to get important health updates or warnings on the current Corona health situation. Acronyms are a special type of functional name. Compared to these, they have the extra benefit of being more approachable if they are identical to human names, as in the above examples.

Finally, a dictionary name can basically be any noun, verb or even adjective you find in a dictionary, offering a lot of space for your creativity. They are multi-functional as they are often used as human names, like Amber, or hint to what your chatbot can do, such as Concierge. If you opt for such a name, make sure that it is linked semantically to your chatbot’s use case or relates to your company’s flagship product, as does Levi’s’ Indigo. To me, names such as Melody or Concierge seem rather randomly picked as they tend to evoke wrong associations. I’d rather expect a music-related service behind Melody and not a medical chatbot as is the case. Thus, make sure your chatbot name conveys the right connotations and does not mislead users. If you plan to localize your chatbot, consider that dictionary names might create a special challenge for translation.

Image by Pete Linforth on Unsplash

The How-To’s

The name of your chatbot should be tailored to your chatbot’s:

• Company brand, tone, and voice
• Personality (character)
• Use case (tasks)
• Target audience
• Industry

You can use these criteria as the basis for your naming process. They provide the overall framework that guides you through the process, determining the general path to follow, and establishing the outer boundaries that you should avoid to cross.

Imagine, for example, that you want to name a financial assistant chatbot. Most probably, you will not opt for a humorous or funny name. Before getting to the fancy part of designing a chatbot name, decide for one of the name categories shown above. The following tips might then help you in the remaining part of the process:

1. Choose a name that is attractive, welcoming, and emotionalizing. Chatbots are made for humans, not for machines. First-time chatbot users might not be as willing to interact with a chatbot called Car Insurance Agent as with one that introduces itself as Carina. Once you have come up with a name, make sure to use it consistently, in other words, don’t release your chatbot and change its name afterward. This might confuse users, and they might lose trust.
2. Pick a name for your chatbot that is easy to remember, write, and speak. This can help users look up and find your chatbot, for example, in their Messenger inbox. Don’t use jawbreakers, especially for voice assistants, no matter how good they sound. Also, don’t have users guess how to pronounce a specific name correctly as this could affect wake word recognition. If you plan to localize your voice assistant, consider how its name will be spoken by people of different languages. How will French or German users pronounce an English name, for example? Will your voice recognition engine be able to handle dialects, accents, wrong intonation, and the like?
3. Think of a name that is unique. If you release your chatbot as a generic “Wine Bot” in a long row of other Wine Bots, this might not be as appealing and distinct as you would want it to be. Even if your Wine Bot is better than all other Wine Bots around, it does not stand out in the first place. Unique also means that once you’ve come up with a name you should double-check on the Web or do some legal research that a chatbot, product or company with that same name does not already exist, to avoid any copyright issues.
4. If you plan to make your chatbot multilingual and keep its original name for all languages and target audiences, the task of naming your chatbot gets even more complex as you need to consider any associations or implications that the name might create in the target culture which are not present in the source culture. Names that work in one culture might be completely unacceptable or even unethical or offending in another culture. There are a lot of awkward or embarrassing product naming examples from the past. Mitsubishi’s Pajero translates as “jerk” into Spanish, and Mazda’s LaPuta as “prostitute”. Little wonder that these cars did not quite hit the Spanish market. Similarly, if you want to use a witty or funny name, consider that humor does not work the same in all cultures.
5. Also, be careful with puns on words. “Penny”, for example, was a financial coach chatbot that was quite popular back in 2015. “Penny” isn’t a bad name at all, it’s a human and a dictionary name at the same time and through its semantic meaning as a monetary unit creates a connection to the chatbot’s use case. But this only works in English, and to restrict it further, only in English-speaking countries that use or at least used pennies as a monetary unit.
6. Consider that different countries use different levels or grades of formality. While people in some countries might be more casual and address each other by their first name, even in business scenarios, this might not be as common in other countries. You therefore need to decide if your chatbot gets just a first name or if it also gets a last name. This makes it very formal but might be recommendable for use cases where the stakes are high, for example, in the banking sector. This might help increase the user’s confidence in your chatbot carrying out financial transactions that involve a certain risk.

Final Thoughts

In the end, there’s a lot of aspects to consider when naming a chatbot. Choose the name for your chatbot as carefully as you would for your newborn or, more technically, for a product that you want to become a sales booster for your company. Your chatbot might be just that: it can help you make a deal — or break it. It’s also a great idea to test your chatbot name with different audiences, especially, if you’re going to release your chatbot in multiple target languages and cultures.

Keep in mind that the name of your chatbot is like the icing on the cake, adding that little something extra. Not even the most clever and attractive name in the world will help if the chatbot itself is not designed well.