2016 was the year of the Chatbot. Microsoft, Facebook, and Google have all released chatbots, demonstrating the importance of artificial intelligence on the web.
What is a chatbot and what can you use it for? “A chatbot is a service powered by rules that a user interacts with using a chat interface. The service can be any number of things, ranging from form to function and live in any major chat product” (Matt Schlicht, Chatbot Magazine).
Practical Uses for Chatbots
Scheduling a week can be one of the biggest pains for the planner-type of users. A chatbot coordinating with another person or bot can automate the calendar system. After coordination, the bot can send the user a plan of their week. Users can also block out busy times so chatbots can schedule around these timeframes.
An example of a scheduler is Meekan. Meekan can schedule lunches, check calendars, book conference rooms, or find flights. In Slack and on screen, Meekan seems like an ideal digital assistant.
Chatbots can simplify the introduction of a new user to an app or website. Chatbots provide a quicker response and better user experience by being “there”. Because the bot lives in the app, it can respond immediately without requiring a user to pick up a phone. This allows businesses to operate without requiring 24/7 customer support for onboarding.
Dolores Landingham is an employee onboarding bot in Slack for 18F. The bot was named after the West Wing character and is adapted to support users in different time zones. Users can subscribe to “benefits” or “legalstuff” and will not be flooded with excess information. With Dolores, employees no longer need hours of in-person training or mountains of paperwork when starting a new job.
Connecting users to websites
Because Chatbots live in a platform, they can track users’ behavior patterns. Chatbots may suggest potential partner websites or social networks as recommendations. As a recommendation source, chatbots make it easier for people to find similar sites.
Blair, a Singaporean voice-integrated chatbot, was designed on the Facebook Messenger platform as a restaurant recommender and booker. Blair is a personalized OpenTable and has potential to partner with wine bars, dessert parlors, or other hospitality services.
Advertising benefits from the rise of chatbots. Chatbots can identify users’ shopping patterns. From these behaviors, chatbots can suggest items the user may enjoy in the platform. The chatbot can even act as the user’s personal shopper.
Spring, a shopping startup, currently has a customer service bot that responds to users’ requests 24/7. The bot allows users to buy, save, or ask questions about items it recommends from brands that the user follows.
When trying to achieve a new goal, people like to be accountable or may never end up achieving the goal. Often, friends hold each other accountable for goals. Sometimes a person may find a goal embarrassing or want to keep their intentions private.
A chatbot can motivate by using personalized, conversational notifications to the user. The user can also set up reminders in intervals of user preference. The chatbot is always there for the user because it is personalized to the user and engages the user with personality.
I tested Opal, a motivational bot prototype on BOTTR. Opal has prepared responses I could click on, which she responded to. If I chatted saying that I was down, Opal would respond with a positive message. However, Opal was 1-dimensional and unable to motivate me to work out. Opal could motivate in terms of positivity, but not goals.
Chatbot Experience Essentials
If bots are not personalized to the user, a usability problem arises. Chatbots should address the user –“Hey, Amelia!” and instead of asking generic questions, provide information the user wants to know. The bot should be able to respond to customized questions and responses.
MIT’s chatbot, Eliza first demonstrated this attention problem back in 1964. Eliza played a psychotherapist by asking standard questions and rephrasing answers as questions. If the user decided to be unpredictable and “go off script”, Eliza would break.
Eliza’s problem is still prevalent today. Opal the motivational bot was only able to respond to prepared responses. The bot could not handle unpredictability and broke when I asked a 2-dimensional question about motivation. The bot could only handle happiness quotient. It could not understand both happiness and goal setting.
Current bots’ increased personalization are creating better user experiences. Blair is able to suggest restaurants and book dinner for users. In the future, Blair will be able to book other hospitality and service partnerships and anticipate users’ wants. Similarly, Spring can anticipate shoppers’ likes and become each user’s personal digital shopper.
Bots can also help with professional goals. Because Dolores Landingham is personalized to an employee, the bot can guide the employee in HR and paperwork questions. In the future, Dolores can automate timesheets, checking in, benefits, and other “busy paperwork” to save time and increase employee productivity.
Solution: Personalize with Conversation
To personalize a chatbot, conversational design is essential. Designing conversation responses involves intention and figuring out linguistic patterns, tone, and interactions. Examples of conversational design include drafting scripts or language mapping.
What exactly is a language map? Language maps are conversations broken down by sentences and words. The most common type of language maps are Syntax Analysis. Syntax Analysis breaks up strings of words to determine syntactical relationships.
With syntax maps, the user can form different types of responses and meanings from a statement. If desired, the user can create multiple responses, duplicate the idea, scale to a broader meaning, or focus on a specific generalization.
For chatbot use, syntax maps are very helpful because they can allow users to predict responses and create alternatives for the bot to chat if the user goes off-script. Although hard to predict for user error, the chatbot has potential alternative response paths it can take and will not break.
Attention must be maintained from the beginning of the chatbot experience. Once the user’s attention is lost, the chatbot’s game is up. Engagement has to happen throughout the experience.
An example of engagement loss is Meekan. Meekan has to gain permission to a user’s calendar through a sign-up link. The sign-up link causes friction and takes the user out of Slack, into a separate page. Although I clicked the link, Meekan lost me after two or three clicks that did not allow me to give the bot permission. After testing out Meekan for 4 minutes, I gave up using it because I could not get past the permissions page.
Engagement loss also occured when I tested Opal. Once Opal was unable to answer my two-dimensional question, I stopped paying attention to the bot and exited BOTTR. The immediate response was, “Oh, Opal doesn’t work.”
Solution: Engage with Visuals
The chatbot has to be eye-catching, but not a real person. Designers must use visuals or conversation to communicate that a chatbot is a robot. When testing the CARL bot, I discovered that users liked the image of the bot as an illustration. Many users expressed their uncomfortable feelings of a bot with a human image.
Users automatically associate chatbots with people or users they know. Users create and impose an identity onto the chatbot. An IDEO study analyzed women’s responses to a “coach” chatbot. Female users projected a pushy, male coach identity onto the chatbot. The “coach” was actually a female human controlling the bot. Even without intention, people humanize bots to connect bots with what is familiar.
Chatbots have the potential to create better user experiences. They are quick, responsive, and always available. Currently, chatbots play the role of digital assistants. No longer do people need a provider to find house cleaners, laundry cleaners, or physical assistant to schedule appointments.
As technology further develops, chatbots can provide an increasingly personalized level of service and remove the in-person assistants’ roles completely to make users’ lives easier at no price. Further chatbot potentials may include more voice integration, personalization, and memory storage to address a user’s current needs.