It’s time for conversational AI across the board. This should be pulled by industry and pushed by the field. There’s no time to waste. We need to (really) talk with our intelligent brethren as soon as possible. This technology is essential for the wide scale adoption of AI and machine learning applications.
Everyone knows bots are stupid. They’re also annoying. I have no idea why a company would develop and launch a customer service bot they know will frustrate their customers. Actually I do – and so do you. It enables them to reduce the number of human agents they pay with actual money – even though they know their customers prefer talking to humans.
So how do people feel about bots? Listen to Chris Elliott:
“It’s time for a reality check. Chatbots are killing customer service.
“Obliterating it, maybe.
“If you’re a customer, you probably already know that. The computer programs that conduct clumsy conversations with you when you have a customer question are maddening. But if you work for a company, maybe you won’t believe chatbots are killing customer service until you see the evidence … well, now you have it, thanks to a survey released this morning by CGS, a business applications, learning and outsourcing services company.
“But are chatbots driving customers away? Maybe. After I reported on the rise of travel chatbots in my weekly Washington Post column, I heard from many readers who said the thought of talking to a bot was a turn-off. Perhaps they had the same experience I did when I tried to engage the program in a simple conversation.”
I’ve also had horrible experiences with bots, and so have you. Chatbots are often incredibly dumb and circular. The flawed design is astounding. Many bots present a menu from which you choose a problem area, but once you select “other,” you’re looped back into the same menu. Over and over again.
Are bots improving? They are, especially in well-bounded areas where questions can be semi-accurately predicted. But as soon as a customer leaves the script, well, that’s when frustration skyrockets. The requisite skill? Knowing how to get human agents by tricking bots into a handoff. But that’s not the answer.
Pieces of Conversations
“Asking: Engaging and seeking information.
Informing: Giving information.
Asserting: Stating something as true.
Proposing: Putting forward argument.
Summarizing: Reflecting your understanding.
Checking: Testing understanding.
Building: Adding to existing ideas.
Including: Bringing in others.
Excluding: Shutting out others.
Self-promotion: Boosting oneself.
Supporting: Lending strength.
Disagreeing: Refusing to agree.
Avoiding: Refusing to consider argument.
Challenging: Offering new thoughts to change thinking.
Attacking: Destruction of their ideas.
Defending: Stopping their attacks.
Blocking: Putting things in the way of their arguments.”
How many “interfaces” have anywhere near these capabilities? Sure, this is too much to expect from a conversational interface but is the general direction wrong? Not at all.
Conversational Chatbot Design
What about chatbots? If we reduce the above number of human-like capabilities and think just about bots, here’s what Cobus Greyling wants – what he describes as conversational AI design principles:
“When we take turns to speak, also referred to as dialog turns, interrupting each other is avoided and the conversation is generally synchronized. This is our way as humans to manage the state of the conversation. As humans we do this intuitively and effortlessly.
“Why A Persona?
“You must see a persona as a design tool, and this tool assists in the writing of a conversation. Prior to start writing the dialog, you need to have a fairly complete understanding of who is communicating to the user.
“What constitutes a persona? A few elements are used, like tone, script, personality and you should know what your persona will do or say in any particular conversational situation.
“Speech recognition is not the challenge. The challenge is understanding, extracting meaning and intent and conversational entities.
“Your conversational interface needs to keep track of context in order to understand follow-up intents … unless the user changes the subject, we can assume that the thread of conversation continues … this allows for follow-up intents to be detected with greater easy in the customer conversation.
“A lack of variation makes the interaction feel monotonous or robotic. It might take some programmatical effort to introduce variation, but it is important.
“Users Are Generally Informative
“Users are bound to supply more information than what you might expect. This will necessitate you to handle quite verbose user dialogs. Especially at the start of the conversation.
“Keep The Dialog On Track
“Your conversational UI will be domain specific, hence you will need to manage the dialog in a subtle way to ensure users understand the purpose and aim of the interface. You might not always be able to handle all cooperative responses from a user. But you should always be able to use lightweight and conversational exception handling to get the dialog back on track in a way that doesn’t draw attention to the error.
“Move The Conversation Forward
“We all had conversation with bots which are sticky, repetitive, rude or plain unhelpful. You expect your user to be cooperative and informative, and your bot must be the same. Always sharing a dialog which is intended and helpful in moving the conversation forward and to conclusion.
“Stick To Your Domain
“In any conversation, saying too little or too much are equally uncooperative. You must try and facilitate your bot’s comprehension by trying, via the script, to keep the user’s response brief and concise. With optimal relevance to the current context.”
These design principles are solid. They move us toward human-to-human communication but with a machine, which is the goal. Singularity issues aside, this is what we want and need – however “simulated” (for now) it might be. We must respect technological limitations while pushing them the right direction.
Conversational AI to the Rescue
When I need help, the app should “understand” me, know exactly what I do, what I mean, what I need and what my perfect outcome should be (in multiple languages in ultra real time). Is that too much to ask? Today it is. But that’s where this needs to go – and not just for customer service, but for all interaction with intelligent systems. It also needs to be proactive, where the systems tell me what I need, when I need it and how to get it. Or, just tell me it got it for me — that’s it’s done, and I can return to whatever I was doing. Until we get there, AI and machine learning will be limited to well-bounded, deductive inferential tasks – what the field calls supervised learning with tons of labeled data that invokes linear regression to repeat human decision-making. While this is OK, it’s not what humans ultimately want or need. Supervised learners are today’s technology slaves. What we want are real partners.
The whole “total experience” (TX) world needs a wake-up call. Customer service bots are an essential part of TX – often the entry into the “experience” companies create for the “benefit” of their customers. Is this how guests should be greeted? Is this how first-time guests should be greeted?
Incredible Strengths & Significant Limitations
AI and machine learning are transforming personal and professional processes at a pace I never anticipated. I believe it’s the most important technology (thus far) of the 21st century. But there are problems. The dirty little secret? Yes, most of the most-powerful applications are in well-bounded domains where supervised learning with regression is the preferred approach. The field needs more before it truly transforms our personal and professional lives. Conversational AI is a huge step in that direction. We need to get there before the bots kill us with anger and frustration.