5 Ways to Build a Better ‘Bot
Empathy Is the Secret Sauce for Next-Gen AI Solutions
Our world is increasingly powered by artificial intelligence. By the end of the next decade, AI will have generated $15.7 trillion in economic growth and touched virtually every area of our lives. But while AI is an incredibly powerful problem-solving tool, we’re still a long way from the talkative, genuinely intelligent AI personalities that many of us grew up dreaming about.
The truth is that while machine learning enables computers to spot patterns and carry out complex tasks, human-like intelligence (or even just the ability to convincingly fake it) remains elusive. Sure, some hotels now have robot butlers who’ll help with your luggage — but they’re lousy conversationalists, and wouldn’t have a prayer of passing the Turing test.
To find AI systems that are genuinely capable of talking to humans, you need to look to more specialized tools designed for precisely that purpose. I’m talking, of course, about chatbots and AI assistants: those virtual sprites and avatars that pop up to offer help and advice when you’re trying to navigate a website, contact customer service, or use a new software tool.
It might sound strange to hold up chatbots as the future of AI — but the way people interact with AI technologies, and digital tools more broadly, has never been more important. As AI seeps into more areas of our professional and personal lives — including sensitive fields such as healthcare and telemedicine — we urgently need frictionless conversational interfaces that are genuinely helpful and pleasant to use, rather than clunky, annoying, or downright creepy.
To create those technologies, we’ll need to create AI systems that aren’t just smarter and more functional, but that are also more empathetic, more emotionally intelligent, and more attuned to the needs of their users. Here are 5 key lessons for companies looking to innovate with conversational technologies:
Don’t try to fool your user.
The “uncanny valley” is real, and nobody likes to feel they’re being tricked by a machine that’s “pretending” to be human. Make it clear from the first moment that users are interacting with a ‘bot, on the other hand, and you’ll often find that users relax into the interaction. In fact, many users wind up feeling more comfortable interacting with a non-judgmental AI chatbot than with a human being. In sensitive health applications, for instance, patients are often more forthcoming when discussing their symptoms or concerns with a virtual avatar, precisely because they know they aren’t being judged.
Paradoxically, even though users know they’re talking to a machine, they still like to be treated with sympathy and respect. The mechanics of conversations make a difference, so pay attention to niceties such as saying “Thank you” after the user gives you information. And don’t overlook the linguistic markers of empathy: if someone expresses frustration then having your tool echo their problem back to them, or simply say “I’m sorry” or “I understand,” can be an effective way to foster more productive and rewarding interactions.
Know your audience.
Not everyone wants the same thing from a digital interaction, so make sure your tool meets users where they are. Giving people the option to toggle between text interfaces and speech, or to change the faces of an avatar to match their preferences, can eliminate frustration points and help users to feel ownership of their digital interaction. The best tools flexibly deliver rich results across a range of formats — an especially important consideration if you’re delivering content in rural or developing-world contexts where limited bandwidth places a constraint on digital interactions.
Think about content.
Many conversational AI tools are intended to walk users through a script that’s derived from preexisting written materials. That’s fine, but it’s important to think about whether a chat-based system is a right format for a given task. If you’re delivering hundreds of pages of content, it doesn’t matter how smart your AI is: the user will tune out. Avoid walls of text, and don’t be afraid to link out to detailed explanations or other external resources rather than trying to squeeze everything into a chat-based interaction.
Let users call the shots.
We still remember Microsoft’s Clippy assistant as annoying and intrusive because it acted like it knew better than us what our needs were. The best avatars suggest and guide, but also take correction and redirection, and never assume that they know best. Empathy boils down to aligning your tool’s language and responses to the user’s changing needs — so let the users express themselves, and give them a chance to steer the conversation in the direction that makes the most sense for them.
Of course, none of these principles will lead to a conversational system that’s actually intelligent in the human sense. At the end of the day, even the most sophisticated AI tools primarily leverage linguistic patterns, domain-specific knowledge, and AI processes to interpret incoming language and deliver appropriate responses.
But because the human brain is wired to respond both emotionally and cognitively to conversational patterns, such mechanistic AI processes can still yield remarkable results. Even when people know full well that they’re dealing with an unthinking, unfeeling machine, they’ll have a qualitatively different — and, in many ways, better — experience if they perceive that machine to be treating them sympathetically, responsively, and thoughtfully.
Empathy, in other words, can be effective even when the user knows rationally that it’s only simulated. Building conversational AI tools is really about creating an illusion — but it’s a remarkably powerful illusion and one that can be harnessed to deliver a far more compelling and effective experience for users across a wide range of areas, from eCommerce to fintech, and from healthcare to customer support. By making AI systems incrementally more lifelike and empathetic, with richer characters and smarter natural-language processing, we’re learning to engage, inform, and assist people more effectively than ever before.