All Talk And No Buttons

9 Mar 2017

Posted by Pete Trainor

We seldom realize, for example that our most private thoughts and emotions are not actually our own. For we think in terms of languages and images which we did not invent, but which were given to us by our society. Alan W. Watts

We spent so much time focusing on creating glass barriers between people and their goal that we may have unintentionally missed the single biggest thing that separates us from other species. No matter the gender, race, or religion of a person—unlike other beings—we have a sense of curiosity, we try to understand the meaning of life, search for acceptance and gratification, and get to grips and accept the inevitability of our mortality... oh... and we like to talk... a lot!

What if we could use digital design to help us all get back to having a good old-fashioned conversation instead of all this ‘browsing’ and ‘clicking on stuff”. People are pretty good at conversing. We use conversations to make or maintain relationships, to share or receive information, and to persuade each other to do things.

On the grand scale of things we are evolutionary babies; all of 200,000 years old for goodness sake. Modern civilisation as we know it is only about 6,000 years old, and even though we’re getting good at all this technical stuff, the best thing we’ve ever learned to do right from the off was to ‘have a chat’. We evolved to do it with combinations of spoken words, written text, nonverbal sounds, physical gestures, and facial expressions—and we’ve had a lot of practice at it. So why don’t we just digitise that instead? Amplify what we all do brilliantly, and stop creating lots of digitised ‘stuff’ that we don’t naturally understand. Do it for my mum. She hates the Internet, but she loves talking!

Your health needs it too—there’s enough research that shows people who talk more are often happier than those who don’t. Chatting is good for you. A good gossip every day releases endorphins that make you feel naturally happier. We can chat ourselves happy!

This is predominantly why a ‘conversational experience’ is always going to be more effective. All this noise around Artificial Intelligence, Virtual Assistants and ChatBots isn’t hype, it’s just obvious - People were already ripe for it by virtue of being conversational creatures to begin with. We know how to chat. We’ve been talking to our devices for years, but it wasn’t until very recently that those interfaces also became able to respond with any kind of credibility—we just hit the perfect intersection between technology and humanity. Dialogue-based interaction, either spoken or written, using your phone, tablet or computer (and very soon, the in-ear interface) is here and is not likely to go away. But of course with all these new ideas, the risk is that designers just dive on it because it’s a shiny thing without truly understanding the psychology behind what makes a good conversation; and that’s where we’ll also see a lot of failures in this space. It’s easy to forget that we’re not digitising behaviour—we’re behaving in a digital world. It’s time to go back to basics and learn to talk again.

As simple as a conversation might seem, being human is a lot more complicated than it first appears and so conversational interfaces require designers and developers to leave behind current practices and adopt an entirely new mindset. I started the book by talking about looking outside your comfort zone. This is another fine example of that. You need to look towards language, social psychology and linguistics for answers. The familiar design patterns we used when we were chopping up pictures and making them wobble around screens won’t work in a conversation-driven interface. As visual design is demoted in favour of words, what you say and how you say it become more crucial than ever.

Recent success stories like Amazon’s Alexa and’s Amy Ingram are great examples of a conversational technology that amplifies the personality of the bot. They spent a lot of time on the personality stuff so people would feel comfortable with it. But, with the rise of messaging platforms like Facebook Messenger, Telegram, and Slack—bots that communicate with people by sending and receiving messages for all sorts of uses—it’s going to be fascinating to watch how this space evolves. More often than not, designers are spending almost no time crafting the perfect personality—the bit that makes things like Alexa work well.

Ai powered chat is reaching the point of being ‘OK’ at holding down conversations with people, but there’s no way they’ll pass a Turing Test (to pass the Turing Test an Ai has to be capable of lying as well as telling the truth!), but they probably don’t really need to be more than OK. In fact, the obvious flaws in Ai may be precisely what makes them work well. They just need to observe and react to what’s being said to them, not lead or attempt to be smarter than they actually are. The time is ripe for conversational experiences, precisely because we can still identify it for what it is—we don’t feel deceived. As Asimov predicted, “we like robots that can be readily identified as robots” and so the key to the conversational experience may be a transparently bot-like personality and simple responses, rather than something so smart it’s intimidating or uncomfortable.

Conversation is really a term that refers to the back and forth interaction in which two parties come to understand each other. That’s the rub of what makes a conversation good—the back and the forth. So designing experiences that create that dialog, although hugely complex, is what you want to be striving for.

You don’t need to be a social scientist to get a conversational experience right on the basics, just go back to what you already know and you’ll find your groove. Remember that a conversation goes nowhere unless you have a partner who listens to what you’re saying and responds in a way that keeps the conversation going. A good conversation is constructed by a speaker and a listener each doing their part. A great conversation is constructed with respectful, interesting, and enriching content. You learn something. You teach something. Your knowledge increases. Your curiosity is piqued. You relish the time spent together and that is what you should be striving to design–something that makes people feel like their time was invested well.

Let me give you the 8 areas that I focus on that might help you create some engaging conversations and help you to get something more than a different type of interface shaped like bubbles out of your product.

Always open well

The opening gambit is also the key to all success. Get comfortable with Daniel Kahneman’s ‘Two-Question Technique’—always start by asking people about something positive in their life. Only after they reply should you ask them how they’re feeling about life in general. A positive answer on the first question will lead to them feeling more positive about their life in general when you ask the second question. Any emotionally significant question that alters a person’s mood will have the same effect.

By being specific and leading the conversation in this way, you can also overcome the single biggest flaw of the conversational experience—natural-language processing. Humans don’t speak or write in perfect grammar that is free of typos. They use slang and nuance to express what they want—aspects that can make even the most powerful computer struggle. Machines just aren’t good at engaging with all the quirks we’ve picked up over the last 200,000 years: accents, bad grammar, and colloquial expressions. The main issue with natural language processing is that even if your Ai gets 80% of the sentences right, it will fail in the remaining 20% in a very stupid, non-human way. Even if your error messages are witty and fun, people will get frustrated. We’re getting closer to having machines that can understand those oddities, but we’re still quite far from it. Flip it—get the conversations perfect for the 20% of common queries and not the 80%.

Design your conversation to give information

This includes expressions of emotions and discussions of intent, as well as simple task-oriented requests—“Please give us more information so we can help you better.” Also, knowing when someone is giving you information is important because it leads to the question of what they want or expect you to do with the information. In the case of a customer telling you they are unhappy, they may be signalling that they want something to change—or they may just be venting and wanting to be heard. But if you don’t factor in the reaction to your request for action—it’s not a dialog, it’s just a prompt disguised as chat bubble.

Design your conversational experience to get information

Someone who says something in order to get information won’t necessarily pose it in the form of a question. “Tell me about…” is one way that a request for information can arrive in the form of a statement. The information being sought may be your opinion, your assessment, or your best guess. It’s not always going to be hard and fast facts that someone wants. They may be gauging the mood to ascertain your willingness to act before they ask you to do something.

Get people to do something

It is imperative that the conversation you create is designed to get someone to do something. This is where we get into the guts of why people say things: to create actions on the part of other people. These actions are normally created by giving information—but in those cases, the desired actions may not be readily apparent. The person asking may be counting on a certain response. Other times, the speaker’s motivations are obvious: “come here”, as a parent might say to a child or “kiss me, fool”, as someone might say to a lover.

Make people feel good

If you don’t design your conversations to make someone feel good, then don’t bother doing it at all. Compliments are the simplest manifestation of this. It’s just common, human courtesy and all too often the experiences emerging that use the conversational paradigm leave out the basic courtesies that separate us from bovine.

Get people to talk about themselves

You have to encourage people to talk about themselves—mainly because it can trigger the same sensation of pleasure in the brain as food or money. People do love to talk about themselves once you tip them over the edge. In some of the experiments we’ve been conducting with conversational interfaces over the last 18 months, one of the things we’ve really started to observe in the data is that a lot of people are actually more comfortable speaking about themselves to a bot—even one that’s a bit dumb—than face-to-face or using voice. It’s because people don’t have to be spontaneous in their replies and because they don’t feel judged for the speaking skill-set they have. In a lot of cases, people also feel like they can communicate more effectively using words on a screen than in ‘real’ conversations.

We studied 140 volunteers using one of our mental health bots, {SU}. We told half of them it was controlled by a person—like a puppet. We told the other half it was computer-controlled—fully automated—and there was no human on the other end. The volunteers who thought they were talking to a computer engaged in less “impression management” and displayed emotions like frustration and sadness more freely and more quickly. People actually enjoy self-disclosure when it feels anonymous or safe.

Give feedback and ask more questions

If you use questions to guide people toward the errors in the thinking process and allow them to come up with the solution themselves, they’re less likely to feel threatened and more likely to follow through. Don’t assume to know all the answers—you won’t, so ask for guidance and advice. New research shows that guidance seeking is a surprisingly effective strategy for exercising influence when we lack authority. In one of our experiments, when we focused the chat on the goal of getting a sale, only 8 percent of chats reached a successful agreement. But when we asked the buyers for advice on how to meet their goals, 42 percent reached a successful agreement. Asking for advice encouraged greater cooperation and information sharing, turning a potentially contentious negotiation into a win-win deal. Seeking advice is among the most effective ways to influence peers, superiors, and subordinates.

Be a good gossip

Finally: Gossip (but positively). Research shows what you say about others colours how people see you. Compliment people and you’re likely to be seen positively. Complain and you’re likely to be associated with those negative traits you complain about. When you make small talk and gossip, listeners unconsciously associate you with the characteristics you are describing, ultimately leading to those characteristics being “transferred” to you. So only say positive and pleasant things about the focused situation, and you are seen as a nice person/bot. In contrast, any negativity, and people will unconsciously apply the negative traits and incompetence to the conversations (and ultimately you/your product). Bots have to make a negative sound positive.

Speech is a very important aspect of being human—a whisper doesn’t cut it. Designing conversational experiences is really hard because the concept of a conversation is by its nature very simple and we pride ourselves in designing complex solutions to simple challenges. Building conversational interfaces doesn’t just come with technological challenges—like Natural Language Processing—it also has lots of social ones. As the designers of these conversations, it’s our responsibility to solve this part of the problem—to craft the perfect personality and conversation.

Remember this—you can’t get away from the value of human expertise, wisdom, and our unique problem-solving ability; it’s what evolved us to where we are today. So start planning your conversational experiences by powering them with real humans on day 1, doing 100% of the work. Introduce natural language powered bots later to automate the 20% of the most frequent questions, or just to do the on-boarding part. Then later—much later—when you have a trove of dialog data, you can move up the automation to take over more of the requests. Embrace the Mechanical Turk.

It may be our actions that define us, but it is our reaction that changes the course of things. Study what makes a good conversation not what technical platform is the biggest and cleverest and you’ll start to find the sweet spot for making your idea resonate with the humans consuming it.

Pete Trainor
Posted by Pete Trainor

Pete Trainor is an author, technologist, accidental polymath, mental health campaigner and co-founder of 'Us Ai' in London. He talks all over the world on creative & social technologies, data, artificial intelligence and the physiological & psychological effects on their audiences. Over the last three years, Pete has helped to pioneer an entirely new approach to Ai focused services, one that looks at ‘self-evolving systems’ and ‘minimum viable personality’ to help solve societal and human issues. His recently published, bestselling, book, “Hippo - Human Focused Digital” takes a philosophical look at technology and design, challenging us to look inwardly at the self when designing future technologies. Pete regularly appears in UK national and international press as an analyst on emergent technologies, and tech markets. Pete also chairs the Ai Think Tank for BIMA. He has a very simple philosophy: Don't do things better, do better things.

View profile
Read next

Newsflash: your agency doesn't stand-out

9 Mar 2017

Posted by Robin Bonn

Keep your finger on the pulse

Join our mailing list to be the first to hear about cool opportunities, hot events and more.

Password Reset

Please let us know your email address and we'll send a new password straight to your inbox.