The (Ancient) Art of Conversation: Styling Beyond the GUI

What happens when your ability to persuade *is* your UI?

Looking to antiquity to reframe our vision of the future, this talk uses the 5 Canons of Rhetoric from ancient Greek philosophy as a framework to examine how we will design and build chatbots, hybrid UIs, and hidden UI interactions going forwards.

Invention, Style, Arrangement, Delivery and Memory: these canons were conceived to win hearts, minds, and votes. In digital, our measures of success are engagement, traction and conversion – and when our computing interactions are reduced to simple conversations, our entire arsenal may once again be rhetoric. Let’s learn what philosophy can teach us about designing for the future of computing.

Let’s think about how bots and conversational interfaces could look like in the future.

First though, let’s go back in time to ancient Greece and the beginning of democracy. There was an argument going between Plato and the Sophists, who were teaching rich people to speak in public. They were the dark pattern designers of the day; and Plato worried they’d fool people into making decisions against their own interests. Aristotle rejected Plato’s view and wrote a book about persuasion (The Art of Rhetoric).

Part of the idea was persuasion at scale: how to convince a large number of people to do something. This is basically the same thing people are trying to achieve with chat bots and other conversational interfaces like recommendation engines.

Conversational UI has had a lot of bubbles and fizzles. People don’t have a good reason to trust bots or expect a good result. Some of these failures can be solved by finding new heuristics. We can look to those early philosophers for inspiration.

(NB: Laura warns this is cherry picking the most relevant parts, not a full brief on Greek philosophy. It’s worth reading further if you are interested. Although perhaps not diving straight into the original texts which tend to be pretty dense language.)

Aristotle’s Rhetoric has three rhetorical appeals:

  • Logos – the use of logical argument (convincing people with facts)
  • Pathos – the use of emotion (convincing people by evoking an emotional response, which can be both positive and negative emotions)
  • Ethos – the degree to which the audience perceives the speaker as credible. Intrinsic ethos comes from the domain; while extrinsic are signifiers like “I have been working in the field for 20 years”.

The five cannons of rhetoric (Cicero in ancient Rome)

  • invention – finding an interesting way to get into the domain (comedians do this a lot)
  • arrangement – the order of the speech. In form design, clustering related fields contextually is an example.
  • style – the design of the language you can call on (Bus Uncle tells you about public transport but trolls you at the same time)
  • memory – recall, the ability to recall the speech; having a bot remember previous conversations
  • delivery – the tone, the speed and inflection of the speech

A challenge and an opportunity – conversational UI can be deep or shallow, it can adapt to context.

One thing to take away: you already are a great orator, you are using rhetoric in your marketing. A good book to read is Words Like Loaded Pistols

We should one day expect our interface to communicate with speech and reason, not just colour and shape.

(bouncy electronic music) – Hi, thanks Ron, for that.

Okay, so let’s just smash straight in, because we’ve got twenty minutes and we’ve got 2,500 years of history to cover, so let’s see how much I can get out there.

So, I’d like to invite you to go on a trip with me to ancient Athens.

So, this is around the end of the fourth century B.C., so that’s like Christ is born, you go back four more, for 400 more years.

And at this stage, we have a bunch of people experiencing the joy that is direct democracy. So, that means one person, one vote.

It is definitely men, landowners, people who live near the city, but that to the side. It’s reasonably represent, rather.

It’s reasonably direct rather than representational, and I’m not having my speaker notes show up.

That’s okay, I’ll just do without.

So, at this stage, we have people essentially able to walk up to the Assembly, vote, and the majority holds the day, so it’s the great experiment that is democracy. And at this point, we have a really interesting, kind of catty academic fight going on between these teachers, called the Sophists, and this guy you may or may not have heard of, Plato. He’s reasonably well-known Greek philosopher, maybe the most famous philosopher in the world, but that to the side.

And the Sophists, their these guys.

They go around Greece and they teach people how to do public speaking, and in some cases they ghostwrite for them. So, they right their speeches and help them make them more persuasive.

And Plato doesn’t like them for a lot of reasons. One of them is that he charges pretty high fee. Oh, sorry, the Sophists, rather, charge very high fees, so it’s really inaccessible to the poor elements of society. They’re also kind of focused on fancy words and floof, the sort of insubstantial elements of the talk, as opposed to the meat, the interesting sort of chewy stuff underneath that problem.

So, he doesn’t like them, and he says even, “Sophists might be considered “the dark-pattern designers “of their day,” because what they do is they convince people who are poor and maybe less privileged in the society to vote for things that are not great for them.

So, you can maybe think of all American Trump, just quietly. But what they do is they convince people to vote for the most eloquent speaker, as opposed to maybe the right outcome.

So, he says, “In politics we presume that everyone who knows how to vote,” oh sorry, “how to get votes knows “how to administer a city or a state.

“When we are ill, we do not ask “for the handsomest physician, or the most eloquent one.” So, of course, all the Greeks are like, “Sick burn, Plato.” (softly laughs) And you know, a lot of people agree with him, and about another couple years along, this guy comes along, Aristotle, and he’s a student.

He studies under Plato for a number of years, and he mostly agrees with Plato about all of his ideas about the world, but they kind of have this argument about this specific issue, this issue of rhetoric, of using language and words to persuade people to do things.

And ultimately, Aristotle ends up rejecting his mentor’s rejection of the Sophists. So, that’s a nice little logical jump for you. So, he says, basically, I think rhetoric is actually useful, because things that are true, things that are correct and just, are naturally superior to things that aren’t. So, if the Assembly, if the body of people was making bad decisions, it’s because of their advocates, right? It’s because the people speaking up for the worse options have better rhetorical skills, have better ability to persuade people.

So, that’s terrible.

That’s a terrible outcome.

So, if we wanna vote for the, if we want the side of righteousness and justness to win, we need to learn how to advocate for ourselves. So, of course, all the Greeks are like, “Sicker burn, Aristotle.” (audience and Laura softly laughing) Okay, sorry.

That’s the worst of my gifts, I promise. (laughs) (audience laughing) So, at this point you’re thinking, “All right, cool story Laura, “but what does all this have to do with UI.” And it’s a great question, and I would argue it’s scalability.

So, what all of these guys have in common, first the Sophists, and then Plato, and then Aristotle, and then about two centuries after him, Cicero in Rome, and then many, many hundreds of years of academics after that, what they’re inherently interested in is the work of convincing people to do something at scale.

And what I mean by that is, I can sit and have a chat with Adam and convince him this is the right way to code up an accessible auto-suggest, but it’s much faster if I can speak to a room of several dozen people, or even several hundred people, and convince you all at the time.

So, it’s the ability to convince people to do things, to agree with you, to think your way, and it’s scalable. So, what other demeans might we think of where we’re worried about convincing people to do things at scale. Well, basically, everything that is a user interaction, right? So, chatbots, obviously.

It’s literally scaling a conversation.

Personal assistants like Siri, Cortana, Google Now, what are they doing if not scaling the work that the PA used to do? And then you’ve got Google Home, which is again, is a conversational assistant, and it can do lots of tasks for you.

And then, of course, there’s more traditional forms of UI like shopping online, and if you remember Anne Elizabeth’s talk, we were looking at this thought of getting a shopping experience in a retail store. Well, this is trying to scale that recommendation thing, but in a digital format.

And then, there’s things like rubor advisors, which is just quietly what I do, which is trying to scale having a chat with your financial advisor, and stuff like e-health, right? So, when you go onto WebMD and say, “Hey, I woke up this morning “and I felt a little bit creaky,” it’s trying to scale that conversation of having a chat to a nurse and hopefully keeping the hypochondriacs at bay, right? So, conversational UI.

At the moment, there are some really great examples in the field, but there’s a lot of meh.

There’s a lot of things aren’t really landing, and I’ll support that by saying I’ve started to see some literature coming out in the press. Or not literature, articles.

We won’t call it literature, (laughs) but there’s definitely there’s some people starting to talk about the backlash against chatbots, and certainly at the beginning of the year, I think it was Feb., Facebook announced that they had had a 70% failure rate of their chatbots in totality.

And what that means is in their ecosystem of chatbots, they were failing to answer, or to respond appropriately, to interactions with people 70% of the time out of 100. So, not terribly good, right? So, I’d argue that we can better, and we can do better without leaning on natural language processing, or better AI, or waiting for those things to catch up, right? I think we can do better by looking at the work these extremely smart philosophers were doing 2,500 years ago to give ourselves some new ways of thinking about what conversational UI could be right now. So, let’s get stuck into it.

We’ve got Aristotle, Aristotle’s Rhetoric, sorry. Talking about rhetoric is really a tongue twister. If you ever go home, just say, “Rhetoric, rhetoric, rhetoric,” several times in the mirror.

It’s a really good one if you wanna get your tongue going. So, Aristotle’s known for many things.

He was actually one of those kind of all-rounders. He kind of invented a lot of sciency things that we do now, but in terms of this domain, he’s well-known for the three rhetorical appeals.

So, I’d just like to highlight that I’m just kinda cherry-pick some of the best-known things here. This is very much a bird’s-eye view of philosophy, so you shouldn’t go away and think that I’ve taught you everything there is to know about this domain.

So, if you are interested, I recommend that you go and learn a little bit more, but I’m just gonna cherry-pick some of the stuff I think is most interesting and most relevant to what we’re doing right now.

So, the three rhetorical appeals are logos, pathos, and ethos.

So, let’s start with logos.

What is that? It’s basically making an appeal to logic, and when we say logic, we mean inductive reasoning. It’s not like a mathematical, logical proof. It’s not necessarily like logic in the way we think of it these days, but rather it’s like trying to draw conclusions from generalities and from examples. So, if I were to bring that into the digital world, it might be something like, “Hey, we think this is the better landing page “out of the split pass we did, “because it’s converting better.

“It’s got 15% more conversions “and it’s got 20% more click-throughs “on the call to action.” And if Mike Sharp is still in the room, you’ll note I won’t tell you the sample size, ’cause I don’t know if it’s statistically meaningful or not, but (briefly chuckles) we’ll put that to the side. So, this is the Spock appeal, right? This is appealing to logic.

This is saying, “We are rational people, “and we can make rational decisions.” Now, there’s a whole sort of subclass of logical appeals, which I’m gonna dig into, ’cause I don’t have time, but again, I just wanna highlight that you can go have a look at them, and I think that they actually could be really interesting ways of constructing copy, or constructing language that you could be using in these sorts of interfaces.

So, an example in the wild is, I asked Siri, “Why did the chicken cross the road,” and she saya, “Maybe the chicken is standing still, “and it is the road that crosses under it.” And this is obviously facetious, and this is what I would call kind of a reverse syllogism. A syllogism is where you say, “We assume that we have “an idea that we agree is true, and I’m gonna make “an inference based in that.” But in this instance, I assume that the road was still. She kind of flipped that on me and she sort of said, cheekily, “What if it was the road that crosses under the chicken?” So, it’s silly, but you get the idea.

Pathos is the second appeal, and this is the appeal to emotion.

So, this is when you try and make your audience feel something deep in their gut, and you wanna make them feel that emotion in order to drive an action.

So, there’s no point in me making you guys feel, I don’t know, terribly horrible about something, and then not having an outcome I want from that. So, it’s not just emotion for it’s own sake. It’s emotion for the sake of driving an action. So, an example of that is this little guy saying, “I love you.

“I hope you love me, too.” So, pathos, as I said, it drives the audience through emotion.

That could be things like love, or pity, humour, fear, or anger, and there’s lots of examples in the wild of what this looks like, but this one, I really love, surprisingly enough.

This is a little chatbot called Catbot, and it doesn’t have logic or natural language processing or anything going on, but when I chat to it, it just gives me really cute GIFs, and adorable little Unicode, and I just love this thing. And I literally sat there for 45 minutes chatting with it to see what it could do.

So, this is a great example of what we call a delightful interface, right? I’m not getting anything out of it.

It’s not getting anything out of me, but I’m just really loving my time here.

And I feel like we could have more experiences like this in chatbots.

This is an example.

Oh, I wanna talk about this before I play it. So, this is a robot.

This little guy here, he’s called Kismet, and he’s a non-conversational, non-logical robot. And what he can do is make sounds, and he has a little face that has ears and then eyebrows, and he responds, and he can do turn-taking.

So, you talk to the robot, and the robot sort of talks back, but there’s no logic, and there’s no words. There’s nothing behind it, but it’s an incredibly evocative experience. So, I’ll just play this for you real fast.

You can see.

– No, you are.

– [Laura] Aw. (Kismet mumbles)

– [Long-Haired woman] Go? No, I’m not gonna go.

– [Kismet] Ooh, yay.

– Nyah, you’re so silly.

– [Narrator] Through this unique approach, the team created a robot that is able to produce moment-to-moment, and seemingly natural, human responses.

– [Kismet] Here you go, mas.

(boy babbles) – [Narrator] This enables Kismet to appear to have social intelligence.

– Okay, we’ll stop it there.

You see the look on that kid’s face? He is totally, absorbed, right? He’s just so, so into this thing, and I feel like this is the kind of experience we want people to have when they interact with the stuff we design. So, just putting it out there, we could have conversational UIs.

It could be something that’s purely voice, or it could be a chatbot on the screen that has nonverbal responses that are cute, and that could be fun.

I like to imagine that there could be a personal assistant called Mumblecat, and I say, “Hi, Mumblecat.

“What the weather like today?” And Mumblecat says, (mimics cat meowing).

And I say, “Oh, well that’s cool.

“What have I got on for today?” And it says, (meows and purrs).

And it wouldn’t be meaningful, but I would so have a good time with that thing, right? So, I feel like we can be creative, and we don’t have to be bound by what our ideas are of what a traditional conversation looks like. So, ethos.

Ethos is the final appeal to rhetoric, rather that Aristotle defined.

So, ethos is how believable the rhetor, or the speaker, or in this case, the interface, is to the audience. So, it’s very specific to the domain.

So, I could be chatting to you guys about tech, and you kind of go, “Oh well, she’s been working in tech “for a while, so I kinda believe it.” If I went to a medical conference and I tried to talk to some surgeons about surgery, they’d be like, “Who the hell is this chick, and why is she talking to us?” So, ethos is very specific to domain, and there are two kinds: intrinsic and extrinsic. So, intrinsic is the ways in which I try to flag or signify to you guys that I belong to this domain. So, could be using acronyms you recognise, like TL;DR or wearing Converse and not high heels, or lots of things, right? There’s lots of ways that you can signify that you belong to a domain and that you are inside that domain, as opposed to outside it.

And extrinsic is the sort of CV points, right, like how many years you’ve been at a job, or what education you have, or what sort of badges or stars you have on your CV.

So, it could be things like talks you’ve given, or blogs you’ve written, or places that you’ve appeared in public where people already believed you, so they’re more likely to believe you in future, because they kind of go, “Oh well, they must kinda know “what they’re talking about.” So, if I say to Apple, “I love you,” and she says, “Oh, I bet “that you say that to all your Apple products.” This is kind of a silly example, but it’s actually intrinsic ethos.

So what she’s doing is referencing a song, which is “I Bet You Say That to All the Girls,” and that’s an inside reference that I recognise, and I’m like, “Oh yeah, okay.

“It’s kind of a cultural reference.

“We’re kind of vibing on the same wavelength, man.” And this is chatbot that I quite like from Nat Geo, and it’s an Einstein chatbot.

It’s not very smart.

It doesn’t actually have very many responses, but because it has this, what I call, borrowed extrinsic ethos, where I’m believing that this is a little bit smarter or a little bit more interesting than just a stupid chatbot, because it looks like Albert Einstein, right? I’m chatting to it about gravity, and it’s kind of nonsensical.

It’s not really giving you that much, but I’m certainly more engaged with it than if it was just a completely unskinned interface, right? So, that’s Aristotle en grace.

We’re gonna sort of hop forward two centuries now. We’re gonna pull some stuff from Cicero.

So, he’s the next big character in the evolution of rhetoric, and he defined these five canons which, I think, nicely translate to what we call five heuristics for conversational interfaces. So, these are invention, arrangement, style, memory, and delivery.

So, let’s go through them one by one.

Invention is what it is you have to say.

So, I think, particularly in this day and age where we have the entirety of the web to beg, borrow, and steal from, there’s really no excuse for not having interesting copy or good arguments to hand when people ask you questions.

And obviously, if you’re a chatbot on Shopify, I don’t expect you to know about Einstein’s relativity, but I do expect you to know about what Shopify products I have available to me. So, an example could be this guy Oppov, right? So, he’s a little chatbot, and what he does is he basically serves you political news, but it starts by asking you, ooh, sorry, I meant to do this guy.

It says, “Where do you fall on the political spectrum? “Are you left or right?” And then it gives you articles from the other side of the political spectrum.

So, it’s serving you content you might not expect. It’s serving you news that kind of tries to broaden your worldview and maybe challenges you a little bit, pushes your boundaries.

So, that’s a good example of being inventive or coming up with content that’s outside what that user might expect. Arrangement is the order of things.

So, I didn’t have a good example of this from the wild, from conversational UI, but we see this a lot in, particularly, web form design.

So, you know that terrible form you had on your website that you had to break up into chunks and cluster it? Yeah, I can see some nods in the back.

You cluster it so it’s contextual, and you don’t make them go through 14 fields in one go. That’s your arrangement.

That’s working out what pieces of content sit nicely together, and that’s making sure that you don’t kind of jump around in a really abstract or hard to follow way. So, arrangement is important, because if it was weird for a conversation, it will be weird for your interface. And the best advice I have for this is practise your experiences as conversations and work out what feels weird to say, because if it doesn’t work for one to one, it’s not gonna scale into an interface.

So, style.

This is the language you choose, the little flourishes. You might call it the niceties that we think of in interaction design, but for your words. So, I have an example I love for this guy.

This is a chatbot called Bus Uncle.

It’s based in Singapore, and what he does is, it’s just a nicer-skinned, better version of the tramTRACKER here in Melbourne.

So, it’s just helping you work out what bus you need to hop on and how far away it is.

But what it does, it’s really genius, is it uses this Singapore English, Singrish, Creole as the language style.

So, you can see here, it says, “Which street lah, bus stop lah, “or send me location also can.” So, the people in this domain, they know this language. It’s like the language they speak everyday, and it’s really charming to them that they get to hear it spoken back to them.

And he’s also incredibly cheeky.

So, here he says, “Okay, I found you.” and then he trolls the user.

He’s like, “Four hours, ha ha ha ha.

“No lah, four minutes.” So, he’s making it a really funny experience to work out how far away your bus is.

And in some cases, like over here, I don’t know, you can’t really see.

That’s actually a menu button here, but you basically say, “Here I am.” He says, “Oh, I can’t find you.” and then he forces you to tickle him before he’ll tell you where you actually are. So, look down here.

You can see the address has shown up.

So, he’s a little cheeky.

He makes people engage with a little bit more than they might otherwise, and he certainly makes what otherwise might be a really dreary task a lot more fun. Okay, so memory, especially the ability to stand up on a stage and talk from memory, but I think, in user interfaces, there’s a lot of interesting applications we can think of. And it could be things like knowing the context of your previous requests.

So, just like Elizabeth was saying before, if someone really hated a suggestion that that chatbot came up with it, maybe don’t send them three more that day.

That’s gonna be annoying.

Or, if you’ve said, “I don’t want you to track my traffic or my requests because I’m kinda privacy compliant or privacy worried, then you’re very careful to not sort of go against that user’s wishes. And I think, to be honest, in this age of Hoovering up data, memory could be what the interface agrees to forget, right? It’s quite important that our interfaces are respectful of us as humans, and certainly if we’re expected to respect them, we want them to sort of acknowledge our agency as well.

So, that could be things like, we already have incognito tab in Chrome that won’t track our history.

So, it could be the same thing.

You could say, “Hey interface, I don’t want you “to track the next request,” because it could be something not very worrisome like, “Oh, I’m trying to find a gift for my husband “and I don’t want him to know that I’m searching.” But for whatever reason, it’s important that the interface doesn’t records our actions if we don’t want it to.

So, this is a cheeky example again.

But this is Siri saying, I’m asking her, “What is the best way to get rid of a body.” And she did actually used to have an answer for this, but now she says, “I used to know, but I’m very politically gonna not tell you,” so (chuckles). So, delivery.

Delivery is speed, intonation, it’s all the things that make speaking.

it’s about accessibility and usability, as well as persuasiveness, right? It’s how much can you understand what this person has to say? So, I’ve got an example here that we’ll listen to. – [Google User] Okay, Google.

Why is the sky blue? – [Google Now] According to UCR Math Department, a clear, cloudless, daytime sky is blue because molecules in the air scatter blue light from the sun more than they scatter red light.

– So, I cut it off there.

That was literally me just holding a microphone up to a Google Pixel and, so sorry, the quality isn’t great. So, that was kind of a run-on, right? It was a little bit hard to get your head around, and she kind of pronounced it a bit weirdly. So, I recorded this again, just with the voiceover command from my Mac. – [Mac Voiceover] According to UCR Math Department, a clear, cloudless, daytime sky is blue because molecules in the air scatter blue light from the sun more than they scatter red light. – So, it’s not perfect, obviously, but having pauses really helps you understand what they have to say, and that’s a thing we really need to get better at as we build out these interfaces going forwards.

So, just to go back to the beginning, this sort of ethical conundrum that Aristotle was struggling with is rejecting his master’s, rather his teacher’s, sort of primary thinking, and by rejecting that, going on to say, “I composed a masterpiece “of the formal study of rhetoric.” And he felt really bad about this, and I think the way this applies to our current UI is, he’s struggling with whether language is gonna be used for good things, morally correct things, or bad things, like what the Sophists were doing.

And I think that applies equally to interfaces today. Our interfaces, if we take it as given at least, that technology’s agnostic, and maybe you wanna argue with me about that later, but if we take that as a given, then we as the rhetors, as the people building these interfaces, are responsible for both the good outcomes and the bad ones. And I think, certainly, it’s important that we think hard about the ethical implications of our interfaces as we go forwards, and we make that we’re comfortable with the outcomes we’re engendering out in the wild. But I say this as a challenge and an opportunity. I think we have this longstanding challenge of UI, which is it’s really hard to design for noob users, people who’ve never seen this thing before, and power users, experts, but when your interface is a conversation, it’s only bounded by the curiosity of that user, right? So, the conversation can continue on as long as you have content to give them and as long as you can give them interesting interactions, and that seems to me like an amazing way to fulfil one of the original visions of the web, which is a matrix of knowledge, and information, and sharing, as opposed to just an endless series of shopfronts, which, not to say anything bad about shopfronts, but I think we can do more than that on the web.

So, if you take away one thing, I hope it is that you believe that, by now, you all are rhetors, that you are using rhetoric in your dailies lives, in your marketing websites, and even to convince your workmates to instal the newest, coolest bot on Slackbot. And if you do agree this, I think there’s some great reading I can recommend, particularly this one. It’s got such great examples, and it’s really engaging reading.

And if you are struggling with any of this copyrighting or content creation stuff, I highly recommend taking a look. So, just to finish up.

So, Aristotle said, “It is absurd to hold that a man should be ashamed of an inability to defend himself with his limbs, but not ashamed of an inability to defend himself with speech and reason, for the use of rational speech is more distinctive of a human being than the use of his limbs.” Well, I’m gonna argue that it’s not long before we’ll be saying, “It is absurd to hold that an interface should be ashamed of its inability to communicate through shapes and colours, but not ashamed of its inability to communicate with speech and reason, for the use of rational speech is distinctive of a great user experience more so than the use of its icons.” Thanks very much.

(audience applauding) (bouncy electronic music)