Thoughts about robots and UX
TL;DR: When we want to use a bot to enhance the user experience of your product, you must aim at a conversation between two humans. First of all — see if a Bot fits this task. Then, think what’s the Bot personality would be. And last: act out the conversation to see if it works for the user. Oh, and there are a lot of quotes of Marvin the Paranoid Android from The hitchhiker’s guide to the Galaxy.
Last week I attended two presentations in regarding Conversational UX. The first was given by Amir Shevat, followed by that of Dr. Jacob Greenshpan. Now, in the UX world it is customary to say that you’ve been doing [ux-hype-word] for over 400 years now, and that you actually was the first to sit down with Plato and Socrates and talk about it. I admit: the first time I heard about Bots and Conversational UX was about five month ago.
In November of last year I happened to sit with Yaron Ben Adiva, Nadav Rikover, Amir Shevat and Dr. Jacob Greenshpan in a restaurant somewhere the Bay Area (can’t recall the name). We sat there, eating good American food and being Israeli loud. Amir had just recently joined as Director of Developer Relations at Slack, after doing similar positions at Google. ‘The Bots are upon us’ Amir said. ‘Google, Apple, Facebook, Microsoft, Slack are all into it, so it will happen’. That sparked a conversation between Amir and Dr. Greenshpan, about what are the challenges bots pose in the UX context. And when these two speak, I do something I’m not accustomed to: just listen.
Fast forward 5 months later, we’re in Tel Aviv, at AWS pop up loft event. Amir is about to give a lecture, his 10th in about 3 days (The energies on this guy should draw the attention of the people at Tesla). Amir didn’t talk about bots, but on personal assistance. Within a year, he claims, we will all have some kind of personal assistance. A bot is just a user with a machine behind it. And wouldn’t it be nice if we could tell someone to fill in our expanses for us? Or just ask someone to find out what is the best time to make a meeting? This is possible now already and if not it’ll be within a matter of months. But the problem is not how we make machine fill out forms and checking data. It’s how we make humans trust them and wanting to use them.
The solution is in how we humans see the service the bot gives us. Our mental model of it: When we call or send a message to our boss’s PA, we don’t give her commands. We open with a ‘hello’ or ‘hi’; we ask politely, or impolitely to arrange a meeting with our boss and several others. We even know if the PA is a she or a he and we adjust the wording that we choose. In short, we don’t perform a task, we make a conversation. And this is the secret to creating a successful bot. Enable a conversation.
As Amir says:
“We’re making UX for tens of years. We make conversations for tens of thousands of years. Make your product a conversation”.
Conversations are human. And just like a waiter don’t just take your order but introduces himself, so does a bot should introduce itself to the user, tell them what they expect. So now the question is how ‘humanoid’ should our bots be? Is putting eyes on them and make the smile like Clippy enough? Enter Dr. Jacob Greenshpan, a cognitive psychologist by title and one of Israel’s leading UX professionals.
When Amir finished his presentation Dr. Jacob Greenshpan came to the stage. Dr. Greenshpan’s gives more of a spectacle than a presentation. I’ve seen him make the whole audience stand up and dance to understand how the Apple Task bar work magic on us. By the end of his presentations you feel that you want to know more about everything. So I naturally was excited.
For me there were several important messages in the presentation:
1. Humans are limited
The reason we need bots is because we’re limited. Cognitively speaking and otherwise, we have limitations. That’s the reason we need a Bot in the first place. But we need to remember this when we utilize a bot. For instance, if we use a bot to read out a list, most probably the human will fail to remember it. Think of IVRs. If we remember that human memory is not efficient then all we need to do is have the Bot remind the human every now and again what the options are.
Marvin: “Reverse primary thrust, Marvin.” That’s what they say to me. “Open airlock number 3, Marvin.” “Marvin, can you pick up that piece of paper?” Here I am, brain the size of a planet, and they ask me to pick up a piece of paper.
2. Select the right assignment for the bot
Although bots are the hot thing right now, and for some of us it is easy to make one, we need to ask ourselves whether or not the task in front of us is suitable for a bot. One of the questions that can guide us is whether or not the assignment is spatial by nature (locating, navigating, etc) or semantic (dictating, answering questions in forms etc). Dr. Greenshpan gave the example of getting directions from a bot: our brain will struggle between remember the instructions and translating them into something visual. On the other hand, a bot can give us a great weather forecast or help us find a flight.
For instance: I want to go from Tel Aviv to Dublin. All flight and ticket websites offer me somewhat the same: give us destination, give us dates, give us who’s flying — and we’ll find you a ticket. Well, almost. What we’ll do is we’ll show you a list and you’ll have to sort through it, scan information and employ some decision mechanism. But this is a great semantic mission. I can ask a bot ‘what are the best tickets from Tel Aviv to Dublin’? The bot will then tell me: ‘well, define best’ for which I will say ‘cheap price, but not more than one stop and good sits’. From here the conversation will roll by itself. Even more: I will be willing to do this assignment throughout the day with the bot popping every now and then with a possible answer. On the other hand, in front of the screen, I need to resolve this right now, which will make me impatience.
It is also easy to test if this user task is suitable for a bot in the design stage, by letting someone play the bot and see when the person paying the user got tired from the explanation or what information they got right.
3. If you build a bot, users will develop a feeling towards it.
One of our limitations as humans, or maybe benefits is that we project human feelings on almost everything. Tamagotchi didn’t need baby face for us to treat it as such. Jacob proved this by showing the audience strawberries and indeed we all felt something towards them: either happy, or hungry or many other human feelings that none of us ever felt towards Clippy even though it had eyes.
Amir told us how people tell him to give his meeting setting bot a raise. I noticed how Amir raised his voice and spoke slower to his Alexa tube during his presentation, as if he wanted to help her complete the mission. Also when speaking to Google, I sometimes find myself say ‘Thank you’ even though it is not supported, so I’m left hanging (It does know how to react when I say ‘Never mind’ when it fails, which is a start). I also speak slower, with a different pitch, so that my English will be clearer to it. I never have this patience with a slow-working app on my phone: if it doesn’t give me an answer in nano-seconds, I will aggressively tap the screen again and again (I’ve also thrown one or two phones against an occasional wall. Not a smart thing to do).
Conclusion: the humanity of the bot is not in giving it eyes, a smile or a face. It is about the small behavioral things, which leads to the fourth take away.
4. In the eyes of the user, your bot will have a personality. You can (and should) decide which it’ll be.
Marvin: You can blame the Sirius Cybernetics Corporation for making androids with GPP…
Arthur: Um… what’s GPP?
Marvin: [despondently] Genuine People Personalities. I’m a personality prototype. You can tell, can’t you…?
We know as UXers that emotions are a big part on how we perceive and use our services and products. It’s only logical that it’s the same for Bots. So when we decide on using a bot for an assignment, we should define what kind of personality the bot has and what kind of emotions we want to evoke during the conversation with the bot.
Conclusion:
Bots are suitable for (mostly) Semantic user tasks as long as they (the bots) have the right personality and the right approach to the user. If we think how to use them before we make them, we will improve the user experience of our service.
