Are you okay with not knowing if you’re talking to your friend or a bot?
This article was co-authored by Stella Jeong.
Let’s the get the obvious stuff out of the way first. The chat ecosystem as it stands today is very fragmented. Most of us feel this pain every day — irrespective of what platform you are on. Yes, it’s a mess. But that’s not what we want to talk to you about today.
Instead, we are going to tell you a story of when we tried to switch to Google Allo, and it was the first time we were creeped out by a chat application.
Some background: Allo is a new(ish) chat application by Google. It introduces a bunch of new features, like Google Assistant integration, (very cute) stickers and gif search. But most of all (in this post) we will be talking about Smart Replies — automated recommendations that are based on the conversation being had before it.
Part 1: Getting Started
We go to the same school. Our association with each other started off as a mentor-mentee relationship, and has since evolved into a professional and personal friendship. Before this, we had been using Facebook Messenger (with a little of text messaging) to communicate. When we decided to switch, one of the first smart replies Anish got was “I love you”. Uhhh what? We laughed it off.
It didn’t take long for the creepiness to begin. Looking back, it’s interesting how forgiving we were.
Part 2: The fun part
Next, we had a little fun. Allo wasn’t all about the suggestions.
We tried some of the stickers — many of which are objectively cute by the way.
In retrospect, we have always used a lot of stickers and emojis. Perhaps because, they represent a form of communication that has never existed before — to express ourselves visually — with little to no effort on the part of the user.
This was the exploration phase. We tried having an entire conversation using just auto-replies and emojis — just to see how that would turn out.
It was fun for a bit and quickly got old.
Part 3: The creepy sets in
The moment we started to talk about something — something other than exploring the application itself — the creepiness of the auto-replies hit us hard.
When the recommendations were completely off base — they were sometimes funny but mostly annoying. When it got them right, they felt like judgements — “this what the AI thinks you should be saying.” Most of all, we had no idea whether that response was from the AI or from the person.
We tried two things:
1. We tried to turn them off. Allo doesn't give you that option.
2. We promised not to use them. But that still made us uneasy.
In the end, the constant reminder that you might be talking to a bot and not my friend was something neither of us was okay with. And just like that, our little experiment with Allo was over.
Some Final Commentary
Anish: Looking back at this entire exchange, almost a month after it happened, has been interesting. As a designer, there are a bunch of lessons to be learnt here.
Stella: I, too, learned a few things from our weird Allo experience. When I de-activated my Facebook account and attempted to leave social media, the idea of inauthentic experiences has really stayed with me.
Anish: Those of us who build products, are often so fixated on how to build it — to look at the smallest of interactions. We forget to ask the big questions. What are the social and cultural consequences of building that product? The big questions like “should you build this”? What was that Bill Buxton quote?
Ultimately, we are deluding ourselves if we think that the products that we design are the ‘things’ that we sell, rather than the individual, social and cultural experience that they engender, and the value and impact that they have. Design that ignores this is not worthy of the name.
— Bill Buxton
Stella: I’m taking a bit of what Anish mentioned, but I’m broadening his idea. Even before we consider if we should build it, I believe we need to start reconsidering why we’re ideating certain designs. Allo was a novel experience, but I would argue in the end, was it really needed? Was it a strange band-aid to our impersonal communication problems?
Anish: Being a bit of a nerd, I know that this “feature” started from Google’s Inbox. In the context of email, I believe smart replies worked quite well. I have used them a few times. The problem, however, is that chat is very different from e-mail: a) E-mails are longer, so smart replies are often a starting point. There is no way someone can get by just using smart replies. b) E-mails are more formal and in my case — used for work-based communication. I couldn’t care less if someone started their email using a recommendation from Google.
There are many different ways in which smart replies make sense — when you want to quickly respond to someone on the train or the bus or when your driving. One might argue the recommendations are required when you’re trying to design conversational UI’s.
We need to be careful when we take ideas as is, without asking ourselves — does it even make sense in this context?
Stella: I feel like with bots becoming smarter and wittier we actually lose a bit of what makes communicating so special. Perhaps I’m a bit too… anti-AI, but I’m interested in keeping human aspects with actual humans (think Westworld here). I want to actually come back to the roots of HCI and design: it’s ultimately used by humans and trying to remove humans from the equation can often create odd experiences — like this one.
I think I’m starting to see that designs that make me address something about myself usually leave lasting impressions on me. So, what if I started to design to encourage vulnerability instead of hiding behind design and technology to further make some false, self-created identity of myself? Allo really allowed me to hide behind the veil of self-automated conversations. I want something completely opposite of that.
Anish: The final idea thing I want to say is to be transparent. What finally drove us away wasn’t the recommendations — it was A) that we couldn’t tell if the reply I was getting was auto generated or something the person typed out. B) and we couldn’t find any way to turn it off. I also do want to give google some credit here. I liked the app — the stickers, the interface — it was all very well done. Designing an application from scratch is hard. Having the power of search inside a chat service has a ton of potential! If I could use it without the smart replies, I think I would.