5by5: Conversational UI, Patrick Jones

Interview with Patrick Jones, Principal, AXD at IA Collaborative

By Thomas Brandenburg

As AI, conversational UI begs the question of design ethics. Investment in this space is coming from a tiny handful of stakeholders who are predominantly wealthy white men geographically bulwarked in Silicon Valley.

AI is going to touch every aspect of our lives, and will take human culture into completely new waters. It could be good. It could be really bad.

 
What language (keywords or phrases) do you like to use to explain Conversational UI to an audience who is not familiar with it?

Our collective idea of tasks-by-intelligent-machine are deeply enframed by long-standing representations of artificial intelligence in science fiction. We all recognize some common tropes: dropped articles, gerunds, gapped pronouns, and monospaced cadence. Even in famous examples like HAL, the computer speaks with flat affect despite being almost fully conversational in every other aspect of language.

We’re still trying to absorb and accept the uncanniness of artificial intelligence. It’s easy to think of conversational UI as being forever locked in a lower stratum of existence absent of emotion and the wonderful rhetorical nuance that comes with it. Siri and her silicon friends abide below, but they are striving for the higher realm.

Ironically, customer service interactions with a real person don’t necessarily feel natural either. Call center scripts are often marked by the same flat affect and tone as computer voices. Assuming that everybody has had this experience, I describe conversational UI as bot-chat that feels like chit-chat; speaking with an AI entity that feels like speaking to a friend, with all the subtext and linguistic in-betweenness that comes with the everyday uses of human language. It’s the opposite of SciFi Voice; you, the user, can convey a need in the terms that make sense to you, and expect it to be understood in those same terms, just as when speaking to a friend.

 

Conversational UI doesn’t act as a hyper-efficient way to avoid talking to another human in order to, say, buy a shirt or book a flight. Instead, it acts like a human when a real human isn’t available.

 


Given we’re entering a new era of computing, what do you consider to be the promise of Conversational UI? 

Until just recently, computing was not particularly designed to complement the experience of being a human—it was not socially and culturally engaged in the world. Conversational UI that allows us to get basic information quickly so we can interpret it in the context of our own situation is already commonplace; for example, shopping and ordering food—these are tasks marked by great convenience but which don’t really touch on human meaning.

The real promise is how conversational UI can respond in circumstances of complex need when another human is not available. For me, this comes down to anticipation grounded in empathy. When needs arise that go beyond mere tasks and post significant problems, our first instinct is to “talk to a real person.”

The potential for conversational UI to read between the lines and connect the dots, perhaps more accurately than a human, means that it could be the basis of human-centered technologies that can perform in the most sensitive situations—a traveler abroad in a megacity who feels lost; a farmer who suffers a heart attack in the back forty; a parent who needs to alter and reorganize their entire day because of an emergency. In these scenarios, conversational UI doesn’t act as a hyper-efficient way to avoid talking to another human in order to, say, buy a shirt or book a flight. Instead, it acts like a human when a real human isn’t available. This is how it will transcend mere transaction and intersect with the social and cultural nuances of human life.

 
Do you think there are any specific aspects that make it difficult to adopt conversational UI for end-to-end experiences.

I’m captivated by thinking of conversational UI as language, rather than a mode of speech. It is a manner of speech, yes, but simultaneously, conversational UI asks us to construct language in a way that feels separate, or perhaps distinct, from everyday grammar. This extends to phonology and inflection, and even seems strange relative to the most formal uses of language. Just watch the next time somebody nearby dictates a text message into a phone.

Right now there seems to be a large syntactic demand on how we engage with conversational UI but for a relatively small payoff. We tend to use bots in cases when we need to complete tasks using procedural logic by applying filters, such as making a restaurant reservation (OpenTable) or ordering shoes (Spring).

In person, we would never follow the same exchange of information and conversation flow. So the most analogous examples in human speech are cases when good information, unambiguous language, and efficiency to minimize possible alternative paths forward are required. Imagine if an EMT did not use procedural logic to treat a patient while rushing them to the hospital.

It’s not that we can’t design for conversational UI in a way that hides procedural logic. We probably can, especially as procedural logic gives way to procedural learning in ML applications. Considering the end-to-end experience, because these speech patterns are limited in existing interpersonal conversations to precision-intensive situations, using highly formal decision trees to work through mundane needs is a design constraint at best. The extent to which it poses difficulty for adoption, especially if a static UI can handle the same data inputs without extended conversation, remains to be seen.

 
Do you think that you can design interfaces to build trust?

Recently, interaction and visual designers have established design traits that convey trust by achieving a level of legitimacy through good design. I say “traits” because I don’t think these things have been codified as standards, as they have in non-digital interfaces like wayfinding. The trouble is that any website, app, or digital channel can be re-skinned to appear trustworthy on the surface—literally, the plane of the screen—regardless of what lies beneath. Credit Karma after its design refresh, which involved brighter, flat colors and the adoption of a Circular-like font, appears more trustworthy than the originally design platform, even if nothing changed in the backend. In a way, one couldn’t be blamed to conclude that all it takes is color and nice typography.

On the one hand, yes: we can design interfaces to build trust. On the other, we have to think beyond the interface and, therefore, beyond the screen. Ultimately, trust is secured by a seamless journey from need and initiation to outcome and impact.

Interfaces that build trust help us understand our behaviors beyond the interface, where impact is measured. If Credit Karma reports a decline in your credit score, it should help you understand why so you can change the way you spend using credit cards.

Or, if Dexcom can help you understand the health impact of daily habits through accessible data visualization, you can change your diet to manage a condition like diabetes. This is why design research is so important to conversational UI. Only by understanding user needs systemically can we design interactions that create the right impact in real life and in doing so, affirm user trust.

 
Tell us about the metrics that your team works toward to capture the value of conversational UI? How do you share it within your organization?

Our recent work has been focused on improving KPIs for natural language processing as our partners’ ML algorithms grow more mature. A great deal of our work is in the logistics space, so our goal is to define user intent immediately through image recognition and by anticipating likely adjacency pairs. In a way, ironically, the goal is to avoid conversation or, in more user-friendly terms, to fill in the gaps and point the user directly to the most practical, real-world solution.

Due to confidentiality, I can’t be more specific, but I’m guessing this is something every reader can relate to. Value is lost if the algorithm gets it wrong, so like everybody else, we’re trying to make it flawless and error-free.

 

Do you have any additional thoughts you would like to share with our readers?

As AI, conversational UI begs the question of design ethics. Investment in this space is coming from a tiny handful of stakeholders who are predominantly wealthy white men geographically bulwarked in Silicon Valley.

AI is going to touch every aspect of our lives, and will take human culture into completely new waters. It could be good. It could be really bad. Conversational UI is something magical, yes, but as designers we need to rigorously reflect on the function of technology in human experience, and take a similarly rigorous position on it. In innovating products, services, and experiences around conversational UI, how do we do so without creating new forms of unthinking, consumption, irrelevance, and waste?

 

related 5by5s

5by5: Conversational UI, Ilana Shalowitz

 

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s