I can’t believe it’s already been almost a month since CI2020 ended. I’ve been lax in posting with the inevitable work chaos and catch up following being away at a conference – my apologies.
I was very happy to see some old friends, cheer on a few more, meet a few people I’ve only known virtually, and of course meet a bunch of great folks for the first time. The quality of the talks was very high and I was pretty excited to see so many talks that felt fresh and new at the same time they felt like some “old school” speech interaction design insights.
I was originally planning on sharing a number of detailed notes on every presentation I saw, but these talks weren’t broadcast, so what good will that do you if you can’t actually go back and watch any of the talks?
No, instead I’ll riff on something I told a friend over email commenting on the conference. In the company of so many people who’d worked in speech for a long time, despite my talk being well received, I felt like I was just saying, “don’t be a dummy, just do speech right, you know how to do it,” to a room full of people who might equally have said back, “Right. Duh.” But I think people appreciated the “right” part, and taking a point of view on it, as well as coming from a place of acknowledging my own bias (ulp, accidentally in one of the bits I happened to have ad-libbed) and being willing to come from a place of sincerity as well as acknowledging imperfection but demanding excellence.
A common theme I took away from the most thoughtful and thorough presentations from Nandini Stocker, Anna Wichansky, Sneha Kannegati, Jayson Webb, Samrat Baul, Teresa O’Neill, and Wolf Paulus, among others, was a contrasting view to the Lean approach. Rather than throwing a Minimum Viable Product out there and letting users find out whether it works (emphasis on the Viable there being what a software builder can ship) I took away that there are a lot of people working on speech and language experiences insistent on customer need being met.
Not to say that any of those folks were saying we needed to build huge perfect applications the first time. On the contrary, the folks at Oracle sounded like they put together a pretty scrappy system for classifying utterances as well as posting expenses via chat in a very short time. To paraphrase Nandini at Flipkart, who talked about finding success building a design team by deemphasizing the “unicorn” multimodal designer who can do speech, IxD, visual design, and maybe a bit of scripting, our success in designing and building natural language systems will come from respecting the problem space in order to meet the user where they are.
The more we let people who are passionate about their individual craft, whether that’s data analysis, great writing, building psychographics, or creating beautiful visuals, the more people will contribute in the way they’re best able to be interested and excited about solving customer problems. We then find where we overlap with our teammates, and are better able to leverage and take advantage of different approaches and ways of knowing, toward a common goal. And because it takes a whole village of folks to make a robust language application, part of that challenge is accepting that speech is a team sport, and finding the best ways to be a team.
Thoughtful is a difficult thing to be at a lot of jobs, when we’re not encouraged to go deep on what we’re best at and show our strongest and truest selves. The idea of it being easy or simple to make great language experiences is persuasive, especially since there’s still lots of money to be made (and saved) in Actions, Skills and chatbots. When we know however that easy and simple isn’t usually how language work gets done, but is the result of care and pride in our work when we are thoughtful, it gets a lot easier to do that work. And when we value and appreciate that there are diverse ways of approaching and understanding customer needs we get better at communicating with each other, which can only help us hear and see our user needs the better.
It was very exciting to hear so many people talking about this idea. And to acknowledge my own talk on the bias of thinking you’re right and know what you’re talking about (slides here), I don’t think it’s only because I happen to agree with that point of view. Where I work now, it’s expected that leaders Are Right A Lot so I was especially excited to find this year at Conversational Interaction that people were happy to come right out and say what they thought was right, and show their work. I’m definitely excited to continue all of the conversations we started there and see what’s in store for next year.