Marry Data Science with User Research. Ethical Design Depends on It.

Design teams don’t set out to build products unethically.

And yet, unethical products get built.

Oftentimes, they’re backed by “perfect” data. The numbers are solid, the algorithms work—but there’s a vital disconnect that can lead us to do harm.

“Today we divorce people from their data,” Ovetta Sampson says, “and that gives companies license to forget about the people behind the data...It allows us to divorce ourselves from the responsibility of what that data can do.”

It’s high time we fix the broken marriage.

From her first experiences programming with her father on a Commodore 64, to her 18 years as a professional journalist, Ovetta has spent most of her life exploring the relationship between people and data. And she’s since become a champion of breaking down the walls between data scientists and design researchers.

Too often, data scientists view researchers as doing “soft work” with little rigor. And too often, researchers see data scientists as “numbers people” who don’t understand their users.

We sat down to ask Ovetta about extending the olive branch—and why doing so is the key to more responsible tech.

dscout: When did you first begin to notice the divide between data science and design research?

Ovetta: When I was 9 years old, my dad bought a Commodore 64 and spent nights learning how to program it. My dad worked a lot so I didn’t get to spend a lot of time with him—but I would go down in the basement while he was using Fortran and COBOL. I was like, “I want to learn that!” And he showed me how to program.

It gave me this really amazing connection to the technology that has always been seared into me. It taught me from the beginning to not divide my humanity from tech. Data, numbers, and people were always intertwined. I learned later, in journalism school, that data was always part of evidence. I was taught that if your mother says she loves you, get a second source.

But when I started speaking about my work at conferences, I met other researchers who were completely cut off from data science. One researcher told me that she didn’t even know where the data scientists were located in her building. That’s when it dawned on me that the way I see data and people—as a natural coupling—isn’t the way companies see them.

After speaking with researchers at some of the biggest tech firms in the world and hearing their stories of being separated from data scientists, I knew that companies were divorcing data from people just in the way they organized their product development teams and processes.

When I did my talk at Strive in Toronto, one of the first things I had the audience repeat is that "all data is created by people. And all people create data." And the reason why I had them repeat that is because I feel like today we’ve divorced people from their data. We’ve separated the two, which gives companies license to forget about the people behind the data. And that fuels some unethical practices we’ve fallen into.

What does the discrepancy between the data and what people say tell you?

That we're people and we're fallible—just like data is fallible. Just because that person says that it was a sunny day and the almanac says it was raining doesn’t mean that person is wrong.

Data has always been a kind of mirror for people and how they feel. Sure, people don't always tell the truth, but it's not like they're not telling their truth. They’re just telling it how they remember it.

It might be sunny because of what happened that day to that person. Maybe they’re associating the joy that happened that day with sunshine. Does that mean that they're giving you inaccurate information? Maybe. But if the almanac predicts that it's going to be rainy in five days and it's not, the almanac is still giving you accurate information. It was an accurate prediction at the time.

You have to blend data and the people together to get as close as possible to the truth. And as a researcher, I'm trying to find the alignment between those two and to learn how we can use that data point or information to make an experience with a person better.

So to really get at the truth, you can’t rely on “just data” or “just opinion.” Data and firsthand accounts are inextricably linked.

Exactly. When you start divide them, you get into trouble.

I had a revelation about this when I was working on a project to build a tablet app for seniors. I was sitting in the living room with this elderly woman who used a tablet and the internet to research her lineage. As she showed me a photograph of her family tree, I realized it featured one of America’s first databases: a slave roll auction list that included her grandmother.

So it dawned on me—that's a database. It's a list of "property" with data attached—how many pounds her grandmother weighed, how tall she was, the $400 it cost to buy her.

But that database is an example of divorcing data from a person's humanity. You forget you’re looking at people and you convince yourself you're just looking at numbers.

That’s a poignant example—and a haunting one at that. Why do you think we tend to keep making these mistakes?

Because it's easier to say, “I'm just the engineer” or ”I'm just the numbers guy.” It allows us to divorce ourselves from the responsibility of what that data can do to people.

Just take a look around at what’s happening in tech today. Company after company has to pay out for data misuse or data breaches. When you think of people in aggregate, it is very easy to forget that there's an individual behind every one of those data points. With breaches today, you hear that millions of records had been accessed. It's so nebulous, so abstract. You don't think about the people behind those numbers.

But that's the first thing a researcher thinks about. This is why you need data scientists and researchers together. They look at the world differently and that's a good thing.

When you divorce people from the numbers, then you think it's okay to divorce the data engineers from the researchers. And then you get products that end up harming people in a way that you didn't think that it would.

So it’s important for designers and data scientists to never forget about the people behind the numbers. But what’s something that UX researchers should keep in mind about data science?

UX researchers need to get over their “qual versus quant” dichotomy. That's a played-out war. We can't afford that. We can't afford to be “Sharks versus the Jets.”

For us to create human-centered products in an intelligent era, we have to have people who understand human behavior and motivation working alongside people building algorithms. To do that, we have to have UX researchers who understand how algorithms are built. We can no longer work in a silo.

I call it marriage. And like Audrey Hepburn said, “If I get married, I want to be very married.”

What do you think keeps user researchers from “proposing?” Are there common misconceptions that UX researchers have about data scientists?

There's a scene in Spider-Man: Homecoming where Spider-Man is trying to drive a car he’s never driven before. So his best friend gets into a chair with a computer and starts telling him how to drive.

And that’s the misconception. That data scientists are just the math people you just call in when you need to crunch numbers or do something technical. At IDEO, we thought of data scientists as algorithm designers. They're designers too—just as much as a design researcher or interactive designer. They need to understand research and the humanity behind the things that they're building.

Do you see any misconceptions that data scientists have about researchers?

That what we do isn't rigorous. The myth is that, "Oh, you just sit around and talk to people." At the end of a project, I have literally talked to about 80 people. That's a lot of people! And not only have I talked to those people, I’ve coded everything they said, I’ve found patterns, and I’ve added those patterns together to come up with a theory or insight.

That's a scientific understanding of building a hypothesis—no different than any other scientific hypothesis. It's just that my evidence is people. There's rigor in that. I'm trying to find discrepancies. I'm trying to find design energy. I'm coding all that together.

Oftentimes we'll see companies and organizations drunk on data. Do you have any practical strategies for these companies to help transform them and help them realize the value in marrying?

One of the things I really recommend is to encourage the individual side hustle. All of us have passion projects outside of the work we do. Often, if you add data scientists to the mix, your passionate project becomes injected with jet fuel.

For example, if one of your passion projects is DJing or music, then find a data scientist who is just as interested. They can help you build an AI engine to create new band names or even compose songs.

Working with data scientists outside of the work setting is incredibly effective. Just getting to know them as people and work on projects really creates a bond that you can take back to the workplace and start working on internal projects together.

I'm obsessed with synthesis and trying to accelerate it, so one of the things I started working on with a bunch of data scientists was a speech-to-text program to help us read transcripts and find patterns faster for researchers.

Now I'm working with a data scientist to help create a chatbot that can recognize fake news. So I’m using what I know about real news from my journalism background, and then using an AI to become a fake news detective.

It sounds like people should just be friends with one another.

Totally. Break down those barriers. It’s hard because both UX research and data science are locked in these mysterious boxes. A lot of people don't know what researchers do and a lot of people don't know what data scientists do, which tells me they have a lot in common.

Where do you think the fields of data science and UX are going into the future? Or where do you hope they will go?

I’m almost beyond thinking about the human-centered AI world, and I'm more in a human/machine cooperative world. That’s where we're doing a lot of affective computing and teaching sentient machines how to have a relationship with humans. Right now we have this slave/master framework where it’s like, “Car, do this for me.” But because of the dynamic nature of AI, we’re going to have to end this.

It used to be I would program my device and it would act on the programming. But now, it’s not just me and my device. It’s me, my device, my car, the other cars, the city camera, my iPod, and a host of agencies operating within the same space.

The framework that Western design has imposed on AI needs to be more community-oriented—a kind of network of design.

From what I see, we can't just focus on the user and the machine anymore. How the user and machine relate has to be a more cooperative relationship.

That means we’re going to have to teach computers and machines how to be more affective. More relational, and less reacting to what the user commands it to do. It's going to require more engagement of people in their machines because of the nuances that we're gonna get to in the future. It's not just going to be me and my device anymore.