IDEO is aiming to build human-centred AI
Exclusive: Ideo’s Plan To Stage An AI Revolution
The prominent design consultancy acquires Datascope, a Chicago-based data science company, to help merge machine learning and human-centered design.
BY DIANA BUDDS 6 MINUTE READ
Today, the design consultancy Ideo announced that it has acquired Datascope, a Chicago-based data science company. The acquisition is meant to help Ideo, one of the country’s most prominent firms, respond to demand from AI-powered clients and can be read as a harbinger of where the design industry is headed.
Tim Brown [Photo: courtesy Ideo]
“We have spent the past hundred, or thousand, years designing artifacts,” Ideo CEO Tim Brown says. “The things we designed were relatively dumb and all of the intelligence in the relationship between us and the artifact came from the human being. Algorithms and technology are taking on their own intelligence. That’s a fundamentally new design problem. We’re designing relationships now as opposed to designing artifacts. How does the traditional discipline of design, the new discipline of data science, and the new technologies of machine learning come together to form these relationships?”
Datascope’s 15-person team of data scientists will move into Ideo’s Chicago office. Founded in 2009 by Dean Malmgren and Mike Stringer, Datascope has consulted with Ideo over the past four years on a number of projects around the world, like improving safety on a national school bus system that transports over a million children and future of mobility concepts for Ford.
Datascope and Ideo have a new buzz phrase for their application of human-centered design to the world of algorithms and data science: “augmented intelligence.”
“Ultimately, we’re using the term ‘augmented intelligence’ to really focus on the fact that we’re extending and enhancing people’s capabilities through technology, as opposed to thinking of technology as a separate thing, or replacing people’s capabilities with technology,” Stringer says.
Algorithms have always been designed, but most often by engineers and computer programmers, not by designers who have been trained in the traditional sense to respond to the needs of people. Brown draws parallels to the birth of interaction design in the 1980s, a field which Ideo cofounder Bill Moggridge coined.
“Data and algorithms and machine learning—all of these things collectively together are a new medium for design,” Brown says. “They can be crafted and shaped by people who think about what the world should be like, who have the imagination to do that, and have the technical skills to actually build the things. I think the design industry is getting excited about [data science], as it realizes [data science] is really a different way of shaping the world.”
Mike Stringer (left) and Dean Malmgren (right) [Photo: courtesy Ideo]
With the acquisition, Ideo plans to further integrate data scientists into its project teams; for example, data scientists working side by side with design researchers, engineers, and interaction designers. This is all in the service of better serving the requests of clients, who are coming to Ideo with questions on what do do with all the data they have, according to Brown. It is also about solving what Brown thinks are the design problems of our generation: systems. (This systems-based approach influenced Ideo’s decision last year to sell a minority stake to a design collective owned by a Japanese holding company.)
“We do a pretty good job at designing artifacts,” Brown says. “We have some great chairs and mobile phones in the world, but we have a lot of systems that are not so great. Systems aren’t designed once and then left to run. They act much more like biology than machines and we need to make these systems super smart to evolve and learn–but they need to learn in collaboration with human beings.”
So while Ideo, as a design consultancy, is using data to improve the work it performs for clients, Datascope, as a company of engineers, sees ample opportunity for learning how to better design the practice of data science, from what type of raw data is collected, how it’s collected, how algorithms are designed, and so on.
“We’ve been talking from the perspective of design, and how [data science] influences design, but this is the edge of data science, too,” Stringer says. “Data scientists have an enormous amount to learn from human-centered design community. This is also the frontier for data science: to figure out how to apply the skills and abilities that data scientists typically have in service of people’s needs.”
We hear time and again about the negative–often damaging–effects of artificial intelligence. Since artificial intelligence and algorithms are based on mathematical theory and data science, many people–from the people who enlist the services of algorithms to big-data evangelists to the general public–assume that they’re objective. (Who do we trust more for recommendations on what to read? Facebook, or our real friends?) We presume that since AI is rooted in numbers, it’s trustworthy. In reality, these systems can be discriminatory, they can undermine our democracy, and they are so opaque that we don’t even know what they were designed to do.
In 2016, ProPublica published a story about how risk-assessment algorithms used to determine criminal sentencing and bail were racially biased and predicted risk incorrectly. It’s not that algorithms are inherently bad or biased; it’s that they’re mostly opaque, the data inputs are biased, or the people programming algorithms are unaware of the potential biases that are influencing the algorithms. To Stringer, this reinforces why there must be a stronger marriage between designers, in the traditional sense, and software engineers.
“When you see these negative examples of AI and data science in the press, we see them as opportunities for human-centered design,” Stringer says. “[The ProPublica story] is an example where a human-centered design approach is critical to explore what the negative outcomes could be in advance so we don’t end up perpetuating those biases [in the algorithm].”
Ideo has been the biggest proponent of the term “human-centered design” and has built its entire business on furthering the approach. The organization believes design is uniquely poised to shape “the world to meet human beings,” as Brown puts it, offering solutions to everything from climate change to social and economic inequality.
So far, though, human-centeredness hasn’t been able to rid technology of bias, and some designers believe “design thinking” has reached the end of its usefulness for solving systemic problems, like racial inequality. Will applying a human-centered approach to data science be able to solve problems in AI that it hasn’t been able to address in other disciplines? Naturally, Brown thinks so.
“[Human-centered design] hasn’t always worked in specific cases, but in the general case the quality of the products, services, and experiences we have today are far higher than they would otherwise had been without it,” Brown says. “Of course, on that journey sometimes designers, engineers, and business people get it wrong. But in the general case, we have far better products experiences than it would otherwise be. I think it’s the same case with AI. There will be specific cases where they get it wrong–partly because [the people developing AI] didn’t understand, partly because they were learning, partly because they asked the wrong questions at the beginning of the process–but I believe if we apply human-centered design in a masterful way, we will, in the general case, improve the quality of what this technology can do.”
While Ideo and Datascope weren’t able to share specifics details about upcoming projects, Brown mentioned that the future of mobility, the circular economy, and education are areas in which they continue to work and which require a strong integration between design and data science.
“Creating algorithms and systems that are in the service of people requires building humans into the loop,” Malmgren says. “It’s observing interactions and taking those learnings and reapplying them to how we’re thinking about algorithms. So it’s the constant loop of people, data, and algorithms and continuing to iterate that.”