Emphasizing Humanity in a World of AI
“Software is eating the world” — Mark Andreessen.
Software is a driving force in the new economy, but how do we maintain the focus on humanity rather than software in the midst of never-ending Agile development? Peter Bull has a few thoughts about the importance of returning humans to the center of innovation and a case study to show how Human Centered Design and data science must inform each other for true innovation to happen.
What’s Going Wrong With AI?
Building AI and software-driven products and services is a driving force behind the world economy, but problems creep in when we focus not on the human element of development but development for its own sake. By releasing developments without considering the real outcomes, we run the risk of ignoring or even creating ethical quandaries. Consider:
- Strava — accidentally revealed the locations of secret army bases by mapping popular running routes
- Facebook — exposed users to data violations through Cambridge Analytica influence by creating information gathering algorithms designed to customize user experience.
- Microsoft — created a Twitter bot that spewed hate speech within days of interacting with the open internet and had to be shut down.
How Humans Get Involved
How can machines take over a process both effectively and ethically in the era of Software 2.0? Bull refers back to a classic L. Bruce Archer quote, “Design is what you do, not what you’ve done” to illustrate a fatal flaw in some developers’ thinking. Design comes from putting outcomes first rather than methods or outputs, a critical distinction to make for human-centered design.
If you’ve ever designed a dashboard no one looked at or accidentally created a solution with a negative impact, you’ll understand this concept. Human-centered design allows products and services to be more effective and more ethical. Let’s look at how this plays out in the real world.
Human Centered Design: A Case Study
Bull worked on a project designed to bring digital money services to a traditionally unbanked population in Kenya. Because they had no bank account associated with their exchanges, this cash economy couldn’t take advantage of traditional services like Venmo or PayPal. Bull embarked on a human-centered design project using real-world data to solve this issue. His primary concerns were:
- to build deep empathy
- to use creative problem solving
- to engage users to understand motivation and unmet needs.
This kind of qualitative insight builds truly viable solutions. Bull expressed this in a quick formula:
Desirable (human) + Viable (business) + Feasible (tech) = Good design (successful solution)
You can’t start with exciting tech or the business outcome, however. You start with a human focus. Here’s how Bull’s project went.
Phase 1: Inspiration
talking to users, brainstorming, generating ideas.
Around the beginning of the project, cell phone use exploded in the target region. What operators began to notice was no one used a subscription plan. Instead, they paid as they needed minutes. This translated to using those minutes as a form of currency where a customer would trade extra minutes for goods. Shopkeepers knew someone would come along and buy those minutes, creating a healthy flow of money through the shop or kiosk.
This began a series of both subscriber and agent solutions based on the real actions of how those two groups interacted with digital money.
Phase 2: Ideation
narrowing ideas + prototyping iteration
The subscriber concept was easy. What the team began to spend a lot of time on was incentivizing agents. They realized through data flow that money concentrated with the agent. Keeping subscribers coming to digital money would require better systems for the agents handling the transactions.
They began to test data sets and to observe that data in the field. First, they just watched transactions. Predictors of success began to emerge with the two-week mark being a critical turning point in a new agent’s success. They tested agent transactions through segmentation and designed tools that both intervened to train or modify agent behavior when failure seemed inevitable, and that allowed agents to view data themselves to make better decisions about incorporating digital money.
They also brainstormed ways to help build the critical trust between subscriber and agent that helped ensure success. One piece of essential data found by observing the field was the unique bond between the agent and users of the digital money system. A system was born.
Phase 3: Implementation
run pilots, test, test, test
These programs ran in the field, many with low-fi iterations. They asked users to do simple budget tasks using paper clips. They used different types of loyalty cards to help build relationships between customers and agents, settling on a unique form of lottery system where both subscribers and their agents could win money together.
What emerged from this human-centered design was a unique system of digitizing money that worked the way actual people worked in their own environments. Tech may seem far beyond the way the project culminated, but the result was a system that solved a real need. The data was the design inspiration, but it was the human focus (observation in the field) that led to true innovation.
What This Means for Data Scientists
Innovation is an iterative process. Data science, AI and software inform business decisions because there’s context involved. Without context, data is just data. HCD helps provide the sort of context that makes data valuable from a business standpoint.
(slide) Iterative process — understanding, data prep, modeling, evaluation, understanding, deployment (context for data in business) human-centered design focuses on context.
HCD strengthens data science through:
- humanizing data points
- mapping pain points and critical experiences in the customer journey
- designing interventions to alter behavior and subsequently business outcomes
Data Science strengthens HCD through:
- illuminating what people actually do (versus what they say)
- discovering the real factors that change outcomes
- defining targets for intervention and analyzing those outcomes
Peter Bull encourages you to think from a human standpoint in your data science and avoid those newbie mistakes. It’s important to follow what he calls the four critical components of Human Centered Design to help inform decisions and drive real business innovation:
- Observe data in the field
- Design with — not for — through prototype iteration
- Put outcomes — not methods/outputs- first
- Build consensus for how success is measured
Empathy, evidence, and iteration are the threads that bind HCD and data science, making them better together — Peter Bull
— — — — — — — —
Read more data science articles on OpenDataScience.com, including tutorials and guides from beginner to advanced levels! Subscribe to our weekly newsletter here and receive the latest news every Thursday.
