GUEST:
We are living in a world of data overload. From behavioral analytics to customer preferences, businesses now have so much data at their fingertips that they’re unable to process and consume all of it in a meaningful way. This is where the magic of machine learning comes in. When applied to massive internal company datasets, machine learning technology can derive important insights and provide actionable recommendations and predictions at superhuman scale.
But as automation, machine learning, and artificial intelligence technologies continue to show up in our daily experiences, more and more users are asking questions. How can I trust machine learning-based recommendations? How do I know this prediction is accurate? Will this machine take my job?
Do you have an AI strategy — or hoping to get one? Check out VB Summit on October 23-24 in Berkeley, a high-level, invite-only AI event for business leaders.
As machine learning technology is still in its infancy, hesitations like these are valid. But machine learning, AI, and automation should be viewed as a benefit rather than a burden. In fact, Gartner recently predicted there will be 2 million new AI-related jobs created by 2020. The same report predicts AI will improve the productivity of many jobs and has the potential to enrich people’s careers, reimagine old tasks, and create new industries. However, for AI and machine learning to reach that point and make a meaningful impact, people need the opportunity to interact with the technology and learn to trust it over time.
The first step in building trust starts with the ways people experience the technology. Modern design needs to evolve to build user confidence through a graduated system of transparent recommendations and a partnership-like approach that leaves users in control. Just as it takes time to cultivate human relationships, relationship between humans and machines need to develop slowly. Here are three phases of design that will strengthen human confidence in machine learning, allowing it to reach its full potential.
Phase 1. Electing
The first step in building trust should highlight the smarts of machine learning while letting users retain full control of the outcome. Giving users the ability to review a variety of options and then control the final decision is an important first step in designing for a cohesive machine learning/user relationship.
One example of this kind of user design can be found in photo-tagging systems on social media apps like Instagram and Facebook. While the tool will make suggestions like “Click to tag this image as Jon Snow,” the user has control over who actually gets tagged in the photo. All technologies that use machine learning need to reflect this same level of transparency, putting people in control of the decision-making elements. This first phase is essential in allowing people to experience the benefits of machine learning while keeping them in the driver’s seat.
Phase 2. Learning
Once there’s been an opportunity to interact with the technology in a transparent way, the learning phase of the design experience can begin. In this phase of the photo-tagging example, machine learning can show off what it gathered from phase one, but apply it on a larger scale.
Here, a user might receive this prompt: “We found 15 photos of Jon Snow. Click to tag all photos.” While people may need to deselect one or two photos that were incorrectly identified as Jon Snow, this is machine learning’s way of, well… learning. Once users have experienced multiple interactions with the technology and made choices based on recommendations in the election phase, the technology will be able to learn from the input and update future recommendations accordingly. The learning phase allows users to see firsthand how the technology is responding and adjusting to their feedback.
Phase 3. Predicting
In this third phase, the technology will have built a pool of recorded and learned information from past interactions. Using this newly created repository of information, machine learning technology can contribute low risk, high confidence predictions and take associated actions. Going back to our photo tagging example, at this point in the design experience, the user will be asked for permission to tag all future photos of Jon Snow automatically. This phase of design automates low risk, everyday actions while still giving the end user final say.
Once users have experienced these three phases of design, a level of trust and transparency will have been established that gives machine learning the runway to truly take off. But training machine learning technology to be responsible for low risk, everyday tasks is only the beginning. As users increasingly trust machine learning technology, it can become a hugely beneficial tool that frees them up to focus on the most complicated projects while letting them make the high-impact business decisions that only a human can.
Linda Tong is vice president of innovation labs and product experience at AppDynamics, a company that connects app performance and customer experience
to business outcomes.
