Ben Shneiderman wrote the essay I am posting below; I am the second author. This essay reflects a series of conversations Ben and I have been enjoying for the past year.
Ben is a Distinguished University Professor of Computer Science at the University of Maryland and a Member of the US National Academy of Engineering. His most recent book is The New ABCs of Research: Achieving Breakthrough Collaborations. Contact him at email@example.com.
Captain Sullenberger (Sully) is rightfully celebrated for his amazing skill as an expert pilot. When bird strikes shut down both engines, he managed to land his airliner safely in the Hudson River. Sullenberger’s remarkable performance was due to his long experience as a pilot, quick assessment of the possibilities, and skillful handling of the controls.
Every day expert nurses, firefighters, and parents also solve real problems that save lives and bring comfort to those they serve. Similarly, doctors, teachers, and plumbers use their experience to help others in meaningful ways. These experts act in difficult situations, in which information may be incorrect and incomplete, and in which existing guidance may be inadequate. Still, they press on by blending their experience in a large number of situations with what they know about the current problem, recognizing familiar patterns and identifying anomalies quickly enough to make decisions and take action. Experts also make mistakes, sometimes serious ones, but careful study of these mistakes can refine existing practices, improve training or checklists, and lead to better tools.
Experts are adept at using sophisticated tools, such as 3D fetal sonograms or air-traffic control systems, which increase their comprehension of what is happening and preserve their control. These tools can have high levels of automation, as long as the expert users can comprehend, predict, and control the actions of the tools. And for that to happen, designers need to keep in mind that their technologies and practices are aimed at supporting and strengthening skilled and expert users; otherwise the automation they introduce runs the risk of interfering with expertise and actually degrading performance, as has happened in some aircraft and medical device designs.
Unfortunately, too many designers seem oblivious to the skills and needs of the experts who will be using their creations. Our purpose in this essay is to offer some guidelines for constructing tools and automated aids that improve expert decision making rather than degrading it.
The expert programmers who created and then improved Google’s Translate feature deserve recognition for their contributions. They clearly understood that:
“Artificial intelligence is not about building a mind; it’s about the improvement of tools to solve problems.”
Although they were clear that they were building tools, many journalists mislabeled their work as another demonstration that humans could be replaced. From the 1947 stories of “giant electronic brains” these misleading reports about the role of technology have undermined the trust in human experts. Even the recent Artificial Intelligence-100 report from Stanford claimed that:
“the difference between an arithmetic calculator and a human brain is not one of kind, but of scale, speed, degree of autonomy, and generality.”
We disagree with this central statement. Human thinking is not calculation or reckoning. Human creativity is different from what neural networks and genetic algorithms produce. Even artificial intelligence experts reject the idea that their conference papers should be reviewed by machine learning programs. Ironically, these human experts who are devoted to creating “smart” and “intelligent” machines know that human experts are better at appreciating discoveries and innovations.
We believe that experts excel at least three admirable human traits:
1) Frontier thinking: Humans are capable of dealing with frontiers of knowledge, where incomplete and incorrect information cause confusion and where goal-setting is a key skill. Even in these challenging environments, experts can formulate and solve problems to create something new. Human experts often make astonishing breakthroughs, opening up fresh paths to surprising research destinations. They do more than optimize performance or recognize statistical patterns — they create wholly new products and services, discern distinctive categories, and see new kinds of relationships. Skill at the frontiers of knowledge and capacity to push past the frontier is what humans have always demonstrated. They produce unexpected patentable inventions, engage in debates to promote compelling causes, and form companies to deliver revolutionary products and services.
2) Social engagement: Humans are inherently social, a skill which they use to learn, build trust, help others, and ask for favors. Human discourse is different from asking Siri or Google or Alexa for information. Human exchanges inform each participant, clarify issues, and build agreements. Human collaboration has been a key to its astonishing successes throughout the existence of our species, from the complex collaborations around hunting, foraging, and community building to the modern teams who run companies and develop Wikipedia. Human leadership to inspire participation requires a combination of skills that run large companies, transform cities, and shape national governments. This leadership can also be malicious, corrupt, and war-like, but that is the dark side of being human. The bright side of human expertise shows meaningful respect, seeks genuine compromise, and nurtures trust. Humans are very skilled at building and using common ground to permit efficient communication, and then noticing when common ground has eroded and needs to be repaired.
3) Responsibility for their actions: Humans are responsible parties who deserve honors when they bring benefits and who should be accountable when their decisions lead to harm. Honest reporting and trusted investigation of errors promotes improvements that provide greater protections. Exceptional performances, such as Sullenberger’s, advance understanding, improve training, and push experts to perform at ever-higher levels of achievement. Responsibility clarifies product design since it encourages designers to give users control and provide investigators sufficient data to understand what happened. We accept responsibility, which permits others to trust us. We also engage in other trust-building activities: we try to be predictable; we warn others when our behavior may surprise them, we take unnecessary steps to assist others, and so forth.
As designers come to recognize these human traits, their designs will improve, leading to safer, more effective, and more successful technologies. The sooner journalists understand that excellence in technology design emerges when these human traits are supported rather than supplanted, the sooner their writings will celebrate the ways remarkable human skills can be enhanced by advanced systems.
We see a future in which the appreciation for human experts grows, even as designers create more powerful tools. The best of these tools will boost the performance of human experts using more of their creative capabilities. And for that to happen, the designers will need to find ways for letting all of us, not just the experts, comprehend, predict, and control the actions of our tools.