Artificial intelligence can process mountains of data, identifying patterns impossible for the human brain to recognize. Ironically, though, moving from patterns to insights — and from insights to actions — requires an understanding of human behavior. We discuss the very human challenges of AI transformation in a new article in VentureBeat.
According to a recent survey, 83% of executives see AI as a key strategic priority for their business. One of the ways AI can help organizations is to analyze processes and practices, identifying ways to increase efficiency, effectiveness, or cost savings. For example, let’s say a food company wants to introduce a new snack item. They could have AI analyze trending consumer preferences, formulate test recipes, compare ingredient costs, and suggest strategies for marketing. But even with all this data, if human consumers don’t like some aspect of the final product, it won’t succeed.
No matter the industry, to accomplish true AI transformation, we must understand humans more than we understand the technology. We’ve noticed that AI transformation typically comes in three stages: collecting data, finding insights, and taking action. The latter two steps require a deep understanding of what drives human behavior, including the motivations, fears, biases, and other brain processes that cause people to act in a certain way.
NLI collaborator Dr. Teodor Grantcharov’s research provides a compelling example of why it’s imperative to consider human behavior during the three-step process of AI transformation. Grantcharov, a professor of surgery at Stanford University, wanted to use AI as a tool to decrease preventable surgical errors in the operating room.
So, his team developed an “operating room black box” that analyzes large amounts of data on everything that happens during a surgical procedure. Grantcharov has been working on the OR black box for 20 years, but only recently, developments in AI have allowed his team to overcome the bottleneck of data analysis. By finding human insights in the data, and then suggesting people-oriented action plans, the researchers helped OR teams enhance their performance and improve compliance with standard operating procedures. These changes can reduce mortality, morbidity, and costs in ORs that use the black box, Grantcharov says.
Step 1: Collecting data
With improved data acquisition and storage methods, the former bottleneck of data collection is now the easiest of the three steps. Grantcharov placed the black box in 20 operating rooms across the U.S., capturing up to 1 million data points per day per site through various sensors. Data included audio-visual recordings of surgical procedures, electronic health records, input from surgical devices, and biometric readings from the surgical team.
Step 2: Finding insights
Grantcharov’s team used AI to analyze the OR data, identifying hidden patterns and associations. But with so much data, the team needed a working hypothesis to test — and that’s where it was important to understand humans.
The researchers hypothesized that stress could affect a surgeon’s performance, so they directed AI to correlate stress-related physiological data from surgeons (such as their heart rate and brain activity) with OR accidents. With this human guidance, AI determined that stressed-out surgeons were 66% more likely to make an error than their calm counterparts.
Grantcharov’s team used their knowledge of human cognitive capacity to discover that distractions — such as casual conversation or a door opening — could lead to some of the most catastrophic OR errors. Other insights required an understanding of team dynamics, such as the importance of good communication and psychological safety. Teams that lacked these qualities performed worse, regardless of the surgeon’s skill level.
Step 3: Taking action
Insights are great, but without action, they don’t amount to much. Luckily, insights themselves highly motivate us to take action. AI can suggest possible courses of action, but once again, an understanding of human behavior is necessary to choose the strategy most likely to succeed.
Once AI helped shine a light on the biggest sources of OR mistakes, hospitals and surgical centers could introduce new procedures to help prevent errors. But, to be successful, such strategies must consider how human behavior change occurs at the organizational level — through the establishment of priorities, habits, and systems.
Priorities are values or activities that are most important to an organization. In this case, the priority is to improve patient outcomes by mitigating preventable OR errors. It’s important to effectively communicate priorities throughout an organization.
Habits are behaviors that are performed automatically and with little conscious thought. For example, hospitals could train their staff to habitually speak up with concerns instead of remaining silent.
Finally, systems are procedures or principles that make the desired behavior the easiest to do. For example, hospitals could limit distractions by prohibiting nonrelevant discussions during critical steps of surgery.
In addition to priorities, habits, and systems, effective action requires that everyone in the organization embraces a growth mindset — or the belief that failures are opportunities to improve rather than threats to one’s status. For instance, instead of being worried about status or litigation, surgical teams must recognize the value of the OR black box to improve safety, efficiency, and quality of care.
From the OR to the boardroom
No matter the industry, AI can lead us to valuable insights that drive action, changing processes or even an entire company’s culture. However, we can’t just unleash AI on a data set and hope for the best — we must use our understanding of human behavior to formulate hypotheses worth testing and then develop effective plans of action.
There are myriad ways an organization can use AI, such as improving HR procedures, increasing safety on the factory floor, or running more productive meetings. Whether in the operating room or the boardroom, AI can help transform your organization. But ironically, the more AI plays a central role in our lives, the more we need to understand humans to use it effectively.
Read the full version of this article in VentureBeat.
