Google’s chief decision woman is out.
Cassie Kozyrkov, who has served as the internet company’s chief decision scientist and helped pioneer the field of decision intelligence, is going solo and working on projects to help business leaders navigate the tricky waters of artificial intelligence.
As AI becomes more powerful and more prevalent across industries, Kozyrkov will launch her first LinkedIn course, publish a book, and give keynote speeches about how to make informed decisions. Her goal is to give leaders the tools to think about how they deploy AI, and to help the public hold AI decision-makers accountable for the choices that impact millions of people, she told Coins2Day.
She spent 10 years at Google, five of which as chief decision scientist. Among her responsibilities, she guided company leaders to make informed and responsible decisions regarding AI.
“I’ve always believed Google’s heart is in the right place,” Kozyrkov said . But it is a large company, and outsiders sometimes equated her personal opinions with Google’s stance on a topic. In her new role, she won’t have to worry about how her advocacy impacts a company she represents, she told Coins2Day.
AI is undergoing a massive period of growth, which has caused anxieties about the future for some. Top minds in the AI space recently warned it could end humanity as we know it. This point in time feels like an inflection point in the world of tech. It is essential to have leaders in place that are educated in decision-making and consumers that can hold them accountable, according to Kozyrkov.
Kozyrkov, who grew up in South Africa, received a bachelor’s degree in economics from the University of Chicago. She also has a master’s degree in mathematical statistics from North Carolina State University and a partially completed PhD in psychology and neuroscience from Duke University. Prior to working at Google, she spent 10 years as an independent data science consultant.
During Kozyrkov’s time as chief decision scientist, which began in 2018, Google’s AI division grew substantially. CEO Sundar Pichai unveiled Duplex, an add-on to Google Assistant that can make phone calls on behalf of a user, intended to help schedule appointments, restaurant reservations, and other engagements. Google has made leaps in generating text, images, and videos from prompts, and it is developing robots that can write their own code. It also released Bard, its large language model rivaling ChatGPT. Many of Google’s developments have raised ethical questions from employees and academics, which isn’t unlike what’s happening at other AI companies. Google didn’t respond to requests for comment.
Kozyrkov would not comment on decisions she helped make at Google because of her nondisclosure agreement, but it’s not difficult to think of areas where the company has faced difficult choices when it comes to AI. In building Bard, Google had to decide whether to scrape copyrighted information to train the AI model. A lawsuit filed against Google in July accuses the company of doing so. Google also had to decide at what point to release the technology to remain competitive with ChatGPT but not damage its reputation. It came under fire right after it published the Bard demo video in which the chatbot gave an incorrect answer.
Kozyrkov’s work revolves around the idea that individuals can make choices that affect a lot of people, and those at the top aren’t necessarily educated in the practice of decision-making. “It is easy to think of technology as autonomous,” she said. “But there are people behind that technology making very subjective decisions, with or without skill, to affect millions of lives.”
The best way to make a decision is something humans have long grappled with, and which continues to evolve. There’s Benjamin Franklin’s three-century-old pro/con model, but there are also more advanced ways to answer important questions, Kozyrkov said. While she is targeting business leaders, her methods can also be used to make other important life decisions, like where to go to college or whether to start a family.
Decision-makers should ask themselves: What would it take to change my mind? They should also use data, but prior to seeing it, set criteria for what they will do based on what the data says. This helps decision-makers avoid confirmation bias, or using data to confirm an opinion they already have. It is also helpful to document the process of coming to an important decision—including the information available at the time—to evaluate the quality of a choice after it is made, according to Kozyrkov.
