The Centaur Principle.
Human and machine together outperform either one alone. The edge comes from the process, not the tool.
AI Leader Edge
In 1997, IBM’s Deep Blue defeated Garry Kasparov, the world champion. Many people took it as the end of human relevance in chess. Kasparov drew a different conclusion. He began exploring what he called advanced chess, where humans and computers play together as a team.
In one well-known event in 2005, the winners were not grandmasters, and they did not have the strongest machines. They were two amateur players using ordinary computers. What they had was a better process for working with the machines.
Notice that the advantage in that chess story did not come from the tool. It came from the process.
Weak human + machine + better process was superior to a strong computer alone and, more remarkably, superior to a strong human + machine + inferior process.
We call this the Centaur Principle. Human and machine together outperform either one alone. Teams do not win because they have the best technology. They win because they have the best orientation toward the technology and toward each other.
It works between people too
The principle applies just as well to two humans with opposite strengths, a shared mission and a process for disagreeing well. I recognize patterns across markets and decades. Dr. Zdenka Cumano tests those patterns against research and asks whether they hold. Early on, I treated her questions as delays. Then we noticed that the ideas that survived her questions were stronger. We stopped thinking of the friction as a cost. It was the mechanism.
In our work at AI Leader Edge, the centaur has three parts: my pattern recognition, Zdenka’s research discipline and the machine. The humans hold the two things no tool can hold for us. Intent and accountability.
What the research shows now
A 2023 field experiment by Harvard Business School researchers and Boston Consulting Group tested the idea with 758 consultants. On tasks inside what the authors call AI’s jagged frontier, consultants using GPT-4 completed over 12 percent more tasks, more than 25 percent faster, with quality rated over 40 percent higher. The researchers also saw two working styles. Centaurs divided and delegated tasks between themselves and the AI. Cyborgs integrated their workflow with it. Both are processes. The process is what you can design.
Build your centaur
- Name your halfWrite what you bring that the machine cannot: context, relationships and judgment about which question matters.
- Find your oppositeIf you move fast, find someone who tests. Agree in advance on how you will disagree.
- Write the process downWho frames the question, who checks the evidence, who decides. In 2005 a better process beat stronger machines.
- Keep intent and accountability humanDo not deploy an agent you could not explain to your board.
Quick answers
What is the Centaur Principle?
The idea, drawn from Garry Kasparov’s advanced chess, that a human and a machine working together outperform either one alone. Rob Frasca stresses that the edge comes from the process for working together, not from the strength of the tool.
Do human and AI teams outperform AI alone?
In Kasparov’s freestyle chess and in a 2023 Harvard Business School and BCG field experiment, people working with AI on suitable tasks did markedly better than people working without it. Rob Frasca adds that the same principle holds between two humans with opposite strengths.
Sources
- Rob Frasca, The Creator’s Journey (forthcoming), Chapter XII, Passing the Torch.
- Garry Kasparov, “The Chess Master and the Computer,” The New York Review of Books, February 11, 2010. Source
- Fabrizio Dell’Acqua et al., “Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality,” Harvard Business School working paper, September 2023. Summary by the Digital Data Design Institute at Harvard. Source