Source: Lu & Zhang (2025) — “1 + 1 Bigger Than 2: Information, Humans, and Machines,” Information Systems Research
Human plus machine sounds like it should automatically be better.
After all, the human has judgment. The machine has scale. Put them together and surely we get something beautiful, like productivity with better lighting.
Except collaboration is not magic.
This paper focuses on information, humans, and machines. The title suggests the key idea nicely: one plus one can be bigger than two, but only when the information relationship works.
That is the part many AI implementations underestimate.
It is not enough to put a model beside a human and call it collaboration. The human needs to know what the model knows, what it does not know, and when its recommendation should matter. The machine may need human input to improve or contextualise decisions. If information flows badly, the team can become worse than either side alone.
Very relatable. Anyone who has worked in a team knows that adding more smart people does not guarantee a better outcome. Sometimes it just creates a longer meeting.
Human-machine collaboration has the same problem.
The value comes from complementarity. The human and machine need different useful information, and the workflow needs to let that information meet at the right moment.
If the machine only produces an answer without context, the human may overtrust or undertrust it. If the human ignores the machine’s signal, the model becomes decoration. Expensive decoration.
My takeaway: the future of human-AI collaboration is not simply about better algorithms. It is about better information design. What does each side know? What should each side reveal? When should the handoff happen? That is where one plus one might finally become bigger than two.