Skills of the Future: The Judgment Gap
- Aug 1
- 3 min read
Look at every surgical robot the FDA has cleared. A pattern jumps out.
Almost all of them sit at the lowest rung of autonomy. They do exactly what the surgeon tells them, moment to moment. A small number can execute a preplanned task on their own. Fewer still can adjust that plan in real time.
None, so far, can plan and carry out a full sequence of surgical steps without a human in the loop.
That gap between what robots can execute and what surgeons must still decide isn't a temporary limitation waiting to be engineered away. It's the actual shape of the next decade of surgery. Almost nobody is training for it directly.
Where the ladder actually stands today
Researchers have proposed a five-level framework for surgical autonomy. It runs from simple robot assistance up to full autonomy.
A recent systematic review looked at every surgical robot cleared by the FDA. The overwhelming majority still operate at Level 1. The robot assists. It makes no independent decisions.
A handful reach Level 2. They can execute one specific preprogrammed task.
The most advanced systems cleared so far touch Level 3, conditional autonomy. Here, a robot can generate a patient-specific plan and adjust it as the procedure unfolds.
Nothing on the market yet plans and executes a full sequence of surgical tasks on its own.
That's worth sitting with. The technology is moving fast in labs and research settings. But the frontier of what's actually cleared for patients is still firmly in the territory where a human makes every consequential call.
What survives at each level
The interesting question isn't when robots will reach Level 5. It's what kind of judgment remains necessary at every level along the way, including the ones we haven't reached yet.
At Level 2, task autonomy, the robot executes what it's told. But a surgeon still has to recognize when a task shouldn't be handed off at all. Not every step in a procedure is equally safe to automate. Knowing which ones aren't is a judgment call no system makes for you.
At Level 3, conditional autonomy, the robot can generate and adjust a plan using real-time information. Someone still has to decide whether that plan actually makes sense given context the system doesn't have. A surgeon who's seen five cases like this one knows things about this particular patient that no amount of imaging data captures.
Even at Level 4, where robots plan and execute full task sequences, the research consensus holds. The hardest part of surgery isn't dexterity. It's contextual understanding, the kind of judgment that comes from experience.
Human surgeons remain considerably better than robots at weighing that experience to make complex calls in ambiguous situations. That's not a temporary gap. It's the reason full autonomy keeps getting pushed further out, even as the technical capabilities improve.
The skill nobody is naming yet
Here's what training programs haven't caught up to.
As more technical execution moves to the robot, pure manual dexterity matters less as a differentiator. Knowing when to trust the machine, when to override it, and when not to hand off the task in the first place matters more.
That's a distinct skill from technical mastery. Residency curricula aren't built to teach it. For most of surgical history, technical mastery and judgment were trained together, inseparably, through repetition.
They're separating now.
A resident can be technically excellent and still be undertrained in the judgment calls a Level 3 system requires. Reading a machine-generated plan critically. Knowing its blind spots. Deciding when the context in front of you overrides what the data suggests.
This is the actual skill of the future. Not operating the robot. Knowing exactly where its judgment ends and yours has to begin.
Source: npj Digital Medicine
Related reading:
Comments