Error Signals and Adaptive Systems
- Apr 5
- 3 min read
Well-functioning systems have a quiet kind of elegance. Many moving parts coordinate without central control, small adjustments ripple through, and the whole stays stable even as conditions change. I remember when General McChrystal spoke to us about dynamic complexity at the Maneuver Captains Career course. It was one of the best talks I have heard.
It all resonated deeply with me. Ever since I read Peter Senge's "The Fifth Discipline," I have been fascinated with systems thinking and complexity science. And I have always been sensitive to, and frustrated by, malfunctioning systems.
Everyone has experienced it at some point. Things take longer than they should. Small problems compound. Plans shift at the last minute. People stay late, pick up slack, and push a little harder to deliver on time. The system continues to function, but only because the people inside it are absorbing the strain.
In mechanical systems, that strain is visible. Push an engine too hard and it overheats. Overload a processor and performance degrades. The limits are real, and exceeding them produces clear signals. Human systems are different.
The limits are softer. People adapt. They compensate. They work around problems instead of exposing them. The cost doesn’t disappear. It becomes harder to see.
And because the system still appears to function, the typical response is to keep pushing. Control what you can control, and fix what’s obviously broken. Optimize the parts that seem to lag behind, share lessons learned, and update the process. It feels like progress.
But then it happens again. And again.
The work becomes a moving window of local optimizations based on a snapshot of the environment at the time of the problem. Well-intentioned teams adjusting pieces without ever stabilizing the whole.
There’s a deeper issue underneath this. The cycle above isn’t just bad luck, it’s powered by how we think about systems. We tend to treat complex, adaptive systems as if they can be controlled from the top and fixed piece by piece. That leads to two predictable moves: we suppress raw error signals (for example, “don’t bring up a problem unless you have a solution”), and we favor immediate, local corrections over understanding the system as a whole. Both reduce uncertainty in the short term. Both degrade adaptation over time.
Learning, whether in a brain or an organization, depends on error signals. Something doesn’t go as expected, and the system updates. Without that signal, there’s nothing to correct.
Some of the most effective systems are built around the importance of error signals such as the "andon cord" in Toyota factories and the Aviation Safety Reporting System (ASRS) in aviation. Problems aren’t hidden. They’re surfaced early and often. Error isn’t treated as failure. It’s treated as information.
But many modern corporate environments select in the opposite direction.
They reward:
smooth performance
immediate results
outputs that appear correct
They penalize:
slowing down
uncertainty
work that hasn’t resolved yet
Over time, this changes behavior. Errors don’t disappear. They get suppressed. Signals get weaker. Problems become harder to detect, not easier to solve. The system starts to look stable, but it’s losing its ability to adapt. A flywheel in the wrong direction.
At the edge of the system, problems show up as friction. Something feels off. A process breaks. A result doesn’t make sense. But from that position, it’s not always clear what’s actually wrong.
Is the issue local? Or is it coming from somewhere upstream?
The people closest to the signal often can’t tell. So when the expectation becomes “don’t bring problems without solutions,” something predictable happens. Problems stop showing up. Not because they’ve been solved, but because the system has made it costly to surface them without certainty. Error signals get filtered out before they can be understood.
This is where leadership enters. Not as a source of answers, but as a function of regulation, like the balancing on a bicycle.
In an adaptive system, leadership shapes the environment that determines:
which signals are allowed to surface
how those signals are interpreted
how the system responds
If every signal triggers an immediate correction, the system becomes reactive and brittle. If signals are suppressed, the system becomes blind. Adaptation requires something in between.
Signals must be preserved, but responses must be integrated and interpreted in the context of the whole system rather than treated as isolated problems to fix. This is less like issuing instructions and more like balancing on a bicycle.
You don’t respond to every fluctuation with a discrete fix. You maintain conditions that allow the system to stabilize over time. Without that, the system doesn’t just make mistakes, It loses the ability to learn from them.
In complex adaptive systems, you get what you reward, not what you ask for.
