Some problems get worse the harder you try to solve them. Not because of incorrect goals, but because trying to control causes resistance. Parts of mind fight themselves harder, couples argue more intensely, and communities polarize into warring memeplexes.

What makes this pattern so common is that it follows naturally from how systems process information. When something gets in the way of an outcome that's being steered towards, the obvious response is to try harder to compensate. But this creates self-reinforcing cycles where control attempts consume the resources needed for actual solutions.

The Alternative to Control

There's a more powerful alternative: Opening. It's no coincidence that many effective practices—from therapy techniques to nonviolent communication to modern mediation—all share this core motion. Instead of collapsing the action space to a single dimension and playing tug-of-war to control outcomes or models, systems intentionally shift to share and receive detailed information without pinning down the conclusion in advance or trying to directly shape other parties.

This is the difference between a command and an offer of useful context, or between keeping up the filters to stop information from preventing you from winning and being genuinely curious to learn from whoever seems to be fighting you.

This enables 'third way' solutions that weren't visible from within the control mindset, as the space of approaches is almost never unidimensional.

The Basic Pattern Through the Free Energy Principle

According to the Free Energy Principle (FEP), self-organizing systems work to minimize their variational free energy—roughly, the difference between what they predict and what they encounter. This manifests in two key strategies: changing internal models to better match the world (perceptual inference) or changing the world to better match models (active inference).

The former path, while often adaptive, can create self-reinforcing cycles of conflict when multiple systems are involved. Each system's active inference becomes another's prediction error, creating escalating attempts to control shared variables through action rather than updating models.

Concrete Examples

1. Interpersonal Communication

Consider a classic household tension: dirty dishes in the sink.

Control Mode: "You never do the dishes. You're lazy and expect me to clean up after you!"

In this framework:
- High precision allocated to existing internal models about the other person
- Active inference dominating over perceptual inference
- Narrow sampling of environmental states relevant to prediction
- Resources dedicated to changing the environment rather than updating models

Opening Mode: "I feel overwhelmed when I see a sink full of dishes. I value a clean kitchen and shared responsibility. Can we discuss how to manage household tasks?"

The information dynamics shift to:
- Lower precision on internal models about the other's motivations
- Higher precision on sensory/informational exchange
- Broader sampling of potentially relevant variables
- Resources allocated to model updating rather than environmental control

This isn't merely "being nicer"—it's a fundamentally different information processing strategy. Instead of fighting over variables in a simplified model, it enables discovery and integration of more complete models.

2. Emotional Processing

When we notice an "unwanted" emotion, our default response is often to try controlling it directly:

Control Mode: "No, we can't be tired now. This is important work. Just push through it."

This creates an internal conflict where:
- One subsystem attempts to override signals from another
- The suppressed subsystem typically amplifies its signals
- Both sides focus on the immediate conflict
- Resources deplete in the struggle rather than finding solutions

While FEP literature doesn't specifically talk about emotional processing, its concept of information boundaries ("Markov blankets") has been applied to partitioning the brain.

Direct control attempts may disrupt the normal information flow between subsystems. This disruption can increase prediction errors across the system, explaining why control often amplifies the very states it tries to suppress.

Opening Mode: "I notice there's fatigue here. Part of me feels urgent about this work. What's going on for both these aspects of my system?"

This creates space for:
- Information exchange across functional boundaries
- Each subsystem maintaining its information-processing integrity
- Joint consideration of multiple systemic needs
- Discovery of solutions beyond the original conflict framing

Our internal systems hold vital information for solving problems. When we try to silence or override them, we often cut off the very insights needed to find better solutions. This is confirmed by the often spectacular success of Opening techniques like Focusing and Internal Double Crux at resolving long-standing psychological blocks or self-defeating patterns.

3. Societal Polarization: Guns and Safety

This pattern extends to larger social systems. Consider the polarization around gun policy in the US through the control vs opening lens:

The Urban/Gun-Control perspective often directly experiences:
- Community gun violence impacts
- Density making stray bullets especially dangerous
- Quick police response making self-defense with guns less necessary
- Limited legitimate use cases in their context
- Statistical evidence about accident risks and suicide enablement

Their model includes mechanisms for how easy gun access threatens safety and community well-being.

The Rural/Pro-Gun perspective often directly experiences:
- Guns as practical tools (pest control, livestock protection)
- Long police response times increasing self-defense necessity
- Community norms around responsible gun handling
- Traditions of marksmanship and hunting
- Concerns about government overreach

Their model includes mechanisms for how gun restrictions could threaten safety and livelihood.

The Control Spiral

After shootings, urban groups attempt to control "gun availability," perceiving resistance as industry manipulation or callousness. This triggers counterpressure from rural groups who see essential tools being threatened. They respond by amplifying resistance and stockpiling, which urban groups register as confirming the danger, leading to calls for stronger restrictions.

Each cycle:
- Increases certainty in each group's model
- Reduces ability to hear the other side's actual concerns
- Shifts resources from understanding to fighting
- Makes integrative solutions increasingly invisible

This is the same pattern we've observed in interpersonal and internal examples, but at societal scale—each side attempting to control shared variables rather than update models based on the other's experience.

"But the variable is in the wrong place!"

A system that is optimizing a function of n variables, where the objective depends on a subset of size k<n, will often set the remaining unconstrained variables to extreme values; if one of those unconstrained variables is actually something we care about, the solution found may be highly undesirable.
- Stuart Russell

If you try to optimize over a system which is tracking and cares about things you're not aware of or mindful of, you will by default set things that system cares about to states which are extreme and undesirable to that system. The harder you optimize, the more extreme this will be. A better informed version of yourself would also likely not approve of those outcomes, especially if you value that system being effective, healthy, and cognitively flexible. Remaining open to information about what systems which seem to be resisting you are caring for is necessary to avoid both conflict and catastrophic failures of myopic optimization.

This doesn't speak against trying to shift variables, just against forcing through resistance from other systems you want to be healthy. Variables are often in the wrong place! It's just that other systems have information you don't, and if you meet resistance that's a signal the system you're meeting has something to offer your models, if you can create space for it to share that signal.

From Internal Double Crux, a strategy for resolving internal disagreements.

Why Is Control Mode So Common?

The ubiquity of this pattern suggests fundamental properties underlying information processing systems. While remaining agnostic about the full theoretical framework, we can identify several factors:

1. Local Computational Efficiency
- Control attempts require tracking fewer variables
- Simpler models (direct cause → effect)
- Immediate apparent error reduction
- Lower working memory requirements

2. Resource Management
- Systems have finite computational resources
- Control mode looks cheaper in the short term
- Creates illusion of effectiveness through immediate action

Control remains valuable and necessary for many situations—it's the foundation of agency and effective action. Problems arise primarily when multiple subsystems attempt to control the same variables in opposing directions, creating interference rather than cohesion.

The FEP framework offers additional insights that might help explain this pattern:

3. Inference Asymmetry
- Active inference (changing the world) may require less immediate model complexity than perceptual inference (updating complex models)
- When precision is high on prior beliefs, active inference becomes a dominant strategy
- The path of least resistance may favor action over belief updating

4. Informational Dynamics
- Systems attempting to minimize prediction error may:
- Sample environmental states more narrowly
- Allocate fewer resources to model exploration
- Prioritize immediate error reduction over long-term adaptation

5. Self-Reinforcing Nature
- One system's active inference creates prediction errors for interacting systems
- Other systems respond with their own active inference
- Resources increasingly allocated to action rather than model updating
- The scope of models narrows under resource constraints

Control mode appears to be a natural attractor state for interacting systems processing information under constraints. It's the path of least resistance—locally optimal but globally suboptimal.

Breaking Out of Control Mode

Recognizing control mode is one thing; shifting out of it is another. What enables this transition?

1. Meta-System Recognition
- A meta-system that can recognize these patterns
- Identification of narrowed information sampling
- Observation of resource depletion
- Detection of escalating control attempts

2. Resource Reallocation
- Temporarily increasing resources for broader information gathering
- Allowing for higher uncertainty in models
- Expanding information processing capacity

3. Creating Safety for Opening
- Establishing conditions where models can safely absorb unexpected information
- Enabling information exchange without immediate control responses
- Maintaining productive uncertainty while systems remain functional

4. Establishing Shared Variables
- Finding information carriers that translate across system boundaries
- Creating shared exploration spaces that respect functional integrity

Practical Techniques for Shifting Out of Control Mode

Recognizing and exiting control mode is valuable; practices which successfully shift out of it include:

In Emotional Processing

Curiosity Practice: "Let me get curious about what's actually here..." instead of immediately attempting to change the experience
Noticing Control Impulses: Recognize the impulse to control and deliberately shift to information-gathering mode
Multi-Channel Awareness: Track multiple information streams simultaneously instead of fixating on a single variable
Genuine Uncertainty: Maintain the state of "I don't know yet" rather than premature certainty

In Relationships

Non-Interpretive Observations: Share direct observations without layering interpretations
Non-Demanding Need Expression: Express needs without demanding specific implementations
Genuine Curiosity: Ask about others' experiences with actual interest in model-updating
Solution-Space Exploration: Hold space for creative options beyond the initial framing

The Evolution of Integration

The fundamental tension is that systems naturally default to control. This creates what we might call the Control Paradox:
- Systems seek control because it promises efficiency
- Control attempts disrupt information flow between systems
- Disrupted flow leads to cruder control attempts
- Creating massive hidden costs through deadweight loss

Yet within this wasteful dynamic lies the seed of its own transcendence. Any meta-system that can recognize control patterns, enable temporary reduction in control attempts, and facilitate information exchange will reduce costly conflict and improve system-wide functioning.

When systems shift from control to integration:
- Resources shift from controlling to sensing
- Information bandwidth expands across functional boundaries
- Models can update and expand
- New solutions emerge that weren't visible from within the control mindset

The very failure of control creates selective pressure for the emergence of integrative meta-systems.^[1] The massive deadweight loss of control conflicts provides the energy gradient that drives the evolution of better coordination patterns.

^{^}
I'll gesture to this but not lay out my reasoning in detail, as this post is already long enough, but I suspect a lot of readers who made it this far will be able to retrace the reasoning steps.