systemic-effect

Home · Working notes · Exploratory notes · About

Models Without Feedback

Why They Explain States but Fail at Development and Prediction

Version: v1.0
Status: Working Note
Last updated: 2025-12-29

Abstract

This working note distinguishes between models that describe states and models that explain development. It argues that feedback is a minimum structural condition for modeling change over time. Models without feedback can explain relationships under fixed boundary conditions, but they fail when applied to questions of prediction, adaptation, or system evolution.

Starting Point

A scientific model is a structured representation of selected aspects of reality. Its purpose is not comprehensive realism, but suitability for a specific class of questions.

The decisive issue is therefore not whether a model is true, but:

For which tasks is the model structurally valid —
and for which is it not?

State Models and Development Models

Models can be grouped into two broad classes.

State Models

State models (static or quasi-static) describe relationships at a given point in time or under the assumption that boundary conditions remain constant (ceteris paribus).

They yield statements such as:

If X is higher, Y tends to be higher.
Under certain conditions, equilibrium Z is stable.

Such models are well suited for:

classification and diagnosis
identifying local directions of influence
comparing scenarios under fixed assumptions

They explain states.

Development Models

Development models describe how system states change over time because current outcomes influence future conditions.

They yield statements such as:

If X increases, Y changes — and this change feeds back into X.
Depending on feedback and thresholds, the system may shift regimes.

They explain processes.

Feedback as a Structural Requirement

Development implies that the state of a system at time t affects the state at time t + 1.

This requires feedback.

Feedback means that outputs become inputs — directly or mediated through actors, institutions, expectations, or norms.

Without feedback, a model remains single-shot: it computes an effect, but not how this effect alters the system itself.

Formally:

without feedback → fixed relations between variables → state description
with feedback → variables influence each other over time → development

This distinction is qualitative, not incremental.

Why Feedback-Free Models Fail at Prediction

Prediction claims more than correlation. It claims that:

If X is changed now, the system will evolve toward Y in the future.

This requires at least three elements.

System Reaction

In social, economic, and political systems, actors adapt. Prices, strategies, norms, and behaviors respond to interventions.

Models without feedback implicitly assume a passive environment.

Endogeneity and Path Dependence

Many decisive variables are generated by the system itself. Treating them as exogenous removes precisely those mechanisms that determine future trajectories.

Regime Shifts and Nonlinearity

Dynamic systems often change behavior once thresholds are crossed. Linear extrapolation fails where tipping points, delays, or saturation effects dominate.

Feedback is the mechanism through which such transitions arise.

As a result, feedback-free models tend to:

extrapolate past relations
underestimate adaptation and counter-reaction
misrepresent timing and escalation

They may appear precise while producing spurious accuracy.

Why Such Models Persist

The omission of feedback is rarely ignorance. It is a trade-off.

Mathematical tractability
Feedback introduces nonlinearity and complexity.
Data and identification limits
Mutual causation complicates inference.
Institutional incentives
Many environments reward clean assumptions and statistical clarity over structural realism.

Elegance is gained. Predictive capacity is sacrificed.

Core Claims

The argument can be summarized without polemics:

Models without feedback are state models.
They explain relationships under fixed boundary conditions, not their transformation.
Development requires feedback.
Without feedback, there is no endogenous dynamics.
For prediction in adaptive systems, feedback-free models are structurally inadequate.

This is a validity judgment, not a quality judgment.

Such models remain useful — but not for questions dominated by feedback effects.

A Practical Heuristic

A simple test:

If the subjects of a model respond to the model or intervention,
feedback is prognostically relevant.

Typical indicators include:

learning actors (markets, voters, organizations)
strategic adaptation
expectation and norm dynamics
network or scaling effects

Core Statement

Without feedback, models explain states.
With feedback, they explain development.
Predictive capacity begins where a system reacts to itself.

On the Use of Models

Much attention is devoted to how models are built. Far less to how they are used.

As a result, models designed for state analysis are routinely applied to prediction or policy questions without structural adaptation.

This is not a failure of modeling skill, but a gap in methodological education.

Before selecting a model, the class of question must be chosen. Using a model outside its question class is not approximation — it is misuse.

Status

This document is a working note.

It is intentionally incomplete and open to refinement. Its purpose is not to conclude a debate, but to provide a stable reference point for methodological clarity.

How to Cite

Wende, A. (2025).
Models Without Feedback: Why They Explain States but Fail at Development and Prediction.
Working Notes, systemic-effect.org. Version 1.0.
https://systemic-effect.org/working-notes/models-without-feedback

systemic-effect.org
Working Notes