Is Translation Just Symbols or a Body in Action?

For decades, translation scholars believed that translators work with “representations”: internal words that match external concepts. Imagine having a mental map of every sentence to move from one language to another. This is the representationalist view, seeing translation as a puzzle of symbols.

In recent years, the scene has shifted. Newer, more flexible models introduced the idea that translation also depends on context, interpretation, and probabilities. Here comes Predictive Processing: the brain constantly anticipates what will happen, comparing predictions with real input. In translation, the translator uses these forecasts to pick the most fitting word.

But there’s an even more radical theory: enactivism. Enactivists claim there are no fixed internal representations; translation emerges from the translator’s bodily interaction with the text, context, and cultural norms. It’s a “self‑evidencing” process, meaning the translator demonstrates understanding through the act itself.

The bridge between the two views is the Free Energy Principle. This mathematical formula, used in neuroscience, explains how the brain minimizes uncertainty (or “free energy”) to adapt to the world. Both representationalists and enactivists can use it: the former to model symbolic predictions, the latter to describe bodily adaptation.

In short, translation is no longer just a symbol game, but a dynamic activity where mind, body, and environment collaborate to make sense of different words.

The Free Energy Principle (FEP), introduced by Karl Friston, explains how biological systems maintain internal order in a chaotic world. Simply put, the brain creates predictive models of the world and updates them to reduce the gap between what is expected and what is perceived, called “free energy.” When this gap is minimal, the system reaches an energy equilibrium.

Applied to translation, the FEP suggests that translators constantly anticipate future meaning based on cultural context, register, and authorial intent. This anticipatory process is modeled by hierarchical neural networks: higher levels handle semantic abstractions, while lower levels process phonetic and morphological details. Interaction between these levels generates a hierarchy of predictions that are continuously corrected by sensory feedback (reading, listening, audience interaction).

Enactivism, on the other hand, stresses the role of the body and environment. According to Active Inference theory, translators act to reduce uncertainty not only cognitively but also motorically: gestures, posture, and tone of voice shape text perception and word choice. Translational activity thus becomes a perception‑action loop, where bodily data (e.g., pen pressure, reading rhythm) feed the predictive model.

A practical application of these ideas is computer‑assisted translation (CAT) tools based on deep‑learning models. These systems implement the FEP by simulating free‑energy reduction: they generate probable translations and update predictions in real time based on user feedback. However, enactivism warns that the user interface must allow fluid bodily interaction; otherwise artificial “free energy”—frustration and translation errors—may be introduced.

In summary, integrating FEP, Predictive Processing, and enactivism provides a unified theoretical framework: translation is a dynamic active‑inference process where mind, body, and technology collaborate to minimize uncertainty and produce meaning.