Contemporary political communication is often characterized using metaphors such as "echo chambers" or "filter bubbles". Despite their popularity, growing concerns about vague definitions are complemented by empirical evidence that contests their widespread existence. Yet, today's political media environment does offer ample opportunities for human and/or algorithmic selection processes
with undesirable outcomes.

To reconcile these two observations, I suggest a framework based on so-called feedback loops. It explains how processes can be re-inforcing, yet not lead to catastrophic consequences. This can solve the paradox that extrapolating from typical filter bubble and/or echo chamber models leads to the conclusion of full radicalization within short timeframes, which despite the presence of the technologies for years has not become omnipresent. I argue that typical characteristics of complex systems, such as non-linearity, the co-esisting of positive (reinforcing) and negative (dampening) feedback loops, as well as the existance of endogeneous forces can explain such seemingly contradictory observations.

To systematisize the study of such mechanisms, I suggest to distinguish between human feedback loops, algorithmic feedback loops, and interactions between those. I highlight three specific societal concerns that can benefit from a feedback-loop lens and discuss the spread of mis- and disinformation, the normalization of extreme content, and the radicalization of fringe groups.

The paper offers a series of propositions as well as suggestions for empirical approaches (in particular, the use of simulation, agent-based testing, experience sampling, and data donations), to enable building better theories of political communication in the digital society.