Reworked version (minimal intervention):

Habermas’s reflections on the links between knowledge and interests deserve far more attention than they will be given here. For present purposes, it is enough to note that in “Knowledge and Human Interests: A General Perspective” he writes: “Theory had educational and cultural implications not because it had freed knowledge from interest. To the contrary, it did so because it derived pseudonormative power from the concealment of its actual interest.”

For Habermas, interests are constitutive of knowledge. In the same paper, he distinguishes between forms of inquiry according to their guiding cognitive interests: the empirical-analytic sciences incorporate a technical cognitive interest; the historical-hermeneutic sciences a practical interest; and critically oriented sciences an emancipatory cognitive interest. Habermas situates his own project primarily within the latter. At the same time, he is clear—if not always explicit—that interests operate across all scales of knowledge production, even if such typologies are best understood as idealizations: simplified models that render complex processes intelligible by bracketing contingency, overlap, and distortion.

In his later work, Habermas gradually retreats from this reliance on ideal types. Concepts such as the ideal speech situation are increasingly set aside, partly because they risk pre-empting, concretizing, and standardizing what ought to remain open to contestation. In their place, he turns to looser formulations—most notably pragmatic presuppositions—which aim to remain sensitive to the situated, fallible, and historically embedded conditions under which communication and critique actually occur.

---

What this does (without changing your claim):

• Clarifies that pseudonormative power comes from concealment, not neutrality.

• Keeps your insistence that the knowledge–interest schema is heuristic, not ontological.

• Makes the later shift read less like a repudiation and more like a methodological softening—Habermas retaining the problem while loosening the architecture.